CA2323867A1 - Method for verifying the authenticity of an image recorded in a person identifying process - Google Patents

Method for verifying the authenticity of an image recorded in a person identifying process Download PDF

Info

Publication number
CA2323867A1
CA2323867A1 CA002323867A CA2323867A CA2323867A1 CA 2323867 A1 CA2323867 A1 CA 2323867A1 CA 002323867 A CA002323867 A CA 002323867A CA 2323867 A CA2323867 A CA 2323867A CA 2323867 A1 CA2323867 A1 CA 2323867A1
Authority
CA
Canada
Prior art keywords
head
differential image
region
function
individual images
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Abandoned
Application number
CA002323867A
Other languages
French (fr)
Inventor
Wolfgang Konen
Michael Brauckmann
Andre Liebetrau
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Idemia Identity and Security Germany AG
Original Assignee
Individual
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Individual filed Critical Individual
Publication of CA2323867A1 publication Critical patent/CA2323867A1/en
Abandoned legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G07CHECKING-DEVICES
    • G07CTIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
    • G07C9/00Individual registration on entry or exit
    • G07C9/30Individual registration on entry or exit not involving the use of a pass
    • G07C9/32Individual registration on entry or exit not involving the use of a pass in combination with an identity check
    • G07C9/37Individual registration on entry or exit not involving the use of a pass in combination with an identity check using biometric data, e.g. fingerprints, iris scans or voice recognition
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16Human faces, e.g. facial parts, sketches or expressions

Abstract

The invention relates to a method for verifying the authenticity of an image of a person to identified, which image is recorded during a personal identification process. According to the method provided for in the invention, a sequence of consecutive individual images of the person is recorded and the authenticity of the recorded image is determined if in at least two consecutive individual images of the sequence intrinsic movements are detected, i.e. movements which are carried out by the person to be identified consciously or unconsciously and without changes in his or her centre of gravity.

Description

MEi'HOD FOR VER11=YING THE AUTHENTICITY OF AN IMAGE RECORDED IN A
PERSON IDENTIFYING PROCESS
The present invention relates to a method for verifying the authenticity of an image recorded in a person identifying process and belonging to a person to be identified.
Known person-identifying methods which are e.g. based on face recognition as described in German patent 44 06 020 only compare single photographs of the person to be identfied with reference pictures. Thus only static information is processed in said methods. Such methods have the drawback that an access cantro! operating with such a method can easily be deceived by showing a photo or a mask-like picture of the person.
Therefore, various methods have been developed for reducing the risk of such deceptions.
For instance, it is known from German Utility Model 2 950 X60 that the spectral range of the light used for identifying a person is limited by an IR filter disc to a range in which a photo or a mask-like picture has reflectance characteristics which differ from those of an actually recorded face_ Such an IR filter disc has the drawback that the image quality is.
reduced thereby and that the demands made on the quality of the identifying process increase. 0n the other hand, there Is the risk that an image medium is found with reflectance characteristics in the IR range that are similar to the reflectance characteristics of a person's face in the IR range, whereby a deception is again possible with such an image medium.
Furthermore, the prior art discloses methods in which the picture of the person to be identified is taken with a stereo imaging system. For instance, such a method is used in the face recognizing system described in the amide "In-Your-Face Security" in PCWeek of March 26, 1997. Other methods for detecting the three-dimensional structure of a face either fully or in part are laser triangulation or the so-called shape-from-shading methods disclosed by Attick, GrifF~n and Redlict5 in Neural Computation 8. 1321-1340, 1996; these make it possible to draw conclusions from the gray-level curve at a given direction of illumination with respect to the depth infomlation of the object viewed.
Although deceptions in the case of which an image is held in planar fashion in front of the recording system can be detected with such methods, deceptions with a cylindrically bent image or a three-dimensional mask cannot be recognized as such.
Moreover, the prior art discloses methods in which the pictures of the person to be Identified are taken by means of thermographic methods; see e.g. German patent 051. In this instance, the radiant heat emitted by the person to be identfied is detected and compared with a given reference value for identifying the person. This method, however, has the drawbaGc that the devices far carrying out the methods are very expensive, in particular the thermographic camera used for image-recording the person to be identified. Thus thermographic methods cannot be used in access controls which are to be used in many places, e.g, in cash dispensers torwithdrawing money andlor in banking transactions.
In view of the above-discussed drawbacks of the devic$s known from the prior art, it is the object of the present invention to provide a method for verifying the authenticity of an image recorded in a person identifying process and belonging to a person to be identified, in the case of which the reliability during ver'rfrcation of the authenticity is improved over the prior art vn the one hand and thus the security to be achieved with an access control, and which can be implemented at law costs on the other hand.
This obJect is achieved by a method for verifying the authenticity of an image recorded in a person identifying process and belonging to a person to be identified, the method comprising the steps of recording a sequence of consecutive individual images of the person and of determining the authenticity of the recorded image if in at least two consecutive individual images of the sequence intrinsic movements of the person are detected.
Since in the method of the invention intrinsic movements of a person to be identified, i.e.
movements carried out by every person, i.e. also a person to be identified, consciously or unconsciously and without changes 1n his or her center of gravity are used for verifying the au#hgnticity of the recorded images, the authenticity will only be detected if the consecutively recorded individual images actually originate from a person performing intrinsic movements, i.e. from a living person. Thus a deception of an access control using the method of the invention is no longer possible by showing a cylindrically bent photo or a mask-like picture of the person to be identified, which considerably enhances the security guaranteed by the access control.
Furthermore, the method according to the invention can be implemented at relative low costs because only an image recording means (called recording means in the following) and an image evaluating means are required for canying out the method and because said two means for carrying out the image recognition proper are already provided for.
Consequently, the method of the invention for verifying the authenticity of a recorded image is also suited far use in access controls which are to be used in many places, e.g.
in cash dispensers for withdrawing money andlor banking transactions.
According to a preferred development of the inventive method at least one differential image of two consecutive individual images of the sequence can be evaluated for determining the intrinsic movements.
The amount of data to be processed in the method of the invention can considerably be reduced by the formation of such differential images, whereby the method can be carried out at a faster pace. This has, inter olio, the effect that the demands made on the hardware for performing the method can be reduced, whereby the total costs of the system are immediately decreased.
Furthermore, according to a preferred variant a differential image which results from an AND operation of two consecutive differential images can be evaluated for determining the intrinsic movements.
Such a development yields, in particular, a differential image in which changes are considerably reduced that follow e.g. from a releas~ of a dominating background in two consecutive individual images and ace thus reflected in the differential image of two consecutive individual images. In other words, the differential image to be assessed in the final analysis is freed by this variant from changes tt7at da not result from intrinsic movements.
According to another preferred development each differential image can be binarized prior to evaluation.
This measure, in tum, greatly reduces tt~e amount of data to be processed, resulting in an increase in the speed of the method and thus in smaller demands on the hardware having the same efficiency.
~inarization may e.g. be carried out by means of a threshold value which is determined by evaluating the background of one of the differential images, e.g. by averaging the intensifies of those pixels in the differential image that are in the area of the static background_ This has the advantage that binarization is always determined in dependence upon the instantaneously existing image-recording situation. Thus the method can be adapted to environment-induced changes, for instance to changing light andlor contrast conditions, thanks to such a development.
Furthermore, at least one section of enhanced movement can be selected in each differential image, and it is possible to detect intrinsic movements if said at least one section in the differential image is consistent with a predetermined region in which intrinsic movements are to be expected.
Thanks to such a development of the method of the invention, the amount of data to be processed can be reduced again, which results in the already above-discussed advantages.
To this end, according to a first alternative, the section of enhanced movement can be defined by a vertical and a horizontal extent, the vertical extent being substantially determined by a peak of the function of the products frorn the horizontal projec~ons with the horizontal variances of the differential image, and the horizontal extent being substantially determined by a peak of the function of the products from the vertical projections with the vertical variances of the differential image.

