CA2081711C - Video analysis system for editing a recorded or broadcast, televised program and its use for post-production techniques, especially multilingual post-production techniques - Google Patents

Abstract

The invention relates to a video analysis system for editing a televised program, which is a succession of images of order k.
The system comprises a circuit (so) for measuring the time variance of the composition of the images, for establishing, time-wise, the composition of the images and for establishing an image difference signal and scenic-activity parameters of the broadcast program for a group of at least two successive images of order k, k-1. An analysis circuit (11) makes it possible, from the scenic-activity parameters (ias), to establish data (srdp) representative of the cutting-up into shots consisting of a group of p successive images of the broadcast, televised program.
Application to post-production techniques, especially multilingual post-production techniques.

Description

2 ~ ~ 1'~ ~.

VIDEO ANALYSIS SYSTEB FOR EDITING A BROADCAST OR RECORDED
TELEVISED PROGRAP~i AND ITS USE FOR POST-PRODUCTION.
ESPECIALLY MULTILINGUA~~TECHNIQUES
Improved quality television systems, and future high-definition television systems, provide, with respect to conventional SECAM or PAL television systems, not only an improvement in the quality of the video and audio-frequency signals broadcast but also a variety of exten-sions to the existing functions, or even the introduction of new services.
By way of example, it will be recalled that the D2-MAC/Packets broadcasting system, beyond the improve-ment in the quality of the images and of the sound, especially provides an extension of the capacity of the channels reserved for the transmission of program-accom-panying data, access control, subtitles, the possibility of simultaneously broadcasting several sound-track programs, fox the broadcasting of multilingual programs, and the introduction of an 16/9 ratio enlarged image farmat, this format being required as the basis of televised programs broadcast in high definition in the future.
Although the D2-MAC/Packets broadcasting system is currently operational, the adaptation of this system to high-definition television having to be operational in the next three to four years, the potential of this broadcasting system is however under-exploited, as the implementation of the aforementioned additional services involves carrying out, in a studio, complementary opera-tions for producing or for preparing programs, the technical feasibility of which, and the corresponding costs, have not been completely brought under control.
In the expectation of an evolution in video production techniques and the appearance of a significant associated production volume, cinematographic films constitute an important, if not the only, program ~~$~.~5 ~.~
resource capable of immediately exploiting the potential of the standards of these new television systems.
Moreover, as regards the qualitative aspect, both visual and sound-track, cinema films, in the most part, make it possible to reach, after transfer to a video medium, the optimum quality which the high-definition television broadcasting systems can provide. Furthermore, one major advantage of cinema films produced far viewing in a cinema hall is their image format, which lends itself easily to the preparation of television programs with the 16/9 enlarged format. However, if the cinema film is intended to be broadcast in 16/9 format by means of a television system ensuring a reception compatible with a conventional 4/3 image format using a panorama technique, that is to say by reduction of the corresponding lateral zones, a post-production operation has to be performed in a studio, before broadcasting, so as to establish the position of the refraining compatible with the 4/3 format for each image of the program.
Other functions may be implemented by using and supplementing operations which have already been accomplished for the use of cinema films in a cinema hall.
Within the scope of the compiling of multilingual programs, a saving in costs of making such programs can be obtained with cinema films by collecting, for the same film, the sound-track versions which have already been translated or dubbed in order to be used in various countries. The supplementary operation then consists in synchronizing the various adopted sound-track versions with a common video version as a reference image.
Within the scope of the compiling of subtitles, from the use of subtitled cinema film archives, a similar principle may be applied, according to standards established for television.
The operating procedures and the installations used by companies specializing in the various cinema or videc~film post-production operations have formed the 20~~.'~~~.

subject of numerous publications.
In general, as regards the automation of the post-production tools, it is possible to imagine the introduction of editing stations which enable a human operator to remotely control, from a central workstation, all the equipment involved in making the final product, such as video tape recorders, audio tape recorders, video/audio mixers, etc., to simulate the result of an editing operation before executing, listening to and viewing it by the operator, to store a list of editing decisions and to make, by selective assembly, a final product by automatically executing the stored editing instructions.
However, it will be noted that the automation of the aforementioned editing process is limited to remote control operations of equipment and to finally making the product when the editing decisions, established by the human operator, have been stored.
Of course, in general the post-production opera tion consists therefore in assembling prerecorded images and sound, from original film shootings ("rushes"), or in embellishing an existing program with supplementary information such as additional sound, subtitles, or assistance data, in order to compile a final program intended for television broadcasting.
An important and essential part of the work of the human operator performing the editing, consists in locating on the source media or original recordings, the image sequences, constituted by one or more images which will be used for producing the final product. In tele-vision, video tape recorders use, for this purpose, a time code which associates, with each recorded image, a unique instant in the form of hour, minute, second and image number. The aforementioned time code enables the play--back of the video tape recorders, for final recording or viewing, to be controlled. The sequences to be edited are thus identified by a start and end time-code.

In the particular case of a multilingual post-synchronization operation, a first analysis step is performed by the human operator so as to identify diver-gences between the editing guides or leaders - concatena-tion lists for the sequences of images - of the various dubbed versions and then to define, case by case, the solutions to be implemented in order to treat the inci-dents picked up. The final step is the production of the final medium, in accordance with the video-editing and sound-treatment instructions which will have been adopted by the human operator.
A first drawback of this operating procedure is, of course, the time spent by the human operator in order to discriminate and to pick up the information locating the image sequences, or shots, by viewing the original recordings. It will be recalled that a shot is a succession of images presenting, for an observer or a spectator, an analogy of scenic activity. lHoreover, it will be noted that for many post-production applications, the locating of the shots, or image sequences, is limited to picking up the start and end time-codes of the shot, which are characteristic of the latter, these values being stored for assembling the medium for the final program.
Another drawback of the aforementioned operating procedure relates to the accuracy of the locating, which accuracy may be affected not only by a reading error by the human operator, but also by the fact that, when a video tape recorder is reading at low speed or in dis-crate mode, image by image, which is necessary for viewing by the human operator, the precision of the time code associated with the current image is not always guaranteed. Such a drawback is liable to be particularly irksome when editing requires single-image accuracy, specially in the case of the 4/3, 16/9 panorama technique or refraining, for audio post-synchronization for example.
In the particular case of compiling a multi lingual program, by poet-synchrpnization of preexisting 2~~~.~~~.

dubbed versions, which are considered as original recordings, the current operating procedure relies on a first comparison of two monolingual versions in order to make a provisional bilingual version, and then in comparing the aforementioned bilingual version with another monolingual version in order to produce a trilingual version, and so on and so forth. A drawback of such an operating procedure is a loss of the recording quality of the final multilingual product or program, on account of deterioration in the recordings of the audio-or video-frequency data, which deterioration is generated on successive copies by the recording equipment.
The object of the present invention is to overcome the aforementioned drawbacks by the implementa tion of a video analysis system for editing a recorded or broadcast, televised program enabling the post-production operations to be performed semi-automatically, intervention by a human operator being virtually elimina-ted or reserved to a few special cases.
A subject of the present invention is the implementation of a real-time video analysis system for editing a recorded or broadcast, televised program, enabling, during the reception of a broadcast program or the play-back of a recorded program, an indicator of the time variations of the composition of the image signal to be picked up and the various image sequences or shots of the program in question to be identified, which enables the various shots, such as a '°cut" shot, lap dissolves or other shots of the program in question to be discriminated qualitatively.
A further subject of the present invention is the implementation of a real-time video analysis system fox editing a recorded or broadcast, televised program making it possible to identify video editing discrepancies which can arise between various recording versions of the same program, so as to establish a leader enabling a final program version, optimized by post-production operations, to be produced.

2~~~.°~~.~.

A further subject of the present invention is finally the use of a real-time video analysis system for editing a recorded or broadcast, televised program for the production, from a plurality of different monolingual versions, of a multilingual program by post-synchronization operations.
The video analysis system for editing a recorded or broadcast, televised program in the form a succession of images, each image being recorded or broadcast in the form of audio- and video-frequency data associated with the image of order k in question and of an associated time code representative of said image, which is the subject of the present invention, is noteworthy in that it comprises circuits for measuring the time variance of the composition of the images, by determination, between two successive images of order k-1, k, of an image difference signal and of corresponding parameters representative of the scenic activity of the recorded or broadcast program for a group of at least two successive images, of order k-l, k. Circuits for analyzing the recorded or broadcast program make it possible, from the difference signal and from the corresponding parameters representative of the scenic activity, to establish data representative of the cutting-up into shots, which consist of a group of p successive images, of the recorded or broadcast program.
The use of a video analysis system for editing a recorded or broadcast, televised program in the form of a succession of images, in accordance with the preceding declaration, is noteworthy in that, with a view to producing a composition of multilingual-narrative televised programs by post-synchronization of monolingual-narrative programs, attributes of common video images of the same program, this use consists in performing an analysis step for each monolingual program in order to establish the shot representative data, consisting of a group of pl successive images, where pl denotes, for the monolingual version 1 in question, the __ number of constituent images of a shot Pl in guestion, then in establishing a systematic correlation between the various shots of each monolingual version, these consisting of a predetermined number pl of successive images for each monolingual version and each monolingual narrative, this correlation enabling the time and/or cardinal data, in terms of the number of successive images pl which are minimum and maximum constituents of a given shot P1, enabling all the monolingual narratives in question to be carried. The data are then stored in vector form (Pml, P11, p11, ..., Plt, plt, .,., Pln, p1n), so as to constitute a frame for composing or editing the multilingual program in which, with the order of the shots Pml which can enter into the composition of the final multilingual program, there are associated, for each monolingual version l, 1 representing the order of the monolingual version in question, the order of the shot Plt, the number of successive images plt and the time data fox 'the start and finish of the shot Plt in question.
The video analysis system for editing a recorded or broadcast, televised program, which is the subject of the present invention, finds an application in all post-production techniques such as automation of the locating of changes in shots arising in a program for compiling and synchronizing various accompanying data for programs, refraining by panorama technigue, subtitling, colorization of original black and white films, automating the process for the post-synchronization of films for compiling multilingual television programs from dubbed original recordings. It may also be used in operations for checking duration of transmission broadcasts on parallel antenna recordings, for conformity of the specification of programing companies, for conformity of the broadcasting times of advertising slots or of political broadcasts during electoral campaigns.
A more detailed descra.ption of the 'video analysis system for editing a recorded or broadcast, televised _ 0 program, which is the subject of the present invention, will be given in the follocaing description and in the drawings in which:
- Figure 1 represents a block diagram of a video analysis system, which is the subject of the present invention, - Figure 2a represents a block diagram of a first embodiment, entirely using digital technology, of an element of the system as represented in Figure 1, - Figure 2b represents various histograms characteris-tic of the number of occurrences, for ~ne image, of the amplitude values of the video samples, - Figure 2c represents a block diagram of a second embodiment, using digital and analog technology, of the same element as in the case of Figure 2a, - Figure 3a represents a block diagram for production of circuits for histogram calculation and for quantification of the bank accounts, - Figure 3b represents, at its points 1 to 4, various diagrams relating to the establishment of the histogram implemented in Figure 3a, - Figure 3c represents various models for deciding the level of scenic activity, these levels being sub-divided into absence of activity, low activity, medium activity and high activity situations;
- Figure 4a represents various types of standard shots permitting, by comparison, the cutting into succes-sive shots of a broadcast program, these standard shots being characterized by the mode of variation of their scenic activity during a time interval or a corresponding given number of successive images, - Figure 4b represents, in an analog fashion, the structure of the representative signal of the cutting-up into shots of a recorded or broadcast, televised program, such as supplied by the system represented in Figure 1, which is the subject of the present invention.