Advantageously, each function can be smootried with a low-pass filter prior to the determination of the corresponding peak. This measure helps to avoid high-frequency noise in the difFerential images, which might lead to an impaired determination of the peak and thus of the section of enhanced movement. On the whole, this yields a more robust determination of the section of enhanced movement.
According to a further development of the first altematlve, sections can be chosen from the differential image that are supposed to be in symmetrical relationship with one another, and intrinsic movements can be detected when the symmetrical relationship is verified by a correlation analysis of the corresponding sections. Since intrinsic movements, e.g. in the face of a person to be identified, are symmetrical as a rule, such a design helps to check whether the intrinsic movement detected in a specific region is actually an intrinsic one or just a different movement, e.g. noise or the like.
Moreover, independently of or together with said check, a further check can be carried out as to whether the enhanced movement in the region of enhanced movement is irregularly distributed over said region.
It can thereby be rated out that small changes which might add up under adverse conditions are wrongly detected in their sum as an intrinsic movement According to a second alternative, a function which depends on the vertical position and which is determined for a given vertical position through the product of the hor~ontal projection with the horizontal variance of the differential image, and a function which depends on the horizontal position and which is determined for a given horizontal position through the product from the vertical projection with the vertical variance of the differential image can also be evaluated for detecting intrinsic movements.
Preferably, in the second alternative, it is also possible to smooth each function with a low-pass filter prior to the determination of the corresponding peak so as tv filter high-frequency noise out of the differential images.

According to an advantageous development, it is possible for the evaluation of the above-mentioned functions to determine at least one peak in the function depending on the vertical position, and at least one~peak in the function depending on the hori2ontal position, and finally intrinsic movements can be detected if said peaks are within predetermined limits in the vertical and horizontal direction, said limits being defined by the vertical and horizontal extent of at least one area in which Intrinsic movements are to be expected.
8y analogy with the first alternative, checking rnett~ods which are based on the fact that intrinsic movements are normally exhibiting symmetries can also be carried out for intrinsic movements detected with the second alternative.
For instance, sections which are supposed to be in a symmetrical relationship with one another can be evaluated on the basis of the function depending on t>1e vertical position andlor the function depending on the horizontal position, and intrinsic movements can finally be detected if the symmetrical relationship is verified by a correlation analysis of the con-esponding sections.
To make said checking method more robust with respect to relative deviations of the sectjons of the function which are supposed to be in a symmetrical relationship with one another, the sections may be standardized with respect to one another prior to the correlation analysis.
Ih addldon or as an alternative to said checking method, the function depending on the vertical position andlor the function depending on the horizontal position can be evaluated, and intrinsic movements can be detected if it is determined by way of a correlation analysis that the values of tf~e function depending on the vertical position andlor of the function depending on the horizontal position are irregular over a predetermined range.
Such an additional check will rule out that small changes that may add up in adverse cases are wrongly recognized in their sum as an intrinsic mavement_ According to a prefer-ed development of the above- iscussed development of the inventive method intrinsic movements can be determined in the head portion of the person to be identified for verifying the authenticity of the recorded images.
This development has the advantage that the head portion as compared to the whole body of a person to be identified shows most of the intrinsic movements, which are also the most dominating ones. Thus, the method of the invention can be carried out in the most efficient way In the head region of a person to be identified.
Regions with particularly pronounced intrinsic movements in the head portion are e.g. the mouth region and/or the cheek region andlor the nose region andlor the eye region.
In a preferred development tf~e Intrinsic movements to be determined are intrinsic movements in at least two regions that are symmetrical to each other in the head portion of the person to be identified, and said at (east two symmetrical regions are used for determining the axis of symmetry of the head portion.
With the symmetry axis found in this way it is possible to produce an image which can be used for identifying a person and in which a frontoparallel-rotated head region is transformed into a head region with a rotated axis of symmetry.
In addition or alternatively, an image of the person to be identified which can be used for identifying the person and which is composed of the left head region and the mirrored left head region or of the right head region and the mirrored right head region, respectively, can be produced by way of the symmetry axis found.
Both alternatives have tt~e advantage that improved images can be provided for a later person identlficatlon. For Instance, In the case of the first alternative, recorded images in which the head region of the person to be identfied is inclined relative to the vertical can be used for image recognition. With the second alternative, it is also possible to use those images for person identification in which an image half is glared and which would thus be useless for the Identification of a person.

According to another preferred development of all of the above-discussed embodiments the regions in the individual images or frames in which intrinsic movements are expected can be extracted from the individual images prior to the detection of the authenticity.
This development has the advantage that the images required for detecting the authenticity can be extracted from any desired video sequences.
Thus, the person to be identified has just to step into the visual t9eld of the recording means. This has the advantage that optimum individual images can be selected from a plurality of individual images for respectively detecting the authenticity.
Furthermore, it is not necessary that the person to be identified assumes a fixedly predetermined position so that the image recordings required for identifying the person and for verifying the authenticity can be made of said person.
dike in the detection of the authenticity, advantageously at least one differential image of tow consecutive individual images of the sequence can be evaluated for reducing the data to be processed during extraction of the regions.
Furthermore, during extraction of the regions it is also possible to evaluate a differential image resulting from an AND operation of two consecutive differential images if the differential image to be finally evaluated is to be freed from changes resulting e.g_ from the release of a dominating part of the background.
Likewise, each differential image can be binarized prior to evaluation and the binafialng operation can e.g. be carried out by means of a threshold value which is determined by evaluating the background of one of the differential images, e.g. by averaging the Intensities of these pixels in the differential Image that are in the region of the static background. As already discussed above in connection with the detection of Intrinsic movements, the amount of the data tv be processed can considerably be reduced by this development.
Analogously to the detection of the intrinsic movements the head region of the person to be identified can be extracted In an advantageous manner.

According to a preferred development the extracted head region can be transformed to a predetermined standard size for identifying the person. Variations in size which might lead to problems, in particular with respect to the computing time, in the complicated algorithms for face recognition can thereby be taken into account.
For extracting the head portion at least two head boundaries can be determined in the corresponding individual images or differential images, and the head region can be extracted on the basis of said head boundaries from the corresponding individual images or differential images.
According to a first alternative the determination of the head boundary in the corresponding individual images ar differential images may include the determination of the upper and the left head boundary.
According to a second alternative the right head boundary can additionally be determined.
Expediently, but not limited thereto, each head boundary may be defined by a head boundary line which extends such that the contour of the head is substantially entirely located within the head boundary lines.
According to a preferred development the function of the vertical projections of the binarized differential image can be determined for determining the upper head boundary, and the upper head boundary can be defined by the first maximum of the absolute value of the first derivative of said function that is above a predetermined threshold value.
If high-frequency noise is to be eliminated, the function of the vertical projections could be smoothed with a low-pass filter prior to the definition of the head boundary.
For the determination of the left head boundary the function of the horizontal projections of the binadzed differential image is determined, and the left head boundary can be defined by the first maximum of the absolute value of the first derivative of said function that is above a predetermined threshold value.
Alternatively, for the determination of the left head boundary the differential image can be divided into a plurality of vertically successive strips, the first strip being vertically downwardly adjacent to the upper head boundary determined; in each strip the function of the horizontal projections of the binarized differential image can then be determined, the absolute values of the first derivatives of the resulting plurality of functions of the horizontal projections can subsequently be formed, the sum of the resulting plurality of absolute values can be added, and the left head boundary can finally be defined as the first maximum of said sum that is above a predetermined threshold value.
The advantage of said second alternative is that the head region is only detected through the strips below the upper head boundary, and the shoulder portion positloned thereunder, which might lead to inaccurate results in the determination of the left head boundary, is masked_ In this case, too, the functions) of the horizontal projections can optionally be smoothed with a low-pass filter prior to the definition of the head boundary.
By analogy with the frrst alternative regarding the determination of the left head boundary, it is possible to determine the function of the horizontal projections of the binarized differential image according to the preferred development in which the right head boundary is also determined, and the right head boundary can be defined by the last maximum of the absolute value of the fast derivative of said function that is above a predetermined threshold value.
By analogy with the second alternative in the determination of the left head boundary, it is possible according td the preferred development in which the right head boundary is also determined, to divide the differential image into a plurality of vertically successive strips for the determination of the right head boundary; in this case the first strip is vertcalfy downwardiy adjacent to the upper head boundary determined, the function of the horizontal projections of the binarized differential image Is determined in each strip, the absolute values of the first derivatives of the resulting plurality of functions of the horizontal projections are formed, the sum of the resulting plurality of absolute values is added, and the right head boundary is defined as the last maximum of said sum that is above a predetermined threshold value.
It is here also possible to smooth the function or functions of the horizontal projections with a low-pass filter prior to defining the head boundary.
According to a preferred development it might be intended in the determination of the head boundaries that iwo successive individual images or the difFerential images obtained therefrom are only used for determining the head boundaries if a change between the two consecutive individual images is within a predetermined range.
it is thereby easily possible to make sure even before the determination of the head boundaries that only those differential images are used for a determination in the case of which it is ensured that sufficiently reliable results fvrthe head boundaries can be obtained therefrom. Since the change between two consecutive individual images must be greater than a predetermined value (so that head boundaries can be calculated according to said preferred development), it is ensured that in both images a change going beyond normal noise and thus an actual movement of the person to be identified have taken place. Since the change between two consecutive individual images must be smaller than a predetermined value (for calculating. head boundaries), differential images in which an excessively strong movement is reflected, as is s.g. produced by shaking the recording means or an extremely fast movement of the person to be identified, are left out of consideration in determining the head boundaries.
According to a particularly advantageous development with respect to the necessary computing time, a motional intensity can be calculated for determining the change between two consecutive individual images. In the case of a gray-level image the motional intensity can substantially be calculated through the sum of the gray level of the differential Image. In a binarized differential image the motional intensity can expediently be calculated as the sum of the 1 pixel or 0 pixel of the binarized differential image.

According to a preferred development predetermined head boundaries can be used for extracting the head region in the case where no head boundaries can be determined.
According to a further advantageous development predetermined head boundaries can be used far extracting the head region if one of the head boundaries determined Is not within predetermined limits. Alternatively, the step for determining the head boundaries can be carried out with successive differential images.
It is thereby possible to perform an additional plausibility check of the boundaries determined in the method. In other words, it is checked whether in the region defined by said head boundaries the head of a person would actually have to be expected with respect to the arrangement of the device. If the head boundaries determined are not plausible, predetermined head boundaries are assumed in the end. Thanks to this additional check the quality of the detemiinatian of the head boundary and, in the final analysis, the quality of the verMcation of the authenticity can additionally be enhanced_ Expediently, in the above-described developments for determining the head boundaries, the lower or the right and lower head boundaries can be defined such that a square head region is extracted from the corresponding individual images or differential images.
According to a preferred development of ail of the previously described variants of the method of the invention a stabilized differential image which is evaluated for determining the intrinsic movements andlor for extracting the regions in which intrinsic movements are to be determined can be evaluated on the basis of two consecutive images.
This development can be used independently of and alternatively to the extraction methods for regions in which intrinsic movements are expected. Thus all of the advantages mentioned in cvnn~ction with the determination of the head boundaries could here be mentioned as an advantage.
On the other hand, said development can also be used for increasing the accuracy of the regions with Intrinsic movements, as determined in the extracting process.