Figure 5a represernts a block diagram of an installa-tion permitting the use of a video analysis sy stem in an application for composing multilingual-narra-tive televised programs by post-synchronization of monolingual--narrative programs, attributes of common video images of the same program, - Figure 5b represents a block diagram relating to the system for obtaining the shots of scenic activity, without a definitively established relationship, 7.0 - Figure 6a represents a comparison algorithm with a strict equality criterion for three consecutive shots of a first and a second monolingual version, - Figure 6b represents a comparison algorithm for three consecutive shots with a strict equality criterion for a shot of order N and close or very close equality criterion for the 2 surrounding shots, - Figure 6c represents a comparison algorithm with close or very close equality criterion for three consecutive shots, -- Figure 7 represents a flow chart of a program for.
putting the shots of a first and a second monolingual version into relationship in order to produce a bilingual version, - Figures 8a, 8b and 8c represent an advantageous variant for putting shots into coincidence or for synchronizing various monolingual linguistic versions.
A more detailed description of the video analysis system for editing a televised program, which is the subject of the present invention, will now be given in conjunction with Figure 1 and the following figures.
In general, it may be considered that the afore mentioned program is televised, recorded or broadcast, in the form of a succession of images, each image being recorded or broadcast in the form of. audio- and video-frequency data associated with the image of order k in question and of a time code, denoted by ct, associated 1p with and represewtative of this image. Figure 1 represents, purely by way of non-limiting example, the aforementioned program supplied to the system 1, which is the subject of the present invention, for example by a video tape recorder, denoted by M. The essential signals supplied by the video tape recorder are thus the audio-frequency signal, 'the video-frequency signal and the aforementioned time code ct.
Tn general, it may be considered that the afore mentioned signals are signals supplied in digital form, so as to simplify and condense the following description.
Thus, as represented in Figure 1, the system 1, which is the subject of the present invention, comprises a circuit 10 for measuring the time variance of the composition of the images, this circuit, in accordance with a particularly noteworthy aspect of the subject of the present invention, permitting the measurement of the aforementioned variance by determination, between two successive images of order k-1, k, of an image difference signal arid of corresponding parameters representing the scenic activity of the recorded or broadcast program for a group of at least 2 successive images of the aforemen-tioned order k-1, k.
Furthermore, as has also been represented in Figure 1, the system l, which is the subject of the present invention, comprises a circuit 11 for analyzing the recorded or broadcast program, making it possible, from the aforementioned image difference signal and essentially from the corresponding parameters represen tative of the scenic activity between two successive images, to establish data representative of the cutting-up into shots, these shots consisting of a group of p successive images, of the recorded or broadcast, televised program. Tn Figure 1, the data representative of the cutting-up into shots are denoted by srdp and indicate a corresponding signal representative of the cutting-up of the aforementioned program into shots.

Thus it will be understoad that the system, which is the subject of the present invention, makes it possible, from a measurement of the scenic activity of two consecutive images, the scenic activity being conner_ted with the relative movements from one .image to the next image of the objects constituting this image and, definitively, from the shot for observation by the viewer, thu6 to discriminate a distribution of successive const:.ituent shots of the aforementioned recorded or broadcast, televised program.
It will also be noted that, throughout the following description, the set of aforementioned digital signals consists either of digital signals transmitted by the data-link RUSES between the various constituent elements of the video analysis system for editing a recorded or broadcast, televised program, which is the subject of the present invention, or of corresponding computer files containing the digital data representative of the item of .information conveyed by the aforementioned signals.
A more detailed description of the circuit 10 for measuring the time variance of the composition of the images will be given in conjunction with Figure 2a.
The present description of the circuit 10 for measuring the time variance of the composition of the images, in accordance with 'the subject of the present invention, is given within the scope of the non-limiting example of a 4-Z-2 type digital video signal, transmitted with 625 lines in real time, s0 as to measure from it the time variance of the composition of the images conveyed by this video signal.
For a more detailed description of the shape of the aforemewtioned video signal, reference may usefully be made to the corresponding standard established in 1982 by the CCTR (International Consultative Committee for Radio Communications) and, in particular, to Recommendations 601 and 656 of volume SCI of the Notices and Reports of the CCIR.

In the aforementioned embodiment, the processing applied, in accordance with the subject of the present invention, is restricted to the digital data of the luminance signal of the active lines of the first frame of ear_h image, i.e. 288 lines of 720 samples coded over 8 bits at the rate of 25 images per second.
The aforementioned video signal, which is an analog signal, is denoted in Figure 2a by S(i,j,k).
As represented in Figure 2a, the circuit 10 advantageously includes circuits 100, 102 far staring the video component S(i,j,k) of the image signal, i indicat ing the index of the component sample in the active line of the image in question, j indicates the active line index in the image in question and k indicates the order number of the analyzed image in question.
It will be understood in particular that the circuit 10 comprises an analog-to-digital converter 100 enabling the analog video signal S(i,j,k) to be converted into a corresponding digital signal.
Furthermore, the circuit 10 for measuring the time variance of the composition of the images according to the invention also comprises circuits 101, 102 for storing the component S(i,j,k-1) of the image signal of the image of prior order (k-1) in question.
It may be pointed out, for example, purely by way of illustration, that the delay of an image, in order to stare the video component S(i,j,k-1) of the image signal relating to the image of prior order k-1, can be produced by an image-storing circuit placed in "thaw" mode, the "thaw" mode being the mode for real-time acquisition-restitut ion of a video signal, the delay of an image over the current image being due to internal processing operations.
As also represented in Figure 2a, the circuit 10 for measuring the time variance of the composition of 'the images includes a circuit, denoted by 102, far phasing the video signal S(i,j,k) relating to the image of order k and the video signal S(i,j,k-1) relating to the image of prior ardor k-1, this circuit 102 males it possible, following the storage of the aforea~ientioned video signals, to read these signals correspondingly, in a synchronous, but shifted manner but the single term of an image, that is to say by the value 1 in the parameter k relating to the order of the successive images.
Such a circuit 102 is a conventional type of circuit and will not be described in detail in the present description. In particular, it will be understood that tha aforementioned synchronous reading can be carried out by means of a unit 107, as represented in Figure 2a, the unit 107 enabling, from the signal relating to the video component S(i,j,k) supplied by the analog--to-digital converter circuit 100 and from an exchange synchronization signal, se, a control signal, scorn, to be supplied and the timing of the entire proces-sing of the aforementioned video-digital signals, as described in conjunction with Figure 2a, to be ensured.
As is furthermore represented in the aforementioned figure, the circuit 10 for measuring the time variance of 'the composition of the images comprises a circuit 103 for calculating the image difference signal from video component signals S{i,j,k) and S(i,j,k-1), this difference signal being denoted by Sd(i,j,k) and satisfying the relationship:
Sd(i,j,k) = A.[S(i,j,k)-S(i,7rk-1)) + Sm.
It will be noted that, in general, the calcula-tion of the image difference signal Sd{i,j,k) is per-formed for each sample of the image by digital processing.
Furthermore, the symbol S indicates, of course, the component of the signal processed, that is to say either the component of the luminance signal Y or the component relating to the color difference signals CR and CB.
In the aforementioned relationship, A indicates a coefficient, which can be taken to be equal to 0 > 5, and Sm indicates the average level of the permissible dynamic _, range of the aforementioned component S, the value of the parameter Sm being chosen to be equal to 0.5 when S
indicates the luminance signal Y and equal 'to the value 0 when S indicates a color difference signal CR or CB.
The coefficient A can either be a constant coefficient or a function applied to the dynamic range of the luminance signal, such as for example the absolute value of the difference (S(i,j,k)-S(i,j,k-1)~, As is furthermore represented in figure 2a, the circuit 10 for measuring the time variance of the com position of the images comprises, in cascade with the circuit 103 for calculating the image difference signal, a filtering circuit 104 enabling the high components of the spectrum of the difference video signal Sd(i,j,k) to be reduced by low-pass digital filtering.
It will be noted that this circuit 104 appears particularly useful in the case when the recorded or broadcast, televised program comes from a cinematographic film transferred to a video medium, the high components of the spectrum of the difference video signal., in this case, corresponding substantially to defects such as scratches or dust particles initially present on the cinematographic film, for example. The filtering circuit 104 also makes it possible to reduce the imperfections of prior video processing operations, such as analog-to-digital conversion, disparity in group propagation times between video components, noise coming from the video tape recorder, quantizing noise during digital coding, without the activity induced by significant inter-image movements being affected. The filtering circuit 104 supplies, at its output, a filtered difference signal denoted by Sd(i,j,k).
As if furthermore represented in figure 2a, the video analysis system, which is the subject of the present invention, includes, at the level of the circuit 10 for measuring the time variance of the composition of the images, a circuit 105 for calculating, for each image of order k, from the image difference si~~nal Sd{i,j,k) and more particularly, preferably as represented in Figure 2a, from the filtered difference sigrcal ~d(i,j,k), a histogram, denoted by HSd(k), of the number of occur-rences of each amplitude J.evel of the image difference signal, or of the filtered image difference signal supplied. by the filtering circuit 104.
The circuit 105 for calculating the aforemen-tioned histogram is next followed by a circuit f.or calr_ulating corresponding scenic-activity parameters by comparison of the distribution of the spectrum of each histogram HSd(k) relating to each succession of images k-1, k, with a plurality of distributions or reference models representative of corresponding scenic: activity.
In the rest of the description, the signal representative of the scenic-activity parameters, con sisting of a scenic-activity indicator, is denoted by ias.
It will be considered that, in general, in accordance with the subject of the present invention, the analysis of the image difference signal or of the fil-tered image difference signal Sd(i,j,k) is carried out by the formation of the histogram of the distribution of the video levels for the aforementioned image difference signal by calculating, for each video level, the number of samples which have this level in the image in ques-tion.
For an aforementioned digital video signal, of the 4-2-2 type, using a coding over 9 bits, there are 256 different values possible for each coded sample. It will be recalled, of course, that the image difference signal Sd(i,j,k) or the signal Sd(i,j,k) indicates, on 'the one hand, the luminance signal and, on the other hand, the two color difference signals mentioned previously in the description.
For a digital level n of a coded sample, n being able to assume any value lying between 0 and TV = 255, the , ha.stogram HSd(k) for the pair of images o:~ order k, k-1 is obtained by calculating, for each value of n, a number ~~ :~. °~ ~~ y equal to the number of times when the coding of a sample of the filtered image difference signal Sd(i,j,k) has the value n. Of course, the sum of the 256 coefficients or numbers thus calculated is equal to the total number of samples of the corresponding image difference component on an image in question.
In the case in which the composition of the image does not vary, the case of a fixed image, each sample of the image difference signal is equal to the value Sm given in the previous relationship 1. °fhis characterizes the histogram of the aforementioned distribution of the amplitude levels by a peak for the level associated with the value Sm, all the samples of the image, and by a zero value for the other video amplitude levels.
In the case in which the composition of the image varies slightly, a large proportion of the samples of the image difference signal remains equal to Sm, which is manifested in the histogram by a broadening of the peak centered on the value Sm.
Of course, 'the greater this variance, the more the number of samples equal to Sm decreases and the more the aforementioned peak subsides.
In the case in which, on the contrary, a rapid break arises in the composition of the images, a break such as for example during a cut shot change, the distribution of the video amplitude levels of the image difference signal is random and therefore does not promote the appearance of a peak centered on the inter-mediate value Sm.
The histograms relating to the aforementioned case are represented at points 1, 2 and 3 in Figure 2b.
The aforementioned histograms correspond to histograms of the amplitude levels of the luminance image difference signal, the axis of the abscissas being graduated in amplitude level from 0 to 255, as described previously, and the vertical axis in a logarithmic scale from 100 to 105.