A stabilization can here be carried out by means of a correlation correction.
According tv a predetermined development the correlation correction can be carried out by template matching which comprises calculating the differential image, in the case of which in calculating the differential image the first individual Image used for forming the differential image is shifted relative to the second individual image used for forming the differential Image in such a way that a correction function between the two individual images is maximum_ According to another advantageous development, two consecutive individual images or differential images obtained therefrom are only used in the above-described methods for determining the intrinsic movements if the change between the two consecutive individual images is within a predetermined range.
It is thereby possible to select consecutive individual images aiready at the beginning of the inventive method for obtaining sufficiently reliable results of the verification.
In particular, it is easily possible to make sure already before the determination of the intrinsic movements that for the determining process use is only made of differential images by which it is ensured that sufficiently reliable results can be obtained therefrom.
Since the change between two consecutive individual images must be greater than a predetermined value (so that according to said preferred development the attempt is made to determine an intrinsic movement), it is ensured that in the two images a change going beyond normal noise and thus an actual movement or an intrinsic movement of the person to be identified has taken place. Since the change between two consecutive individual images has to be smaller than a predetemnined value (for calculating head boundaries), differential images which reflect excessive rnovernents, as are e.g. caused by shaking the recording means or by an extremely fast movement of the person to be identified, are not used for determining the intrinsic movements.
Preferably, the change between two consecutive individual images can be calculated through a motional intensity. Like in the determination of the head boundaries, the motional intensity can be calculated for a gray-level image through the sum of the gray levels. 6y analogy, the motional intensity can expediently be calculated for a binary differential image through the sum of the 1 pixel or 0 pixel of the binarized differential image.
Further advantages and features of the method according to the invention will become apparent from the description of detailed embodiments of the invention with reference to the drawing, in which.
Fig. 1 is.a sequence diagram of a first embodiment of the method for verifying the authenticity of an image recorded in a person identifying process, according to the present invention;
Fig. 2 is a sequence diagram of a second embodiment of the method far verifying the authenticity according to the present invention;
Fig. 3 is a sequence diagram of a third embodiment of the method for verifying the authenticity according to the present invention;
Fig. 4 is a sequence diagram of a fourth embodiment of the method for verifying the authenticity according to the present invention;
Fig. 5 is a sequence diagram of a #'rfth embodiment of the method for verifying the authenticity according to the present invention;
Fig. 6 is a sequence diagram of a sixth embodiment of the method far verifying the authentiGty according to the present invention;
Fig. 7 is a sequence diagram for performing a correlation analysis which can be used within the scope of the fourth or frfth embodiment of the method for verifying the authenticity according to the present Invention;
Fig. 8 Is a sequence diagram of a seventh embodiment of the method for verifying the authenticity according to the present invention; and 1$
Fig. 9 is a sequence diagram of an eighth embodiment of the method for verifying the authenticity according to the present invention.
Fig. 1 shows a sequence diagram of an inventive method for verifying the authenticity of an image recorded in a person identrfying process and belonging to a person to be identified.
In a first step SiIQ a sequence of consecutive individual images or frames K(t), t=1, ..., n of a person is recorded. To this end, use is expediently made of prior-art video cameras which record the sequence in digitized form. Individugl images recorded with such a video camera are norrnalfy present in the form of pixels, each individual image comprising c x r pixels if c designates the number of the columns and r the number of the lines of the corresponding individual image.
In the following K(t) therefore stands for the whole individual Image at lima t and iC~~(t) for the pixel in the With column and the r-th row.
The individual images or frames are recorded in the inventive method at a speed of 8-12 imageslper second.
In step Sl2fl, two consecutive individual images K(t-1) and K(t) are chosen from the recorded sequence of individual images_ In the subsequent step 5130, the individual images (K(t-1) and K(t) are compared with each other to find out whether the person subjected to an identitlcation has performed intrinsic movements between two individual images K(t-1) and K(t).
To this end the two individual images K(t-1) and K(t) can be compared with each other pixel by pixel, and the change resulting from said comparison of pixels can be compared with a predetermined threshold value. Thus if the amount of change is above the predetermined threshold value, the person has performed an Intrinsic movement between two individual images K(t-1 ) and K(t), whereas if the amount of change is below the threshold value, no intrinsic movements of the person could be detected.
Said pixel-by-pixel comparison yields satisfactory results if no additional movements are superimposed an the intrinsic movements of the person, i.e. if the person does not move or only moves minimally with respect to the image recording means.
How the comparison has too be made in detail depends on the type of the consecutive individual images. If the individual images are e.g. present as gray-level images, the amount of the difference of the gray levels of two corresponding pixels in the individual image K(t-1 ) and K(t) can be regarded as a measure of the change in a pixel, and the sum of the amounts of said differences as a measure of the change between the two individual images K(t-1) and K(t) that are finally compared with the predetermined threshold value.
If the individual images are e.g. present in binarized form (l.e. in the form of white or black pixels), the change between two individual images follows from the sum of the drl-fering pixels corresponding to each other in the individual images K(t-1 ) and K(t).
Thus, in the case of binarized individual images said change has to be compared with a predetermined threshold value. The binariaation of gray-level images will be described in detail in connection with the second embodiment (step S240).
Finally, if the comparison in step S 130 shows that intrinsic movements of the person have taken place between the two individual images, the authenticity of the recorded individual images K(t-1 ) and K(t) is verified in step S 170.
8y contrast, if the comparison shows that the person has not pertormed intrinsic movements between the individual images K (t-1 ) and K(t), the authenticity of the recorded individual images K(t-1) and K(t) cannot be verified in step S180.
In this case it is possible to interrupt the method and if said method is used within the scope of an access control, access can be denied. Alternatively, as shown by step S190, it is possible to jump back to step S120, and the method for verifying the authenticity can be carried out again with two further individual images, e.g. K(t) and K(t+1 )_ The loop shown by step S190 can be passed through j times, j being a predetermined number which is n-1 at the most.
Alternatively to the comparison of the individual images K(t-1 ) and K(t), a differential image D(t) ~ K(t) - K(t-1 ) can be formed for reducing the amount of data to be processed, and finally the differential image D(t) can directly be analyzed to find out whether the person has performed intrinsic movements between the two individual images K(t-1 ) and K(t).
According to a further alternative of the first embodiment, instead of the differential image D(t) an AND operation of two consecutive differential images D(t) and D(t-1 ) can be evaluated for detecting the intrinsic movements. Said alternative is particularly useful when there are great changes in two consecutive individual images due to the release of a dominating background, for instance a source of illumination, in two consecutive individual images, and when these are reflected in the differential image.
Thus, the differential image to be assessed in the final analysis can be freed by said altemativ~
from changes that do not result from intrinsic movements.
Since in said alternative two differential images D(t) and D(t+1 ) are used, three consecutive individual images K(t-1 ). K(t) and K(t+1 ) are needed for producing said two differential images.
Fig. 2 shows a sequence diagram of a second embodiment of the method according to the invention.
Said second embodiment is predominantly an alternative to step 130 in the first embodiment.
Thus, according to the second embodiment, steps S110 and S120 are also passed through first, i.e. a sequence of consecutive individual images is recorded and two consecutive individual images are chosen from said sequence.

Following step S120 the differential image D(t) is determined in step S130 from the individual images K(t) and K{t-1 ).
In the opUonal step S240 said differential image can be binarlzed according to the following equation:
Ba(t) = O(D(t)-S). (1 s is here a predetermined threshold value, and 0(I) designates the step function, i.e., s(I) = 0 for I < 0, and A(I) = 1 far 1 >_ 0.
The threshold value s may be fixedly predetermined or determined by evaluation of the background.
In the last-mentioned case, the threshold value S should be defined as a function of the upper m lines of the differential image because said fines only show static background, and values different from 0 in the differential image are thus due to noise caused, for instance, by the recording device. Consequently, noise is suppressed if only gray-level values above said threshold value are considered. Thus the maximum value found in the uppermost m lines of the driferentiai image should be used as the threshold value, i_e.
= max Da (t), (2) C,fcm In the next step S250, at Least one region of enhanced movement is determined in the differential image D(t) or in the binarized differential image B(t).
To this end a masking frame of any desired size may be predetermined and guided over the corresponding differential image, and upon each shift the movement is calculated, i.e. the number of the sum of the pixels with the value 1 or, altematlvely, of the pixels with the value 0. The masking frame of the shift with the largest sum finally defines the region of enhanced movement After the region of enhanced movement has been determined in the differential image or the binarized differential image, it is checked whether said region corresponds to a predetermined region in which intrinsic movements are to be expected.
if this is the case the authenticity of the recorded individual images K(t-1) and K(t) underlying the differential image D(t} or the binarized d'rfferentlal image B(t) are verified in step 5170.
If the region of enhanced movement is not identical with the predetermined region, the authenticity of the recorded individual images cannot be verfied in step S180, and the method may be interrupted or repeated according to step S190 (see Fig. 1).
The region in which most of the intrinsic movements of a person have to be expected is the head region. In the head region as such, most of the intrinsic movements are to be expected in the mouth region, the cheek region, the nose region or the eye region.
Preferably, according to the second embodiment in i=ig. 2, the at least one determined region of enhanced movement is therefore compared with the mouth region, the cheek region, the nose region andlor the eye region.
As already implicitly pointed out, it is of course also possible to determine several regions of enhanced movement in the differential image and to check whether said determined regions of enhanced movement correspond to the mouth region, the cheek region, ttie nose region andlor the eye region.
Fig. 3 shows a sequence diagram of a third embodiment of the method of the invention, which illustrates an alternative to step S250 in the second embodiment, i.e.
the determination of the region of enhanced movement, and, moreover, corresponds to the second embodiment.
Hence, in the third embodiment a differential image D(t) or, alternatively, a binarized differential image B(t} is started from. The differential image D(t) yr ttae binarized differential image B(t) will be abbreviated in the following text as differential image K.