-~ 17 -On the basis of the interpretation of the aforementioned video level histograms of the image difference signal, the circuit 106 enables a corresponding processing operation to be performed in order to synthesize such an item of information.
The principle adopted for carrying out the aforementioned synthesis is to compare the histogram curves with preestablished models, the range of the amplitude levels of the video signal being subdivided into several intervals and a summation o~ the level counting operations being performed in order to integrate the histogram over the aforementioned zones. ' For each level range or interval, the total obtained is next compared with a scale so as to determine a scenic-activity indicator as will be described subse quently in the description.
As regards the embodiment in Figure 2a, which embodiment uses digital technology, the presence will furthermore be noted of a circuit 108 which enables controlling preselection signals to be generated from, on the one hand, the signal of the video component of two successive images S(i,j,k), S(i,j,k-1) of the filtered image difference signal Sd(i,j,k) supplied by the filter 104, this circuit 108 enabling a digital controlling signal, denoted by scoot, to be supplied, with a view to carrying out tests on this signal, for example. Further-more, a digital-to-analog converter 109 is provided which makes it possible, from the digital controlling signal scent, to supply a corresponding analog signal enabling the aforementioned control operations to be performed in analog mode.
The implementation of these control signals will not be described as it is not essential for the purpose of the present invention.
It will be noted that the embodiment in Figure 2a can also be replaced by an embodiment using digital-analog hybrid technology. In such a case, as represented in Figure 2c, the same elements carry the same references as in the case of Figure 2a. Lt will simply be noted that, following the storage of the image of prior row k-1, by the agency of the circuit 101, there can be carried out by means of two circuits, the digital-to-analog conversion circuit lOla, 101b, a conversion into analog signals, on the one hand, of the digital-video signal S(i,j,k) supplied by the analog-to-digital converter 100, and, on the other hand, of the signal of prior order k-1, S ( i, j,k-1 ) supplied by the delay circuit 101. The signals supplied in analog form by the converters 101a and lOlb are then supplied to an analog phasing circuit 102, to an analog subtraction circuit 103 and to an analog filtering circuit 104. An analog-to-digital conversion circuit 104a makes it possible, from the filtered image difference signal supplied by the filtering circuit 104, to pass back into a digital signal, the circuits 105 and 106 then being identical to those in Figure 2a. Finally, it will be noted that the circuits for preselecting the controlling signal can be produced in the form of two circuits 108a and 108b, each enabling the control signal scont to be produced in digital mode and in analog mode respectively.
A more detailed description of the circuit 105 for calculating histograms and of the circuit 106 for calculating the corresponding scenic-activity parameters, that is to say scenic-activity .i_ndicators, for each image of order k or pair of images of order k and k-1, respec-tively, will be given in conjunction with Figure 2c. Of course, since Figures 105 and 106 are identical in the embodiments in Figures 2a and 2c, the embodiment in Figure 3a will be able to be used in one or other of the aforementioned embodiments.
As represented in figure 3a, the circuit 105 for calculating, for an image of order k from the image difference signal or from the filtered image difference signal ~d(i,j,k), a histogram HSd{k) advantageously includes an input port 1050 for the filtered difference signal Sd(i,j,k) connected to an input interface circuit, i~ ~ d' ._. 1 Q ...
denoted by 1051, this input interFace aircui_t receiving the aforementioned filtered difference signal, and supplying time reference signals srt, video data signals sdv and clock signals, denoted by clk. ' Furthermare, a circuit 1053 generating service signals is provided, this circuit receiving, on the one hand, the time reference signals srt and on the other hand, the clock signals clk, and supplying service signals ss, which enable the video signals, and in particular the signal sdv supplied by the input interface circuit, to be processed.
Furthermore, a circuit 1052 for sorting the video data signals sdv is provided according to a histogram distribution law using video quantising levels.
This sorting circuit 1052 receives, on the one hand, the video data signals sdv, the clock signals clk, the service signals ss as well as sequencing signals slb supplied by a microprogramed sequencing unit 1061, which will be described subsequently in conjunction with the description of the circuit 106. The circuit for sorting the video data signals sdv supplies video data signals sorted by level sdvtn, according to a criterion for formation of the histogram which will be described subsequently in the description.
Furthermore, as represented in Figure 3a, the circuit 105 for calculating the histogram comprises a plurality of storage registers, denoted by 1054-0 to 1054-3, each register, intended to receive and to store the video data signals sorted by level or by level range, sdvtn, as a function of the levels of these signals, forming a bank of sorted-level values.
An adder circuit 1055 is provided so as to perform the down-count in each bank, that is to say in each register 1054-0 to 1054-3 of the number of occurren-ces of the amplitude levels of the sorted video signals.
This adder circuit 1055 is connected via a read-BUS data-link, denoted by SUS-L, and via a write--l3US data-link, denoted by BUS-E, to each of the registers forming a bank.
~1s regards the circuit 106 for calculating the scenic-activity parameters, this circuit, as represented in Figure 3a, comprises an input port 1060 and a micro-s programed sequencing unit 1061 connected to the afore-mentioned input port 1060. The microprogramed sequencing unit 1061 receives the time reference signal srt and supplies an access signal to the banks, that is to say to the registers 1054-0 to 1054'3, this access signal being denoted by slb arid corresponding to a signal for reading the aforementioned banks as will be described subse-quently in the description.
The circuit 106 also includes a buffer register, denoted by 1062, interconnected via the read BUS-L to each of the registers 1054-0 to 1054-3. This buffer register comprises, on the one hand, an auxiliary memory circuit enabling the bank read signal relating to each bank to be stored and, on the other hand, a conversion table. This conversion table supplies, from the bank read signal, a scenic-activity indicator signal for the pair of successive images of order k-1, k in question, as will be described subsequently in the description. However, it will be noticed that the aforementioned conversion table can be localized at the level of the microprogramed unit and can be chosen from Q preprogramed tables or established according to given criteria, self-analysis, forcing or the like. Finally, it will be noted that the circuit 106 includes a communication register 1063 which is connected, on the one hand, to the microprogramed sequencing unit 1061, to the buffer register 1062 via a BUS-type data-link, and, on the other hand, to a first port 1064 receiving the exchange-synchronization signal se or the control signal scorn and to a second port 1065 supplying a scenic-.activity indicator si.r~nal ias.
~.'he operation of 'the circuits 105 and 106, which carry out respectively the calculation of the histogram HSd(k) and the calculation of the sscenic-activity parameters, as represented in Figure 3a, will be '~i ~ .;~ w~
~~ ~ TJ .~. a zl explained in conjunction with Figures 3b and 3c.
A histogram of the leve7.s of the digital s.~.gnal representing the inter-image luminance differences, far example, is represented in Figure 3b at the point 1 of the latter. Ttie histogram represented at point 1 is a priori arbitrary.
At point 2 of the same Figure 3b is represented the distribution of the levels in the banks and, in the particular non°limiting example in Figure 3b, and of the embodiment in Figure 3a, 4 banks centered on the zero difference. It will be noted, for example, that the width of each bank depends on the number of banks and on a parameter assignment connected with the nature of the images. As will be noted at the point 2 of Figure 3b, to each sorted-level value bank are assigned values which are symmetrical with respect to the level of amplitude value corresponding to that of the average luminance Sm.
Thus, a first bank, bank 0, is assigned a continuous range of values which are symmetrical with respect to the average luminance amplitude level, this average luminance level being identified by Nl2 and n, the value of the coded samples, varying from 0 to N = 255 for a coding over 8 bits, as mentioned previously.
2 ranges of discrete values, which are symmetri cal with respect to the amplitude level of the average luminance N/2, are assigned to each of the other banks.
Thus, the spectrum of the histogram HSd(k) is expressed by bank count-down signals, denoted by sdb(0,1,2,3), each relating to the bank in question. Each bank count-down signal represents, in fact, the number of amplitude signals lying within the corresponding range of amplitude values associated with the bank in question.
Thus it will be observed that, at the point 2 in Figure 3b, a range of values, which are continuous and symmetrical with respect to the value N12, is assigned to the first bank, denoted by bank 0, two ranges of discrete values of symmetrical values are assigned to the second bank, denoted by bank 1, these two ranges being adjacent _ 22 _ to the central-value range associated caith the first bank, two ranges of separate values of symmetrical values are associated with the third bank, denoted by bank 2, these two ranges themselves being respectively adjacent to the two ranges of discrete values associated with the second bank, and 2 ranges of discrete values of symmetri cal values a:re associated with the fourth bank, denoted by bank 3, these two ranges themselves being respectively adjacent to the two ranges of discrete values associated with the third bank.
According to a particularly advawtageous aspect of the system, which is the subject of the present invention, as represented in Figure 3b at 'the point 3 of the latter, the bank count--down signals sdb(0,1,2,3), each representative of the number of occurrences NO of the amplitude signals corresponding to those of the range associated with the bank in question, are quantized onto a number NAS of scenic-activity levels, so as to define, for each bank, a quantized elementary scenic-activity level, nase(0,1,2,3). As represented at the point ~ in Figure 3b, the set of the quantized elementary scenic-activity level signals, nase(0,1,2,3), then forms the components of a vector in fact constituting the scenic-activity indicator, ias, for the pair of images of order k-1, k in question. Each component of the vector con-stituting the scenic-activity indicator, ias, has a value nase(0,1,2,3) referenced in terms of scenic-activity level value, NAS, including for example 5 levels, NAS
indicating the maximum scenic-activity value.
According to another particularly advantageous characteristic of the system, which is the subject of the present invention, the scenic-activity indicator, ias, as represented in Figure 3c, is representative of the absence of scenic activity, of a low scenic activity, of a medium scenic activity and of a high scenic activity.
As represented in Figure 3c, at the point 1 of the latter, the absence of scenic activity constitwting a corresponding decision model is defined by an c: .~. ~ ~. .~.

elementary scenic-activity level relating to the first bank nase(0), greater than the average value NAS/2 of the scenic-activity levels, nase(0) >_ NAS/2, the elementary scena..c-activity level relating to the second, third and fourth banks being substantially zero, i.e. nase(1,2,3) - 0.
As represented at the point 2 of Figure 3c, the low scenic activity, constituting a decision model of the scenic-activity level, is defined by an elementary scenic-activity level relating to the first bank, nase(0), greater than the average value NAS/2 of the scenic-activity levels, i.e. nase(Oj 9 NAS/2, the elementary scenic-activity level relating to the second bank being less than 1, nase(1) s 1, and the elementary scenic-activity level relating to the third and to the fourth banks being substantially zero, i.e. nase(2,3) - 0.
As represented in Figure 3c, at the point 3, the medium scenic activity is defined by an elementary scenic activity level relating to the first bank, greater than the average value NAS/2 of the scenic-activity levels, the elementary scenic-activity level relating to the second and to the third banks being less than or equal to 1, i.e. nase(1,2) s 1, and the elementary scenic-activity level relating to the fourth bank being substantially zero, i.e. nase(3) = 0.
As represented in Figure 3c, at the point 4, the high scenic activity is defined by an elementary scenic-activity level nase(0) relating to the first bank less than or equal to the average value NAS/2 of the scenic-activity levels, i.e. nase(O) s NAS/2, or by an elementary scenic-activity level, relating to the fourth bank, greater than or equal to the average level NAS/2 of the scenic-activity levels, i.e. nase(3)>_NAS/2.
The operation of the circuits 105 and 106, as represented in Figure 3a, can then be illustrated as follows ~ on transmission of the filtered image difference signal Sd(i,j,k) via the input port 1050, the input ~9 r ~-~ 1' _ ~4 _ interface 1051 successively supplies the various samples in order to give the video data signal sdv. on sequencing, by the microprogramed sequencing unit 1061, the circuit 1052 sorts the data and transmits the sorted video data signal, sdvtn, as a function of the amplitude of the samples of the signal sdv, to the register 1054-0 to 1054-3, with which the amplitude range constituting the bank, as defined previously, is associated. The adder 1055 then enables the down-counting, in each correspond-ins register and therefore in each bank, of the present and stored samples, the aforementioned adder generating the corresponding bank down-count signal, sdb(0,1,2,3), which is recorded in the register 1054-0 to 1054-3 associated with the aforementioned bank. At each new triggering by the microprograzned sequencing unit 1061, by the agency of the circuit for sorting the data and for managing the banks 1052, the bank dawn-count values stored in each register associated with each bank are read, the bank down-count signal, sdb(0,1,2,3), previously stored in the corresponding register, being transmitted via the read-BUS, BUS-i~, to the buffer register 1062.. Under the control of the microprogramed sequencing unit, on the one hand, a bank down-count signal, sdb(0,1,2,3), is stored by the aforementioned buffer register and then converted via the conversion table so as to generate, from the bank down-count signal, the scenic-activity indicator signal as described in conjunction with the points 3 and 4 in Figure 3b, and 1 to 4 in Figure 3c for a successive image pair of order k-l, k in question.
The circuit 11 for analyzing the recorded or broadcast program then permits the processing to continue.
As is furthermore represented in Figure 1, the circuit 11 for analyzing the recorded or broadcast program makes it possible, from the image difference signal and more particularly from the corresponding parameters representative of the scenic: activity, to ~' ~~ ~~ ~~ :~.
,~ ~.~ cj .~.
z~ -receive the exchange-synchronization signal, se, the scenic-activity indicator signal, denoted by ias, the method of obtaining which was described previously in the description, these two first signals being supplied by the circuit for measuring the time variance of the composition of the images carrying the reference 10, and the time code signal, denoted by ct, supplied for example by the video tape recorder 3~.
~n general, it will be understood that the circuit 11 for analyzing the recorded or broadcast program comprises a calculator circuit 110 which is interconnected with the circuit 10 for measuring the time variance of the composition of the images, and which receives the three signals, se, ias and ct, described previously.
The calculator circuit 110 enables the scenic-activity parameters to be stored and includes discrimination software making it possible to select the images of order k which are in question or the groups of p images, the scenic-activity parameters of which are greater than one or more predetermined threshold values.
The aforementioned software makes it possible to establish time and/or cardinal data representative of the cutting-up into shots, consisting of p successive constituent images of 'the recorded or broadcast, televised program. Of course, it will be understood that, in accordance with the standard for transmitting the images of z recorded or broadcast, televised program, such as the 4-2-2 standard for example, the aforemen-tioned time date may easily be converted into cardinal data corresponding to the image numbers for example.
An illustrative description of the ogerating mode of the aforementioned software will be given in conjunc-tion with Figures ~a and 4b.
The exchange-synchronization signal, se, enables data corresponding to the time code signal and to the scenic-activity signal, ias, to be acquired, at the image - ?5 ._ frec.Iue:rcy, by the calculator circuit 110.
The discrimination software makes it possible, using as discrimination criterion the scenic-activity indicator as described previously in conjunction with Figures 3b and 3c, for every image of order k or group of p images of order k-(p-l, k), to determine the editing points of this program, these editing points corresponding to two successive shots P-1,P. According to an advantageous embodiment, as represented in Figure 4a, the discrimination criterion is established for example with respect to 3 standard shots of noteworthy scenic activity.
A first standard shot of scenic activ.i.ty, such as the passage from one person to another person in a 25 dialogue situation, is indicated by cut shot, this type of shot having, as represented in Figure 4a, an abrupt time variation of scenic activity. In particular it will be understood that the abrupt time variation of scenic activity is characterized by the slope of this variation and not by its amplitude, over a small number of images, lying between 5 and 10 for example.
A second standard shot, indicated by lap-dissolve shot, has a scenic activity with a substantially linear time transition, 'the lap dissolve being characterized by a range of slope values for a number of images much greater than the scenic-activity transition corresponding to the aforementioned cut shot, that is with respect to the number of images of the order of 10, A third standard shot, indicated by overlay shot, as represented in Figure 4a, has a pulse-type time variation of scenic activity, the aforementioned standard shot then being characterized by a very rapid variation, less than or equal to 2 images in duration, of the scenic-activity level.
Tt will be understood that, in general, the aforementioned standard shots are defined by 'the value of the scenic--activity indicator and the value of scenic-activity which are predetermined, as described previously t in x-elation with Figure 3c and the comparison of the variance of 'this activity over a number of given images, as described previously.
Thus, the calculator circuit 110, by virtue of the use of the aforementioned software, enables a signal to be supplied which is representative, for the cor responding broadcast program, of its cutting-up into shots, this signal being denoted by srdp in Figure 1.
Each shot of row p is thus defined, as represented in Figure 4b, by a start time-code and an end time-code. Of course, it will be noted that, following the correspond-ing processing, the start time-code of a shot F
corresponds to the end time--code of the prior shot P-1.
Of course, 'the aforementioned signal can also include a code relating to the order of the predeterrnined shots and, if necessary, a code relating to the number of constituent images p of each shot. Of course, the signal srdp can then be stored and formed as a file, as has already been pointed out in the description.
It will be noted that the video analysis system for editing a recorded or broadcast, televised program, which is the subject of the present invention, previously described in the description, thus enables the set of time codes relating to shot changes arising in the editing of the aforementioned program to be used automatically, this program having been recorded on a video medium for example.
Thus, a large part of the laborious tasks of manual shot locating by viewing by a human operator can be delegated to the aforementioned system, which provides a saving in time and a saving in post-production costs for example.
63ith the aim of being able to be easily exploited in currently existing post-production installations, the video analysis system, which is the subject of the present invention, can be endowed with a computer com-munication interface circuit so as to permit the trans-mission of the aforementioned information ~~nd especially ~r,.~ >~r ~~. 9 A. _t za -of the files constituted by the signal representative of ctatting-up into shots to a wide range of equipment, such as microcomputers, editing stations, colorization stations or the like.
As will be described hereinbelow in the descrip-ti.on, the video analysis system, which is the subject of the present invention, may be used so as to establish the comparison between various different versions of the same program automatically.
Such a use may be performed by correlation between the leaders relating to the various versions.
Such a use will more particularly be described in conjunction with Figure 5a and the following figures.
The equipment necessary for the implementation of the aforementioned is represented in Figure 5a. Tt will be noted in particular that the aforementioned equipment is configured in the following manners several video tape recorders, denoted by M1 to ML, these video tape recorders being interconnected with the video analysis system for editing a televised program, denoted by 1, as previously described in conjunction with 'the present description. The video analysis system 1 is interconnected with a module 2 for virtual post-synchronization of the programs. The aforementioned use will more particularly be described in an application far the creation of multilingual, recorded or broadcast, televised programs from various monolingual versions. The combination of the video analysis system 1 and the module 2 for virtual post-synchronization of the programs is represented in Figure 5a by a single functional unit, denoted by 3, this unit being supposed to supply a virtual post-synchronization signal, denoted by spsv, which can be used directly by a program composition module ~, as will be described hereinbelow in the description.
zn general, and with a view 'to producing a composition of multilingual-narrative televised programs by post-composition of monolingual-narrative programs, C~ ~:! s~J -a 'F
w .. t7 .,:.. a ::
_ ~t~ ._ these monolingual narratives forming the attributes of common video images of the same program, the aforemen-tioned use consists in performing an analysis step of each monolingual program in order to establish the data representative of the cutting-up into snots consisting of a group of pl successive images, where pl indicates, for the monolingual version 1 in question, the number of constituent images of a corresponding shot P1. Of course, in Figure 5a the monolingual versions, V1, V1 to VL, are transmitted sequentially for example by each correspond-ing video tape recorder ril.
A systematic correlation between the various shots P1 of each monolingual version, each consisting of a predetermined number pl of successive images for each monolingual version V'1 in question, is carried out so as to establish a time coherence between all the shots Pl of each monolingual version and each monolingual narrative.
Of course it will be understood that this correlation enables the time and/or cardinal. data to be adopted, in terms of minimum and maximum number of successive images, pl, constituting a given shot, P1, enabling all the monolingual narratives in question to be carried. The data coming from the aforementioned correlation are then stored in vectar form (Pml, P11, pll, ..., Plt, plt, ..., Pln, pln) so as to constitute a multilingual-program editing or composition frame in which, in the order of each shot Pml which can go back into the composition of the final multilingual program are associated, for each monolingual version 1, 1 representing the order of the monolingual version in question, the order of the shot Plt, the number of successive images, plt, and the time data for the start and end of the shot Plt in question.
In Figure 5a, the multilingual-program composition frame is represented by the sa_gnal for virtual post-synchronization of the multilingual programs, which signal is denoted by spsv, which is just a sequential list of shots without a definitively established relationship, as will be described subse~~uently in the _ ~a .~
description.
The operating mode of the module 2 for virtual post-synchronization of the multilingual programs will be described in conjunction w:~.th Figure 5b.
Pzs pointed out in the aforementioned Figure 5b, the module 2 r_omprises a calculator, such as a microcom-puter for example, interconnected at the output of the circuit 11 for analyzing the recorded or broadcast program, as represented in Figures 1 and 5a.
The calculator, denoted by 20, comprises software developed so as to establish the relationships, shot to shot, between the various versions analyzed and thus to simplify the locating work fox editing the editing instructions of the final product.
The principle used for establishing these relationships between shots rests ozi a comparative analysis, version to version, of 'the distribution over time of the duration of the identified shots.
So as to remove the uncertainty which may arise as regards 'the determination of the relationships between shots, for example when consecutive shots have closely similar durations, a procedure can be used which consists in a real-time calculation of signatures relating to the composition of the video signal before and after the detection of a shot change. These signatures are in the form of digital data and are transmitted to the calculator 20 during a step for acquiring the analysis data. They can be used so as to automatically validate the result of the identification of the relationships, shot to shot, between versions, V1, in question.
The signal representative of the cutting-up operations into shots, srdp, relating to each monolingual version, Vl, especially comprises the series of time codes which corresponds to a shot-change detection, as represented previously in Figure 4b.
The processing software then makes it possible to use, in delayed-time for example, the files of cor-responding raw data of the signal sr-dp so as to automatically establish the editing divergences hetween the various versions of the same film.
The aforementioned software is composed of 3 phases:
- a preparatory phass during which, for each monolin-gual version V1, the raw--data file relating to the monolingual version V1 in question is processed, - a phase for putting the shots into relationship, during which phase the relationships, shot to shot between two versions, are established, - a final phase which supplies, for 'the set of mono-lingual versions V1 in question, which are in intended to form the multilingual version, the sequential list of the shots without a definitively established relationship between c<>rresponding monolingual versions. The sequential :list of the shots is indicated by the signal for virtual post-synchronization of the multilingual programs, spsv, the virtual character of the post-synchronization resulting from the fact that the relationship established between the shots is not necessarily definitive.
For each monolingual version, the preparatory phase consists in processing and structuring the file of raw data, srdp, obtained previously.
Thus, the aforementioned preparatory phase comprises a step for complementary filtering of the raw data, that is to say of the signal representative of the cutting-up into shots, "srdp, in order to generate a ~0 corresponding filtered signal representative of the cutting-up into shots, srdp. By means of the filtering operation in question, an operator has the possibility to define decision thresholds which are more restrictive than those used during the phase for acquiring the raw data. The file is processed in order to eliminate the time codes when the level of the scenic-activity indicators does not correspond to the new filtering criteria.