In step 5350, the function Yv of the products from the horizontal projections PN and the horizontal variances V" is determined on the basis of the differential image K, in formulae:
r 1 z Yv =PH~v" =~~K~y'~nc~~Ka~-K~~ ~, with KT _ ~ ~ Ka , (3) and nc = number of columns of K.
Furthermore, in step 5351, the function YK of the products from the vertical projections Pv and the vertical variances V" is determined from the differential image K
according to the following formula:
_ _ _ _1 z Y" Pv Vv ~~ K~~ ~ ( nr ~ (K~~ - ~~ ) ~, with K~ _ ~ ~ Ka, (4) nr and nr = number of columns of K.
In step S352, the functions Y~ and YH obtained according to the above formulae can be smoothed with a suitable low-pass fitter TPF{ }.
In functions Y" and YH or the smoothed functions TPF~Yv} and TPF~YH}, peaks are identfied in step 5353. The peaks of Yv and Y", respectively, define the vertical and horizontal extent and position of the region of enhanced movement Various definitions which are known in the prior art can be given as a relation between the peak of one of the functions and the corresponding region of enhanced movement_ For instance, the position of the con-esponding region can be indicated by the maximum value of a peak and the extent of said region by the FWHM (full width of half maximum) of the peak. Alternatively, the extent can also be indicated by the fuming points of the peak As a further alternative, the extent can also be indicated by the boundaries which lead to a pradeterrnined area below the peak Standard prior-art methods can be used for finding the individuai peaks. These methods substantially consist in looking for the sections of a function that exceed a predetermined threshold value.
The definition to be used for the peak and the best method for finding tf~e peak strongly depend on the specific intrinsic movement to be detected; the most advantageous definition for the peak and the most advantageous method for finding the peak can be detem~ined by way of tests for each individual case. It should be noted that a detailed description of a speafic definition for the peak and of the method used for finding the peak will be given in connection with the embodiment described in Fig_ 5.
Fig_ 4 is a sequence diagram showing a fourth embodiment of the method of the invention. Said fourth embodiment is an extension of the second and third embodiment.
It is presupposed for the pertormance of the method according to the fourth embodiment that a first region of enhanced movement has already been determined and that it has been checked whether said region corresponds to a predetermined region in which intrinsic movements are to be expected.
If such a region has been found, a further region of enhanced movement which is supposed to be in a symmetrical relationship with the first region of enhanced movement is determined according to the fourth embodiment in step 5461.
For determining said second region, step S250 (second embodiment) may be passed through, or steps S35rJ to 5353 (third embodiment).
It should here be noted that a symmetrical relationship between two regions in a differential image, i_e. a symmetrical relationship in the movements, must be expected when the regions resul~ng in the intrinsic movements are symmetrical with each other.
For instance, when the head region of a person to be Identified is considered, the mouth region in the left and right face halves, the cheek region in the left and right face halves, the nose region in the left and right face halves as well as the left and right eye, are suited as the first and second region within the scope of the fourth embodiment.
When two of said regions that are supposed to have a symmetrical relationship have been found in the differential image, it can be verfied by means of a correlation analysis whether or not the symmetrical relationship exists.
On account of the fact that the intrinsic movements, e.g_ in the head region of a person, are~symrnetrical, a plausibility check of the detemlined regions of enhanced movement.
is thus obtained through the fourth embodiment.
Hence, K it is verified in step S462 by way of a correlation analysis that the symmetrical relationship exists, the authenticity of the recorded individual images can be verified in step S1T0.
In the other case the authenticity cannot be verfied (see step 5180) and the method is thus interrupted with such a result or, as already discussed above, repeated.
The various possibilities of performing a con-elation analysis according to S462 will be explained in detail below in connection with Fig. 7.
fig. 5 shows the sequence diagram of a fifth embodiment of the method according to the invention. Steps S530 to S565, which are shown in Fig. 5, illustrate a preferred embodiment of step S130 according to the first embodiment.
Hence, a sequence of consecutive individual images of a person is first recorded in the fifth embodiment as well, and two consecutive individual images are then selected from said sequence. This is illustrated in Fig. 5 by step S12Q_ do the basis of said individual images, a differential image is determined in step S530 and optionally binarized in step 5540 (steps S530 and S540 correspond to steps and S240 in the second embodiment, so that a description is here not needed and reference is made to the corresponding description of steps S230 and 5240).

According to the fifth embodiment the vertical position and the extent as well as the horizontal position and the extent of the mouth region of a person to be identified are determined on the basis of the differential image in step S550.
tn contrast to step S250 in the second embodiment, in which regions of enhanced movements are determined in the differential image and a check is carried out whether said regions can be assigned to specific regions, a specific region in which intrinsic movements are expected, namely the mouth region, is. looked for in step S550.
It Is then checked in step 5560 whether or not the determined mouth region is within predetermined boundaries. If the mouth region is within predetermined boundaries, the autt-ienticity of the recorded individual images can be verified in step S170;
if the determined mouth region is not within predetermined boundaries, the verification of the authenticity of the individual images must be denied.
Moreover, as an alternative to step S56t3, a correlation analysis can be carried out in step S565 with the determined mouth region for detecting intrinsic movements.
Various possibilities of carrying out a correlation analysis will be explained in detail in connection with Fig. 7.
In a modification of the fifth embodiment, it is also possible to czny out steps S560 and S565 not alternatively, but cumulatively. Thus, according to step S170 the authenticity of the individual images wilt only be detected if the determined mouth region is within predetermined boundaries and if the presence of intrinsic movements follows from a correlation analysis of the mouth region.
Although the fifth embodiment specifically relates to the mouth region, it is of course possible to carry out the method with other regions in which intrinsic movements are expected, or with cambinativns of such regions.

Fig. 6 is a sequence diagram showing a sixth embodiment of the inventive method_ Said sixth embodiment is a preferred development of step S550 in the fifth embodiment for determining the vertical position and extent and the horizontal position and extent of tl~e mouth region on the basis of the differential Image. Thus, it is presupposed that according to step S530 a differential image 0(t) or according to 5540 a binarized differential image B(t) is present On the basis of said differential image, the function Yv of the products from the horizontal projections PH and the horizontal variances V" is calculated in step 650 according to equation (3).
The function Yv can optionally be smoothed with a low-pass filter.
The vertical position and the vertical extent of the mouth region are determined on the basis of said function or the smoothed function in step S651.
The vertical mouth position is here defined as the position of the last (lowermost) maximum of Yv that is greater than a predetermined threshold value. in the case of the mouth region twice the mean of Yv has fumed out to be a suitable threshold value.
The width of the peak in which the peak has dropped to a predetermined percentage of its maximum value can be indicated as the vertical extent of the mouth region.
50% has here fumed out to be an advantageous value_ As an alternative, the vertical extent of the mouth region can also be indicated as the width of the peak which follows from the distance of the two Gosest fuming points at the left and right side from the peak position.
Moreover, the values obtained for the vertical position and the vertical extent can be checked as to their plausibility. To this end a minimum mouth positron and a maximum mouth position can be predetermined, and 'tt can be checked whether the determined mouth position is within said two extreme values. Said procedure can be carried out alternatively or cumulatively for the vertical extent of the mouth region.

The function YH;t,b of the products from the vertical projections Pv~.b and the vertical variances vv;t,b in a horizontal strip extending from the upper mouth boundary t determined in step 651 to the lower mouth boundary b, namely n ''' 1 e, z ~H~,b = PVa.b ' VH;t.b - ~~ KCrJ ' ~ ~ r~, ~~or - KC) b with KQ (t) _ ~ ~ K~r, n = b - t + 1, (5) n r.l and nr = number of columns of K~
is tfien calculated in the next step 652.
The function YH;tb calculated in this way can also be smoothed in an optimum manner.
In the next step Sfi53, the horizontal position and extent of the mouth region are detem~ined.
To this end the absolute value of the derivative of tha function YH:m or of the smoothed function TPF{YHx,b} is determined, i.e.
bHt,b ~~yH+~;t,b - yN+1 t,b~, (S) The left horizontal boundary of the mouth region is then defined as the first maximum of bN that is greater than a predetermined threshold value. The right horizontal boundary of the mouth region is defined accordingly as the last maximum of bN that is greater than said threshold value. In the instant case the mean of bH has been used as the threshold value.
In the case of the horizontal extent of the mouth a plausibility check can also be carried out such that the determined mouth region has to lie within a predetermined maximum region so that the region of enhanced movement is regarded as the mouth region.

Fig. 7 shows a sequence diagram of a correlation analysis, as can e.g. be carried out in step 5462 according to the fourth embodiment or in step S56b according to the fifth embodiment_ Although the procedure according tv the sequence diagram in Fig. 7 is explained with reference to the functions Yv and YH, it can readily be applied to the regions as determined in the fourth embodiment.
In step 5766, function Yv andlor YH is split up into sections which are supposed to be in a symmetrical relationship with one another.
In step S767, it is checked by means of a correlation analysis whether there is the symmetrical relationship assumed in step S76fi.
Steps S766 and 5767 can of course also be carried out for several symmetrical relationships.
If it can be verified with the correlative analysis that there islare the symmetrical relationship(s), the authenticity of the recorded individual images can be verified according to S170.
If the symmetrical relationship cannot be verified with the correlation analysis, the authenticity of the individual images cannot be verified either, and the method may be Interrupted or, altematavely, be repeated for a new pair of individual images.
The following correlations can be checked in step S787 for the already discussed case regarding the detem~ination of intrinsic movements in the mouth region.
Qn the one hand, an intrinsic mouth movement is symmetrical in itself.
Therefore, the correlation of curve YH and of the smoothed curve TPF(YH}, respectively, with the mirrored curve Y" and the mirrored smoothed curve TP1={YH}, respectively, must be greater than a predetemlined value, Expediently, prior to the correlation analysis curve YH iS transformed to standardized variables so that the function YH has the mean value 0 and the.standard deviation 1.
When said standardized variables are designated as n(Y",v), the correlation is calculated according to the formula ncl2-y gtn _ ~, nm+c (YH ) - nnra (Ytt )~
o._~clz where n~(Y,~) is continued for indices c ~ 0 and c ~ n~ with 0.
ft has been found in the case of a mouth movement that a value of gm ~ 0,9 corresponds to a sufficiently symmetrical mouth movement. Thus, the expected symmetrical relationship would be confirmed if gm > 0,9 followed from equation (7).
As an alternative ar in addition to the above~fescribed correlation analysis, it is possible to check the correlation of the image section from the differential image of the mirrored mouth with the mouth itself. When said correlation is smaller than a predetermined value.
it can be assumed that an actual intrinsic movement has not taken place. but the determined movement is due to movements, noise, or the like.
The correlation of the differential image of the mirrdred mouth with the mouth Itself can be determined by the expression h e~ ~~K .~.n ~~K
m - ~+o v cnt-c cr ~ (s) ~_,rz r=t r.t where cnt is the horizontal position and a the width of the mouth and can be determined by cnt = arg m ax gm (9) m It has been found that a value of hM=0.5 is suited for discriminating between intrinsic and non-intrinsic movements_ as Together with the above-described correlation analyses or aitematively, a further correlation analysis is provided for in steps 5768 and S769.
For carrying out said correlation analysis function Y" andlor YN can be split up into predetermined sections in step S768. Said step, however, is only provided for the case that specfic regions of the corresponding function are to be analyzed. If the function Y~
andlor YH is to be analyzed in its entirety, step 5768 can consequently be dispensed With.
In step 5769, it is checked by means of a correlation analysis whether the values of the corresponding function within their range of definition (or in the predetermined sections if step 5768 is carried out) are substantially irregular.
If the values are irregular, the conclusion can be drawn that an intrinsic movement is actually concerned. By contrast, in the case of regular values of the function the conclusion can be drawn that the movements determined are not movements but rather noise, or the like.
for carrying out the correlation analysis according to step S769, correlations between the vertical projections and vertical variances for the right and left side of the mouth can be checked in the case of me mouth region.
The correlation between the vertical projection and the vertical variance for the right and left side of the mouth can be calculated by the expressions eLr ~~ ~ b 2~
~m - nant+c nw~l. C ,~, ~~d - Ko -_o r=t r., ~' 1 r z jm = ~ nerr+oCL Karl Wce~+cCL (Ker - ~) ~_ ~=-on r=c r=t .__ .... .. ,.__ ~vJ~

1n the case of the mouth region it has been found that the correlations im and jm between the vertical projection and the vertical variance must be < 0 to ensure that the detected movement is an intrinsic movement.
In the methods according to Figs. 5 to 7 the mouth region is shown as an exemplary embodiment of the method according to the invention. However, as aln=_ady explained in connection with the second embodiment, the method of tt-ie invention is not limited to the mouth region.
Rather any regions in which intrinsic movements have to be expected, e.g. the cheek region, the nose region and the eye region, or any combinations of said regions, are possible.
If a region which differs from the mouth region is to be analyzed as to intrinsic movements, only the regions in which the functions are formed, the can-elations which are analyzed within the correlation analysis, as well as the threshold values with which a correlation is verified, have to be adapted accordingly to the regions considered.
Since the principles according to which such an adaptation has to be carried out are obvious by virtue of the above-explained example of the mouth region and the fundamentals found in standard textbooks on correlation analysis, a detailed descrfptlon of the adaptation to the ether above-mentioned regions is dispensed with.
Fig_ 9 shows a sequence diagram of an eighth embodiment of the method according to the invention. The steps of the eighth embodiment serve to optimize the individual images after recording for determining the intrinsic movements.
Thus, as shown in Fig. 8, two consecutive individual images are chosen in step after a sequence of consecutive individual images has been recorded in step S110.
In step 5825, it is first of all checked whether or not the change between the two consecutive individual images is within a predetermined range.