,5,7 Ci a _,~, ~tr-I! r~ ?S.
'!~ v. . ~ > . .

The aforementioned filtering step is followed by a step for calculatineJ the duration of the shots and of the transition shots by subtraction of the successive time codes, ct. The duration of the shots is denoted by DUR and the duration of the transition shots is denoted by 'fRAIVS .
This duration can be expressed in terms of number of images.
It will be noted that a rapid movement in a sequence of images can generate, for example, a succession of time codes, each spaced apart by one image and, consequently, a succession of shots of a duration equal to one image.
According to a characteristic of the use, which is the subject of the present invention, these short duration shots, that is to say, less than 10 images for example, are withdrawn from the initial list of the shots, that is to say from the signal srdp obtained as output of the analysis system, which is the subject of the invention, as represented in k'igure 1, in order to be considered as transition shots. Advantageously, this succession of shots constituting a transition shot is arbitrarily associated with the preceding shots, the duration of which is greater than 10 images.
Thus, each shot is identified, on the one hand, by its own duration DUR, and, on the other hand, by the duration of the transition shot associated with this shot, the duration of which is denoted by TRAMS.
So as to simplify the establishment of the relationships, shot to shot, between monolingual versions Vl, the use, which is the subject of the present inven tion, comprises a step for calculating and assigning, to the first shot of each monolingual version V1, a virtual tme code, TCIV, or arbitrary value and, by recurrence, for each subsequent successive shot Pl, a virtual time code, expressed in number of images, satisfying the relationships Tclv(P1) = TUIV(P1-1) + DuRtPI-ly ~~ TRArrs~Pl-1~.

~~ C.~ ~.

Tt will be noted that in this relationship, Pl indicates the shot number of the monolingual version 1 in question.
The aforementioned step is then followed by a step for creating a signal representative of the cutting-up into processed shots, this signal consisting, of course, of files, as mentioned previously in the description. This signal is indicated by srdpt. It will be noted that this signal, or file, allows the identification of each shot P1 and contains, in addition to the real time code, a shot number, the duration of the corresponding shot expressed for example in number of images, the duration of the associated transition shot and a virtual time code TCIV. A file representing the signal srdpt in an analog manner is given by the following table.
TABL$ 1 OUTPUT FILE OF THE PREPARATORY PHASE.
SHOT No. DUR TRAMS TCIV

. . . 10 807 A more detailed description of the establishment of the systematic correlation between the various shots of each monolingual version will now be given in con-junction with Figures 6a, 6b, 6c.
In general it will be considered 'that, for a number L or monolingual versions, the correlation is performed by group of 2 monolingual versions. This correlation between successive shots P1, P1+1, of two monolingual versions V1, V1+1, is performed by comparam five analysis of the duration and/or of the corresponding number of images pl, pl+1, for the shots tn question, or a .~_ of a linEar combination of these durations or of the numbers of successive images over a predetermined number of successive shots P1, P+1,1, ..., P+rl; P1+1, P+1,1+1, ..., P+r,l+1 of both monolingual versions V1, V1+1 in question.
Thus, for a shat of given duration of the version V1 for example, a shot of equal or approximately equal duration is sought in the other version Vl+1.
zn general, the step for establishing the afore mentioned correlation may consist, from the signal representative of the cutting-up into processed shots, srdpt, in performing a putting-into-relationship opera tion, shot by shot, by comparison of the duration DUR of the corresponding shots P1; P+r,l+1, of the two monolingual versions Vl, V1+1 in question.
This comparison is preferably performed by concatenation of elementary comparison steps based on a constraint criterion of equality of the shot durations DUR and/or the transition durations TRAMS of the transition shots, this equality constraint being more or less strict.
The quality of the results arising from the processing in the phase for putting the shots into relationship rests an the nature of the comparison algorithms and on their scheduling. Thus, a comparison algorithm which is limited to the simple equality of shot duration is capable of putting into relationship shots of strictly identical durations but of different contents.
Thus, at the start of the phase for putting shots into relationship, the high-equality-constraint com parison routines are nailed up.
The strongest equality canstraint is produced by a strict equality allocation of the shot, DUR, and transition shot, TRAMS, durations for three consecutive shots in each of the 'two versions Vl, V1+1.
A comparison algorithm with strict equality criterion for 3 consecutive shots is given in k'igure 6a for two versions, version 1 and version 2, corresponding C? :~ ~~
r~ ~ ~ .~. ~ _~_ ~~
m 35 to 1=1.
For the aforementioned version 1 and version 2, we consider:
TAa~ 2 Shot P-1 (previous shot) M-1 Shot P M
Shot P+1 (next shat) M+1 The putting of shot P of V1 into relationship with shot M of V2:
DUR (P-1) - DUR (M-1) AND