Since the change between two consecutive individual images must be greater than a predetermined value, it is ensured that in both individual images a change going beyond normal noise has taken place and thus an actual movement or an intrinsic movement of the person to be identified.
On the other hand, since the change iJetween two consecutive individual images must be smaller than a predetermined value, it is ensured that differential images in which an excessively strong movement is reflected, as is e.g. caused by shaking the recording means or by an.extremely fast movement of the person to be identified, are not used for determining the intrinsic movement because in such a case a distinction between an actual movement and an intrinsic movement can no longer be made in a sufficiently reliable manner.
Step 5825 can be carried out in that a motional intensity is calculated for determining the change between two consecutive individual images.
In case the differential images are gray-level images, the mofional intensity can substantially be calculated through the sum of the gray level of the differential image. In case the difFerential image is binarized, the motional intensity is expediently calculated through the sum of the 1 pixel or D pixel of the binarized differential image.
In step S826, which can be carried out alternatively or cumulatively with respect to step S825, the regions in the individual images in which intrinsic movements are expected are extracted from the corresponding individual images before the individual images are compared with each other. The amount of data to be processed can thus be reduced considerably-A region which can expedienby be extracted because a multitude of intrinsic movements take place therein is the head region.
Thus for extracting purposes a predetermined section in the individual images where according to experience the head of the person to be identified comes to rest can be extracted.

With reference to Fig. 9, an alternative to said method shall be described below.
According to said aitemative the head region of the person to be identified is determined by determining the head boundaries from a differential image. With said determined head boundaries the head region can then finally be extracted from the respective individual images.
Again alternativeiy or cumulatively with respect to steps S826 and S825, respecfively, the individual images can be stabilized in step S827.
Again alternatively or cumulatively with respect to steps 5826 andlor step 825, two or more individual images can be stabilized with respect to each other in step 827 or, in other words, aligned with respect to each other.
To this end a residual shift (I,m) between K(t) and K(t-1 ) can be determined by template matching. For this purpose K(t) and K(t-1) are first superimposed in centered fashion (this is squat to a residual shift (I,m) _ (0,0)). The standardized correlation c(0,0) is then determined. In the next step the standarriized correiations c(I,m) are calculated in dependence upon a variation of the residual shift in the vicinity of (I,m) =
(0,0).
The carrelatio~ correction (x,y) finally follows from (x, Y) = argmax c(~ y)- (11) With this residual correction a stabilized differential image D(t) can finally be determined according to Daft) = K~-x.r-r(t -1) - ~r(t) (12) Fig. 9 is a sequence diagram of the already above-mentioned alternative for determining the head boundaries with the help of which the head region in which most of the intrinsic movements are expected can be extracted from individual images K(t-1) and K(t).

Starting from two selected consecutive individual images K(t-1 ) and K(t), the differential , image D(t) is first determined in step S92'I _ In step S922, the upper head boundary is then determined from the difrerential image p(t). To this end the horizontal projection s"(t) is first determined according to sH(t) _ ~,p«(t) (1 In this context it should be remembered that c designates the columns and r the lines of the differer~tlal Image.
optionally, function sH(t) can be smoothed by filtering with a suitable low-pass ftlter TPF.
The absolute value of the first derivative is then formed from the function sH(t) or the smoothed function TPF{s"(t)}, respectively bH(t) = I s~.,(t)-sh,(t) I (1 ~'he position of the first maximum of b" that is greater than a predetermined threshold value, here half the mean of bH, is now determined as the upper head boundary.
If such a value is not found, the calculation of the upper head boundary can be repeated or set to a predetem~ined value which has e.g. been determined for the individual images K(t-2) and K(t-1 ).
Finally, in step 5293 the left and right head boundaries are determined. To this end the differential image can be divided into q successive strips which extend in vertical direction downwards, starting from the upper head boundary t.
In each strip the vertical projection of the diffierential image is then calculated according to t+(q+~~Z-~
5~ayt) = ~ D~r(t) ~ (15) r.~+qs As in the case where the upper head boundary has been calculated, the function s;,q~(t) can be smoothed by filtering with a suitable low-pass filter TPF.
In the next step the absolute value of the first derivative of the function s,~,q~(t) and the smoothed function TPF{ s;,q'(t) } is fomled and subsequently summed over q:
bY (t) _ ~ Ig~q i (t) _ g~9)~(t~
'I0 Of this function by all maxima are determined that are greeter than a predetermined threshold value, here half the mean bv.
The first maximum (the maximum farthest to the left) is defined as the position of the left head boundary. The last maximum (the. maximum farthest to the right) is defined as the right head boundary.
If function by has less than two of such maxima, the determination of the left and right head boundaries for the next following differential image can be carried nut.
Alternatively, the left and right head boundaries can be replaced by predetermined values which have e.g_ followed from the determination of the head boundary with the help of the differential image D(t-1 ).
Finally, in step S824, the lower head boundary is determined with the help of the upper, left and right head boundaries. To this end the width of the extracted region is determined from the difference between the right head boundary and the left one. The lower head boundary can be adjusted such that a square region is obtained on the whole.

Apart from the variant described in Fig. 9 for determining the head boundaries, alternative embodiments are possible.
For Instance, it is possible to determine only the upper and lest head boundaries according to the above~iescribed procedures. The tower and right head boundaries can then be calculated by way of a predetermined width and a predetermined height, which typically correspond to the head region in the individual images.
Moreover, the head boundaries can also be determined from a differential image which has been obtained as an AND operation from two consecutive differential images.
Furthermore, it can optionally be checked in the embodiment shown in Fig. 9 whether the change between the individual images K(t-~t} and K(t) is within a predetermined range.
Since the same procedures as explained in connection with step S825 can be used far this purpose, a description is here not needed, and reference is just made tv the description regarding step 5825.

Claims

Claims 1. A method for verifying the authenticity of an image recorded in a person identifying process and belonging to a person to be identified, the method comprising the steps of:
recording a sequence of consecutive individual images of the person, and determining the authenticity of the recorded image if in at least two consecutive individual images of the sequence intrinsic movements are detected.

2. The method according to claim 1, wherein determination of the intrinsic movements comprises evaluating at least one differential image of two consecutive individual images of the sequence.

3. The method according to claim 2, wherein for the determination of the intrinsic movements a differential image is evaluated that results from an AND operation of two consecutive differential images.

4. The method according to claim 2, wherein each differential image is binarized prior to evaluation.

5. The method according to claim 4, wherein binarization is carried out by means of a threshold value which is determined by evaluating the background.

6. The method according to claim 2, wherein in each differential image at least one section of enhanced movement is selected, and intrinsic movements are detected if said at least one section in the differential image is consistent with a predetermined region in which intrinsic movements are to be expected.

7. The method according to claim 6, wherein the section of enhanced movement has a vertical and a horizontal extent.
wherein the vertical extent is substantially determined by a peak of the function of the products from the horizontal projections with the horizontal variances of the differential image, and the horizontal extent is substantially determined by a peak of the function of the products from the vertical projections with the vertical variances of the differential image.

8. The method according to claim 7, wherein each function is smoothed with a low-pass filter prior to the determination of the corresponding peak.

9. The method according to claim 6, wherein sections are chosen from the differential image that are expected to have a symmetrical relationship with one another, and intrinsic movements are detected when the symmetrical relationship is verified by a correlation analysis of the corresponding sections.