TRAMS (P-1) = TRAMS (M-1) AND

DUR (P) - DUR (M) AND

TRAMS (P) = TRAMS (M) AND

DUR ( P-1-1 - DUR (M+1 AND
) ) TRAMS (P+1) = TRAMS (M+1) Uonsequently, various algorithms may then by concatenated, such as those for equality allocation between two consecutive shots, equality between the sum of three consecutive shots and one shot, equality between the sum of two consecutive chats and one shat, and simple equality.
Furthermore, it will be noted that each of the comparison algorithms can be activated with a precision criterion of greater or lesser constraint:
- strict equality of DUR and TRAMS, - very close equality within 1 to 2 ~k, - ca.ose equality within 3 to 5 $m This precision criterion enables a fuzzy-logic decision criterion to be introduced> Examples of cam-par.ison algorithms over 3 consecutive shots are given in Figures 6a, 6b and 6c, Figure 6b corresponding to a strict equality criterion for the shot P and with a close or very close equality criterion for the two neighboring shots, whereas Figure 6b corresponds to a comparison algorithm with a close or very close equality criterion for 3 consecutive shots.
According to a further advantageous character-istic of the use, which is in accordance with the subject of the present invention, for each elementary comparison, the comparison is performed on a search window consti-tuting, for a shot P1 of order P of the version Vl, a reduced environment of s shots, Pl+1 to P+sl+1 of the version Vl+1.
The creation of this search window is justified ' by the fact that 2 shots capable of being put into relationship have an adjacent position in their respec-tive files. The shots located at the start of a file in .
ane version will also be located at the start of a file in the other version.
Thus, by virtue of the creation of the search window, the most probable shot is sought in a reduced environment of shots of the version 2 for any shot of the version 1. The search window is determined in value by the value of the virtual time code of the version 1, modulated by the width of the search window.
According to another particularly advantageous aspect of the use, which is the subject of the present invention, the putting of two shots, respectively Pl and P+sl+1, into relationship is immediately followed by a relative refraining of the shorter virtual time code TCIV
of the two monolingual versions Vl, V1+1 in question.
Furthermore, the virtual time code of the next shots of the version, the virtual time code of which has been re-updated, is also re-updated until the recognition of an ~5 already re-updated virtual time code.
It will be noted in particular that this virtual time code re-updating in the two monolingual versions being compared, Vl, V1+1, corresponds t.o a systematic film- or program--portion refraining, an example of which is given hereinbelow in the description. Finally, it will be noted that as the number of shots put into relation ship is increased, so the width of the search window is then reduced.
An example of systematic refraining of film portions is given hereinbelow in relation with Tables 3, 4 and 5t TAEi~ 3 BEFORE SHOTS RAVE BEEN PUT INTO RELATIONSHIP
14000 ~ ~ TCIV = 14500 TCIV = 15000 PL1V1 L1V2 ____ ____ TCIV = 24500 ~ ~ ~ ~ TCIV = 23000 ~L2V1 L1V2 The TCIV of the shot PL1V2 is in the search window (14,000, 16,000). PL1V2 is put into relationship with PL1V1. The PCIV are updated.

4 j ~' ( '~J s '~
. ....

TABLE
AFTER TI3E E'IRST PUTTING OF SHOTS INTO RELATIONSHIP.
TCIV = 15000 TCIV = 24500 _-_ ............. --~_ TCIV = 15000 ___ ....,........ ____! , --.- 23500 --- TCIV ~ 23500 ___ ~ ----The TCIV o~ the shot PL2V2 is in the search window (23,500, 25,500). PL2V2 is put into relationship with PL2V1. The TCIV are updated.
TAF3I~3 5 AFTER THE SECOND PUTTING OF SHOTS INTO RELATIONSHIP.
TCIV a 15000 TCIV = 24500 ___ ,............ ___ TCIV = 15000 --- ............. --- TCIV = 24500 ,............ ___ p s'.
JC~ _ The possible incidents specific to each version, cutting, et.c., require a relatively large search window when the first comparison algorithm is called up.
Experimental studies have enabled it to be established that a comparison between two versions of a closely similar total duration required an initial search window of 3 to 4 minwtes ( that is 4, 500 to 6, U00 images ) , whereas a significant disparity greater than or equal to Z5 $ in the total durations of the two versions requires a window of the order of 10 to 15 minutes (that is 15,000 to 22,500 images).
It will be noted that the width of the search window, connected with the comparison algorithm imple mented, then varies from two minutes to the order of ten seconds.
2n Europe, for the 25 images/second 625 line standard, the film°video transcription is produced by an analysis of the film at 25 images per second.
For the 30 images/second 525 line standard, the film-video transcription is produced at 24 images per second. The duration of the same film is thereby increased in the ratio 25/24. The values of the file coming from the preparatory phase, that is to say, the signal sdrpt comprising the parameters DUR, TRAMS, TCIV, are then assigned an inverse correcting coefficient of 24/25. Of course, the version or versions to be processed are determined by successive trials.
A general flow chart relating to the execution of the step for putting into relationship the shots for two monolingual versions Vl indicated by vers-ion 1 and version 2 is given in Figure 7.
A first step, denoted by 1000, calls up the original signals relating to each version 1 and 2, these original signals being just the corresponding signals, srdpt, representative of the decomposition into cor-responding processed shots.
A step 1001 then enables the comparison algorithm to be selected, an algorithm such as descr3.bed previously ~~?.~~,_~ ~ ~~
a;
li p " ..._ in relation with Figure 6a to 6c cor example. The step 1001 thus makes it possible to obtain a sequential list of the shots capable of being put into relationship with a step 1003, after a step 1002 for activating the chosen comparison algorithm, A test 1004 relating to the existence of a corresponding empty list makes it possible, by a test 1016 for determining the call-up of the last algorithm from the algorithm library, to pass, on a positive response to the test 1016, to subsequent steps which will be described hereinbelow in the description, and on a negative response to this same test, to return to the step 1001 for selecting the comparison algorithm.
On a negative response to the test 1004, without the list obtained following the implementation of the step 1003 being empty, a step 1005 is provided for seguentially analyzing the list, shot by shot, as described previously in the description.
The putting into relationship, shot by shot, can, for example, as represented in Figure 7, compriae a step 1006 for calling up 'the first shot of the version 1, for example, taken as a reference, and for the implementation of the procedure fox determining, over the search window, coincidence of all the shots of the search window of the version 2. The aforementioned step 1006 is followed by a step 1007 for determining the presence of doublets, that is to say the coinciding, in the current search window, of two shots of the version 2 with one shot of the version 1.
On a negative result to the test 1007, a step 100 is provided for checking validity of the shot numbers, the latter being followed by a step 1009 for checking validity of the shot, for example by means of the signature-implementing procedure previously described in the description. On a positive response to the test 1009, the latter is followed by a step for establishing the relationship of shots in the original lists of the versions 1 and 2. The two coincident shots are then ._ ~1 _ considered to be in relationship.
On a positive response to the test 1007, a step 1010 is provided for processing the doublets. On a posit.i.ve response to a test 1011 for validating the shot, the previously described step 1012, for establishing the relationship of shots in the original lists of the versions 1 and 2, is carried out. A negative response to the tests 1009 and 1011 for validating 'the shot permits, as well as following the aforementioned step 1012, the calling up of an end-of-list test 1013, a negative response to the test 1013 permitting the return to the next shot of the reference version, that is to say the version 1, by the step 1014 and the return t:o the phase for putting the shots into relationship. A positive response to the end-of-list test 1013 enables a test 1015 for reactivating the algorithm to be called up. On a positive response to the test 1015, a return to the step 1002 for activating the comparison algorithm is provided, while on a negative response to this same test 1015, there is provision for passing to the last-algorithm test step 1016 previously mentioned in the description. The positive response to the aforementioned test 1016 makes it possible to pass to a step 1017 for checking validity of the shot numbers and then to a shot-reversal algorithm step 1018 which is itself followed by an end step.
A more detailed description of the step for controlling validity of the shot numbers will be given hereinbelow.
The shots put into relationship with algorithms having weaker equality constraints require a validity check so as not to link shots with different contents.
This check is carried out by the analysis of the shot numbers newly put into relationship with respect to the previously identified adjacent shots. By way of non limiting example, the putting of shots of the version 1 and the version 2 into relationship is recognized as being valid if the shot number in each of the two versions i.s flanked by twa smaller shot numb~:rs and by _~~~
two :larger shot numbers.
The validity check of tkie shot numbers with respect to their environment in fact excludes the facility for putting reversed shots into relationship.
Consequently, a special processing operation manages the reversal of shots, this processing operation being implemented at the end of the phass for putting shots into relationship at the shot-inversion algorithm step 1018. By means of this step, the shots put into relation, whose number does not follow the logical shot-numbering order, are withdrawn from the shot list. The step 1010 for processing the doublets is called when, at a shot of the version 1, called the reference version, a relationship can be established with two or more shots ' of the version 2. The indeterminancy as regards the probable shot to be associated with the shot of the version 1 can in 'the majority of cases be removed by virtue of the analysis of the numbers of adjacent shots.
An example of signals or lists obtained coming from the phase for putting shots into relationship between versions is given in the following Tables 6 and ?e _ 43 _ ~t~ABLE 6 VERSION 1: LIST OF THE SHOTS OF THE VERSIONS 2, 3, 4 AND
PLIT INTO RELATIONSHIP WITH THE SHOTS OF THE VERSION 1.

5 'JERS.1 DUR TRAMS VERS.2 VERS.3 VERS A VERS.5 SHOT No SHOT No SHOT No SHOT No SHOT No 1 loo 0 1 1 l0 3 125 0 3 3 6 1.3 89 1 13 13 13 14 All the shots of the versions 2 and 3 were able to be put into relationship with all the shots of the version 1.
- The shots 1, 2, 3, 7 and 9 of the version 4 were not able to be put into relationship with those of the version 1.
- The shots 1, 2, 3, 4 and 5 of the version 5 were not able to be put into relationship with those of the version 1.
- The shots 8 and 9 of the version 1 are not present in the version 5 (cutting).

r v~ ~ ~. r~ r_ _~
T~~~.1 VERSION 5 t PARTIAL VIE6d OF' THE SHOTS Ok' THE VERSION 5 SHOT No DUR TRAMS SHOT No ALGORITHM
VERS.5 VERS.1 INDICATOR USED

6 125 0 3 2 shots, strict equality 7 234 2 4 3 shots, strict equality 8 45 0 5 3 shots, strict equality 9 34 0 6 3 shots, strict equality 10 565 1 7 2 shots, very close equality 11 35 0 10 2 shots, close equality 12 269 0 11 3 shats, very close equality Each file, one per version, essentially contains the list of the shots put into relationship between the version in question and the other versions.
Finally, the final phase of the software loaded into the calculator 7.0 makes it possible to terminate in the ghase for generating the sequential list of the shots without a definitively established relationship, that is to say the list of the shots which it has not been possible to make coincide.
The final phase now enables the difference in duration existing between these shots to be determined in order to establish a new file, which will enable the signal spsv of Figure 5a, or the sequential list of the shots without a definitively established relationship, to be generated.

~ t1 f-y For the shots put unto relationship, the difference in duration between two shots is:
D = (DUR + TRAMS) of V2 - (DL3R + TRAMS) of V1.
The duration calcui.ation is conducted as a function of the algorithm used for putting shots into relationship:
-- a.lgorithm with strict equality criterion: o = 0, - algorithm with very-close or close equality criterion:
0 is calculated according to the aforementioned relationship.
For a sum algorithm, the calculation of v is extended to the two or three shots composing the sum.
Finally, for the non-identified shots, the difference in duration is calculated, on each series of shots not put into relationship, by subtraction of the extreme real time codes.
The corresponding file or signal spsv previously described may then be used in order to produce, by means of the module 4 represented in Figure 5a, a composition of the coherent program for the set of image and sound sources of the aforementioned monolingual versions Vl.
The use, in accordance with the subject of the present invention, wall be described with a view to producing the composition of multilingual-narrative televised programs by composition of h monolingual narrative programs, attributes of common video images of the same program.
Tn such a use, the post-composition operation consists, from the various versions V1 constituted by the video signal ViVl relating to the version Vl, and a corresponding audio signal denoted by AuVl, as well as from a composition frame constl.tuted in fact by the signal spsv and from one of the versions taken as the reference version, the version 1 for example, .in perform-ing a video post-synchronization between the video signals of the reference version and each other version.
Two shots, or groups of shots, F1, P1, with 1 ~ 1, of the reference version and oi: the version subjected to synchronization, are synchronous when the difference in duration of the two shots, or groups of shots, put into relationship is zero.
The post-synchronized shot is obtained by selec tion and storage of one of the corresponding video signals, that is to say signals AuVl, AuVI, with 1 :E 1.
These signals are synchronized respectively to their corresponding version, including their sound-track information. Such an operating procedure enables L-1 pairs of post-synchronized versions to be constituted.
P. two-by-two comparison of the difference in duration of the pairs of post-synchronized shots is performed in order to determine, by intersection, the shots which require no intervention. An analog illustration of the determination of these shots requiring no intervention is given in the table herein-below:
TAB?JE 9 VERS.1 DUR DIFF. DUR DTFF. DUR DIFF. DUR DIFF.
SHOT No VERS2-VERS1 VERS3-VERS1 VERSO-VERS1 VERSS-VERS1 4 p 0 0 0 d'h (') ~ p°.~ .7 '~
a a C_ _ . ~. :~.
Very small differences in duration between shots, less than 10 images for example, may then be compensated for by time-compression or -expansion processing operation of the audio signal, l~uVl, and the 7_arger differences may be compensated for either by suppression of the longest-duration audio signal or by external intervention by a human operator.
The mode of operation of the previously described module 4 is 'thus rendered completely automatic although, if required, maintained under the initiative of the human operator.
Of course, the manual intervention of a human operator is useful when the synchronization between versions has not been able to be completely performed in automatic mode.
These manual intervention operations may consist, by means of viewing and listening, in validating synchro-nization propositions when series of shots have not been able to be put into relationship, but have small differences in duration , in choosing synchronization modes when shots have been put into relationship, but have significant differences in duration, in resolving individual cases of shots not put into relationship, for example in the case of a shot missing in one version or problems of a purely artistic nature.
The module 4 supplies as output a signal con-stituted by a computer file forming the final product, this file being representative of the editing determined automatically or semiautomatically, according to the choice of the user, by the processing program. This file, compatible with the format of the editing machine, not represented in the drawing, contains the sequential list of the time codes of the shots of the versions to be synchronized.
The human operator then edits this list and can validate or modify their editing commands. Such a list comprises the shots, whose difference in duration is large, the shots which have not been able to be put into relationship and the shots whose automatic synchronization solution requires validation.
An advantageous variant of a procedure for synchronizing various monolingual. linguistic versions will be described in conjunction with Figures 8a to 8c.
Such a variant enables artistic and technical characteristics of the work itself to be taken into account. In Figures 8a to 8c, two separate monolingual versions to be synchronized are indicated by way of non-limiting example by V1 and V2, the version Vl serving as the reference for example and the version V2 constituting in fact the version to be compared and synchronized.
The great diversity of the situations and the impossibility of a priori evaluation of the subjective consequences of the synchronization actions limit the automation capacity of these. Experiments have shown however that, most often, the differences in duration result from start- or end-of-shot cutting operations, that they are exact~.ly compensated for and therefore cancel each other out,, or alternatively that they are considerably reducea by grouping together several consecutive shots. T;~e automatic actions for each possible choice of corr~:sponding synchronization can then be introduced into the list of the various applicable solutions, it being possible for indicators or commentaries intended to inform the operator to be introduced. Thus the aperator may have available an assistance for testing various solutions. He can choose one or none of them, and may then decide, if none is satisfactory, to perform the work in manual mode.
In a non-limiting manner, the aforementioned solutions may consist in performing either a start--of-shot coincidence operation with preservation of the shortest shot and end-of-shot cutting of the longest shot, or an end-of-shot coincidence operation with preservation of the shortest shot and start-of-range cutting of the longest shot, or a preserva~:ion of all the shots, the start- and end-of-shot coincidence operation being able to be obtained by processing one of the versions, processing such as, for example, reading at a variable different rate for deviations, the duration of which does not exceed 4 ~ of the duration of the shot, or an association of consecutive shots, the overall duration of which has a difference which is zero or is at most egual to the duration of a few images, for example ten images, as a result of the inter-shot compensations for the deviations.
The various solutions are represented from the two aforementioned monolingual versions V1 and V2, the most probable coincidences of which have been established and made clear, in accordance with the subject of the invention, as represented in Figure $~.. Figure 8b represents diagrammatically the implementation of the first three previously mentioned solutions, namely end cutting, start cutting and reading at variable rate. The image numbers are arbitrary and given purely by way of illustration, the degree of compression 3/735 being able to correspond to a real case. Finally, Figure 8c represents diagrammatically two types of editing by association of shots with a direct copy of two associated shots, respectively a copy of two associated shots and end-of-shot cutting of two images.
A video analysis system has thus been described for editing a recorded or broadcast, televised program, in the form of a succession of images, which is particularly powerful insofar as this system and its use permit a complete automation of the determination of the shot changes arising in a program for the compiling and synchronization of various data accompanying programs, such as 16/3-4/3 refraining, introduction of subtitles or various procedures for post-production of films or of video recordings, such as colorization of black and white ~5 films, for example. The system, which is the subject of the present invention, and its use also enable poet-production operations to be carried out, such as procedures for post-synchronization of films for ~~~~.~.~t~_.
- 50 _ compiling multilingual television programs from dubbed archives considered as corresponding monolingual versions.
Furthermore, the system, which is the subject of the present invention, and its use also enable assistance to the operations for_ checking duration of transmission broadcasting on recordings in para11e1 with the antenna transmission, such as checking the conformity of the specification of programing companies, conformity of the IO broadcasting times for advertising slots or of political broadcasts during electoral campaigns, for example.

Claims (17)

1. A video analysis system for editing a recorded or broadcast, televised program in the form of a succession of images, each image being recorded or broadcast in the form of audio- and video-frequency data associated with the image of order k in question and of an associated time code representative of said image, wherein said system (1) comprises:
- means (10) for measuring the time variance of the composition of the images, by determination, between two successive images of order k-1, k, of an image difference signal and of corresponding parameters representative of the scenic activity of the recorded or broadcast program for a group of at least two successive images, of order k-1, k;
- means (11) for analyzing the recorded or broadcast program making it possible, from said difference signal and from said corresponding parameters representative of said scenic activity, to establish data representative of the cutting-up into shots, which consist of a group of p successive images, of said recorded or broadcast, televised program.

2. The system as claimed in claim 1, wherein said means (10) fox measuring the time variance of the composition of the images include:
- means (100, 102) for storing the video component S(i,j,k) of the image signal where i indicates the index of the component sample in the active line of the image in question, j indicates the index of the active line in the image in question and k indicates the order number of the analyzed image in question, and means (100, 101, 102) for storing the component S(i,j,k-1) of the image signal of the image of prior order k-1 in question;
- means (103) for calculating the image difference signal satisfying the relationship:
Sd(i,j,k) = A . [S(i,j,k) - S(i,j,k,-1)] + Sm where A represents a value-weighing coefficient or a predetermined functions, Sm represents a correction coefficient which is a function of the permissible dynamic range for the component S(i,j,k) in question.

3. The system as claimed in claim 1 or 2, wherein said means (10) for measuring the time variance of the composition of the images furthermore include:
- means (105) for calculating, far an image of order k, from said image difference signal Sd(i,j,k), a histogram HSd(k) of the number of occurrences of each amplitude level of the image difference signal Sd(i,j,k), - means (106) far calculating corresponding scenic-activity parameters by comparison of the distribution of the spectrum of each histogram HSd(k) with a plurality of reference models or distributions representative of corresponding scenic activities.

4. The system as claimed in claim 3, wherein said means (105) for calculating, for an image of order k from the difference signal Sd(i,j,k) or from the filtered difference signal, a histogram HSd(k) include:
- an input interface circuit receiving the filtered difference signal Sd(i,j,k), said interface circuit supplying time reference signals (srt), video data signals (sdv) and clock signals (clk);
- a circuit generating service signals receiving, on the one hand, said time reference signals (srt) and, on the other hand, said clack signals (clk), and supplying service signals (ss), - a circuit for sorting the video data signals (sdv) according to a histogram distribution law using predetermined video quantizing levels, said sorting circuit receiving, on the one hand, said video data signals (sdv) and, an the other hand, sequencing signals (slb) and supplying video data signals sorted by level (sdvtn), - a plurality of storage registers, each register forming a bank of sorted-level values, intended to receive and to store said video data signals sorted by level (sdvtn) as a function of their level, - an adder circuit for down-counting in each bank the number of occurrences of the amplitude levels of the sorted video signals, which circuit is connected via a read- (L) and write- (E) BUS data-link to each of said registers forming a bank, said adder circuit delivering, after reading and down-counting each bank, a bank down-count signal (sdb) representative, for each corresponding bank, of the number of occurrences of the amplitude levels of the sorted video data signals (sdv) relating to the signal Sd(i,j,k) in question.

5. The system as claimed in claim 3 or 4, wherein said means (106) for calculating corresponding scenic-activity parameters includes - a microprogramed sequencing unit receiving the time reference signal (srt) and supplying a bank read signal (slb), - a buffer register interconnected via the read-BUS
(L) to each of said registers forming a bank, said buffer register comprising, on the one hand, an auxiliary memory circuit enabling the bank down-count signal (sdb(0,1,2,3)) relating to each bank to be stored and, on the other hand, a conversion table, said conversion table supplying, from the bank down-count signal, a scenic-activity indicator signal (ias) for the pair of successive images of order k-1, k in question.

6. The system as claimed in either of claims 4 or 5, wherein, values, which are symmetrical with respect to the level of amplitude value corresponding to that of the average luminance, are assigned to said sorted-level value banks, a first bank being assigned a continuous range of values which are symmetrical with respect to said amplitude level of the average luminance, and two ranges of discrete values of values, which are symmetrical with respect to said amplitude level of the average luminance, being assigned to each of the other banks, the spectrum of the histogram HSd(k) being expressed by the bank down-count signals (sdb(0,1,2,3)) relating to the bank in question.

7. The system as claimed in claim 4, 5 or 6, wherein said bank down-count signals sdb(0,1,2,3) are quantized onto a number NAS of scenic-activity levels, so as to define, for each bank, a quantized elementary scenic-activity level, nase(0,1,2,3), the set of the quantized elementary scenic-activity level signals forming said scenic-activity indicator, (ias).

8. The system as claimed in claim 6 or 7, wherein, there being associated with said first bank:
- a second bank to which two ranges of discrete values of symmetrical values are assigned, these two ranges being themselves adjacent to the central-value range associated with the first bank, - a third bank to which two ranges of discrete values of symmetrical values are assigned, these two ranges being themselves adjacent respectively to the two ranges of discrete values associated with the second bank, and - a fourth bank to which two ranges of discrete values or symmetrical values are assigned, these two ranges being themselves adjacent respectively to the two ranges of discrete values associated with the third bank, said scenic-activity indicator (ias) is representative of the absence of scenic activity, of a low scenic activity, of a medium scenic activity and of a high scenic activity, - the absence of scenic activity being defined by an elementary scenic-activity level, relating to the first bank, greater than the average value NAS/2 of the scenic-activity levels, nase(0) ~ NAS/2, the elementary scenic-activity level relating to the second, third and fourth banks being substantially zero, nase(1,2,3) = 0, - the low scenic activity being defined by an elementary scenic-activity level relating to the first bank greater than the average value NAS/2 of the scenic-activity levels, nase(0) ~ NAS/2, the elementary scenic-activity level relating to the second bank being less than 1, nave(1) ~ 1 and the elementary scenic-activity level relating to the third and to the fourth banks being substantially zero, nase(2) = nase(3) = 0, - the medium scenic activity being defined by an elementary scenic-activity level relating to the first bank greater than the average value NAS/2 of the scenic-activity levels, nase(0) ~ NAS/2, the elementary scenic-activity level relating to the second and third bank being less than or equal to one, nase(1) ~ 1, and nase(2) ~ 1, and the elementary scenic-activity level relating to the fourth bank being substantially zero, nase(3) = 0, - the high scenic activity being defined by an elementary scenic-activity level relating to the first bank less than or equal to the average value NAS/2 of the scenic-activity levels nase(0) ~ NAS/2, or by an elementary scenic-activity level relating to the fourth bank, greater than or equal to the average value NAS/2 of the scenic-activity levels, nase(3) ~ NAS/2.

9. The system as claimed in one of the preceding claims, wherein said means for analyzing the broadcast program, from said difference signal and from the parameters representative of the scenic activity, includes:
- calculator means (110) interconnected with said means for measuring the time variance of the composition of the images, said calculator means enabling said scenic-activity parameters to be stored, said calculator means comprising:
- discrimination software making it possible to select the images of order k in question, or the groups of p images, the scenic-activity parameters of which are greater than one or more predetermined threshold values, which makes it possible to establish time and/or cardinal data representative of the cutting-up into shots, consisting of p successive constituent images of said recorded or broadcast, televised program.

10. The system as claimed in claim 9, wherein said calculator means (110) receive as input:

- a time code signal (ct) representative of the succession of each image of order k, which images constitute the broadcast program, - a scenic-activity indicator signal (ias) representative of the scenic-activity parameters for two successive images of order k-1, k, said signal being supplied by said means for measuring the time variance of the composition of the images, - an exchange-synchronization signal (se) enabling data corresponding to said time code signal and to said scenic-activity signal to be acquired at the image frequency by said calculator means, said discrimination software making it possible, using as discrimination criterion the scenic-activity indicator of every image of order k or group of p images of order k-(p-1),k, to determine the editing points of said program, corresponding to the two successive shots, P-1, P, said discrimination criterion being established with respect to three noteworthy standard shots of scenic activity, a first standard shot, indicated by a cut shot, such as the passage from one person to another person in a dialogue situation, having a scenic activity with an abrupt time variation, a second standard shot, indicated by a lap-dissolve shot, having a scenic activity with a substantially linear time transition, and a third standard shot, designated by an overlay shat, having a pulse-type time variation of scenic activity, said calculator means supplying a signal representative, for the corresponding broadcast program, of its cutting-up into shots, signal (srdp), each shot of row P being defined by a start time-code and an end time-code.

11. A use of a video analysis system for editing a recorded or broadcast, televised program in the form of a succession of images, as claimed in one of the preceding claims 1 to 10, in order to replace the steps of searching, by means of viewing, for the editing points of the programs or parts of programs during video or cinema film post-production operations, or checking the conformity of the editing or broadcasting of the final product, or the composition of multilingual-narrative televised programs.

12, The use of a video analysis system for editing a recorded or broadcast, televised program in the form of a succession of images as claimed in claim 11, wherein, with a view to producing a composition of multilingual-narrative programs by post-composition of monolingual-narrative programs, attributes of common video images of the same program, said use consistss - in performing an analysis step of each monolingual program in order to establish the data representative of the cutting-up into shots consisting of a group of pl successive images, where pl indicates, for the monolingual version Vl in question, the number of constituent images of a shot Pl in question;
- in establishing a systematic correlation between the various shots P1 of each monolingual version which consist of a predetermined number p1 of successive images for each monolingual version V1 in question, so as to establish a time coherence between all the shots Pl of each monolingual version and each monolingual narrative, said correlation enabling the time and/or cardinal data to be adopted in terms of minimum and maximum number of successive images pl constituting a given shot P1 enabling all the monolingual narratives in question to be carried;
- in storing, in vector form (Pml, Plt, plt), said data so as to constitute an editing or composition frame of the multilingual program in which, in the order of the shots Pml which can go back into the composition of the final multilingual program, are associated, for each monolingual version in question, the order of the shot Plt, the number of successive images plt and the time data for the start and end of the shot Plt in question.

13. The use as claimed in claim 12, wherein said step for analyzing each monolingual program comprises, from the signal representative of the cutting-up into shots (srdp) obtained for each monolingual version Vl:
- a step for complementarily filtering of said signal representative of the cutting-up into shots (srdp) in order to generate a corresponding filtered signal representative of the cutting-up into shots (~rdp), - a step for calculating the duration of the shots, (DUR), and of the transition shots (TRANS) by subtraction of the successive time codes (ct), - a step for calculating and assigning, to the first shot of each monolingual version Vl, a virtual time code (TCIV) or arbitrary value and, by recurrence, for each subsequent successive shot Pl p a virtual time code satisfying the relationship:
TCIV(P1) = TCIV(Pl-1) + DuR(Pl-1) + TRANS(Pl-1), - a step for creating a processed signal representative of the cutting-up into shots for the monolingual version in question (srdpt), said signal allowing the identification of each shot Pl and containing, beyond to the real time code, ~~a virtual time code, ~ a shot number, ~ a duration, ~~a duration of the associated transition shot.

14. The use as claimed in claim 12 or 13, wherein, for a number L of monolingual versions, said correlation is performed by a group of two monolingual versions, the correlation between successive shots Pl, Pl+1 of two monolingual versions Vl, Vl+1 being performed by comparative analysis of the duration and/or of the corresponding number of images pl, pl+1 for the shots in question, or of a linear combination of said durations or of said numbers of successive images, over a predetermined number of successive shots Pl, P-1,1, ...P+r,l; P1+1, P+1,1+1, ~~~P+r,l+1 of both monolingual versions in question.

15. The use as claimed in one of claims 12 and 13, wherein said step for establishing a systematic correlation between the shots Pl of each monolingual version consists, from said signal representative of the cutting-up into a processed shot, (srdpt), in performing a putting-into-relationship operation, shot to shot, between two monolingual versions Vl, Vl-1, by comparison of the duration (DUR) of their corresponding shots Pl;
P+r,l+1, said comparison being performed by concatenation of elementary comparison steps based on a constraint criterion of a more or less strict equality of the shot durations (DUR) and/or of the transition durations (TRANS), said comparison being performed, for each elementary comparison, on a search window constituting, for a shot Pl of order P of the version Vl, a reduced environment of shots Pl+1 to P+sl+1, of the version Vl+1, the putting into relationship of two shots, respectively Pl and P+sl+1, of the two versions Vl, Vl+1 being immediately followed by a relative reframing of the shorter virtual time code TCIV of the two versions Vl, Vl+1 in question, the virtual time code of the next shots of the version, the virtual time code of which has been re-updated, also being re-updated, until the recognition of an already re-updated virtual time code.

16. The use as claimed in claim 12 or 14, wherein, with a view to producing said composition of multi-lingual-narrative televised programs by post-composition of L monolingual-narrative programs, attributes of common video images of the same program, said post-composition operation consists, from the various versions Vl constituted by a video signal ViVl and a corresponding audio signal AuVl of said composition frame, and from one of the versions taken as the reference version, V1, in performing a video post-synchronization between the video signals of said reference version and of each other version, two shots, or groups of shots, Pl, Pl, with 1 =
1, of the reference version and of the version subjected to synchronization being synchronous when the difference in duration of the two shots, or groups of shots, put into relationship is zero, the post-synchronized shot being obtained by selection and storage of one of the corresponding video signals of the corresponding audio signals AuVl, AuVl, which are respectively synchronized to their corresponding version, so as to constitute L-l pairs of post-synchronized versions, - in performing a comparison, two by two, of the difference in duration of the pairs of post-synchronized shots in order to determine, by intersection, the shots which require no intervention, - in compensating for the very small duration differences, less than a predetermined number of images, by time-compression or -expansion processing of the audio signal AuVl, the larger differences in duration being able to be compensated for either by suppression of the longest-duration audio signal or by external intervention.

17. The use as claimed in claims 15 or 16, wherein the putting into relationship, shot to shot, of each monolingual version consists in performing either - a start-of-shot coincidence operation, with preservation of the shortest shot and end-of-shot cutting of the longest shot, or - an end-of-shot coincidence operation with preservation of the shortest shot and start-of-range cutting of the longest shot, or - a preservation of all the shots, the start- and end-of-shot coincidence operation being obtained by processing one of the monolingual versions, or - an association of consecutive shots, the overall duration of which has a difference which is zero or at most equal to the duration of a few images.