10. The method according to claim 6, wherein intrinsic movements are detected if it is determined by way of a correlation analysis that the enhanced movement in the region of enhanced movement is irregularly distributed over said region.

11. The method according to claim 2, wherein detection of intrinsic movements further comprises evaluating:
a function which depends on the vertical position and which is determined for a given vertical position through the product from the horizontal projection with the horizontal variance of the differential image, and a function which depends on the horizontal position and which is determined for a given horizontal position through the product from the vertical projection with the vertical variance of the differential image.

12. The method according to claim 11, wherein each function is smoothed with a low-pass filter.

13. The method according to claim 11, wherein at least one peak is determined in the function depending on the vertical position, and at least one peak is determined in the function depending on the horizontal position, and intrinsic movements are detected if said peaks in the vertical and horizontal direction are within predetermined limits that are given by the vertical and horizontal extent of at least one region in which intrinsic movements are to be expected.

14. The method according to claim 11 further comprising evaluating, on the basis of the function depending on at least one of the vertical position and the function depending on the horizontal position, sections that are expected to be in a symmetrical relationship, and detecting intrinsic movements if the symmetrical relationship is verified by a correlation analysis of the corresponding sections.

15. The method according to claim 14, wherein the sections which are expected to be in a symmetrical relationship are standardized with respect to each other prior to the correlation analysis.

16. The method according to claim 11 further comprising evaluating at least one of the function depending on the vertical position and the function depending on the horizontal position, and detecting intrinsic movements if it is determined by way of a correlation analysis that the values of at least one of the function depending on the vertical position and of the function depending do the horizontal position are irregular over a predetermined region.

17. The method according to claim 1, wherein the intrinsic movements to be determined are intrinsic movements in the head region of the person to be identified.

18. The method according to claim 17, wherein the intrinsic movements to be determined are intrinsic movements in at least one of the group consisting of the mouth region, the cheek region, the nose region, and the eye region of the person to be identified.

19. The method according to claim 17, wherein the intrinsic movements to be determined are intrinsic movements in at least two regions that are in symmetrical relationship with each other in the head region of the person to be identified, and said at least two symmetrical regions are used for determining the axis of symmetry of the head region.

20. The method according to claim 19, further comprising producing with the help of the axis of symmetry found, an image which can be used far identifying a person and in which a frontoparallel-rotated head region is transformed into a head region with a rotated axis of symmetry.

21. The method according to claim 19, further comprising producing an image of the person to be identified with the help of the axes of symmetry found, said image being useable for identifying the person and being composed of the left head region and the mirrored left head region or of the right head region and the mirrored right head region.

22. The method according to claim 1, further comprising prior to the detection of the authenticity extracting regions from the individual images in which intrinsic movements are expected.

23. The method according to claim 22, wherein extracting the regions comprises evaluating at least one differential image of two consecutive individual images of the sequence.

24. The method according to claim 23, wherein extracting the regions comprises evaluating a differential image that results from an AND operation of two consecutive differential images.

25. The method according to claim 23, wherein each differential image is binarized prior to evaluation.

26. The method according to claim 25, wherein binarization is carried out by means of a threshold value which is determined by evaluating the background.

27. The method according to claim 22, wherein the head region of the person to be identified is extracted.

28. The method according to claim 27, wherein the extracted head region for identifying the person is transformed to a predetermined standard size.

29. The method according to claim 27, wherein extracting the head region comprises determining at least two head boundaries in the corresponding individual images, on the basis of which the head region is extracted from the corresponding individual images.

30. The method according to claim 29, wherein determining the head region in the corresponding individual images comprises determining the upper and the left head boundary.

31. The method according to claim 29, wherein determining the head boundary in the corresponding individual images comprises determining the upper, the left and the right head boundary.

32. The method according to claim 29, wherein each head boundary is defined by a head boundary line which extends such that the contour of the head is positioned substantially entirely within the head boundary lines.

33. The method according to claim 30, wherein for the determination of the upper head boundary comprises:
determining a function of the vertical projections of a binarized differential image, and defining the upper head boundary by the first maximum of the absolute value of the first derivative of said function that is above a predetermined threshold value.

34. The method according to claim 33, wherein the function of the vertical projections is smoothed with a low-pass filter prior to defining the head boundary.

35. The method according to claim 30, wherein determination of the left head boundary comprises:
determining a function of the horizontal projections of a binarized differential image, and defining the left head boundary by the first maximum of the absolute value of the first derivative of said function that is above a predetermined threshold value.

38. The method according to claim 30, wherein determination of the left head boundary comprises:

dividing a binarized differential image into a plurality of vertically successive strips, the first strip being vertically downwardly adjacent to the upper head boundary determined, determining in each strip the function of the horizontal projections of the binarized differential image, forming the absolute values of the first derivatives of the resulting plurality of functions of the horizontal projections, adding the sum of the resulting plurality of absolute values, and defining the left head boundary as the first maximum of said sum that is above a predetermined threshold value.

37. The method according to claim 35, wherein the function or functions of the horizontal projections is/are smoothed with a low-pass filter prior to defining the head boundary.

38. The method according to claim 30, wherein determination of the right head boundary comprises:
determining a function of the horizontal projections of a binarized differential image, and defining the right head boundary by the last maximum of the absolute amount of the first derivative of said function that is above a predetermined threshold value.

39. The method according to claim 30, wherein determination of the right head boundary comprises:
dividing a binarized the differential image into a plurality of vertically successive strips, the first strip being vertically downwardly adjacent to the upper head boundary determined, determing in each strip the function of the horizontal projections of the binarized differential images, forming the absolute values of the first derivatives of the resulting plurality of functions of the horizontal projections, adding the sum of the resulting plurality of absolute values, and defining the right head boundary as the last maximum of said sum that is above a predetermined threshold value.

40. The method according to claim 38, wherein the function or functions of the horizontal projections is/are smoothed with a low-pass filter prior to defining the head boundary.

41. The method according to claim 29, wherein two consecutive individual images are only used for determining the head boundaries if a change between the two consecutive individual images is within a predetermined range.

42. The method according to claim 41, wherein determination of the change between two consecutive individual images includes calculating a motional intensity.

43. The method according to claim 42, wherein the motional intensity is substantially calculated through the sum of the gray level of a differential image obtained from the two consecutive individual images.

44. The method according to claim 42, wherein the motional intensity is substantially calculated through the sum of the 1 pixel or 0 pixel of the binarized differential image.

45. The method according to any one of claims 29 to 44, wherein predetermined head boundaries are used for extracting the head region if no head boundaries can be determined.

46. The method according to claim 29, wherein predetermined head boundaries are used for extracting the head region if one of the head boundaries determined is not within predetermined limits.

47. The method according to claim 30, wherein one of the lower and the right and lower head boundary is defined such that a square head region is extracted from at lest one of the corresponding individual images and differential images.

48. The method according to claim 1, wherein a stabilized differential image is formed from two consecutive individual images, said image being evaluated for one of detecting the intrinsic movements and extracting the regions in which intrinsic movements are to be detected.

49. The method according to claim 48, wherein stabilization is carried out by means of a correlation correction.

50. The method according to claim 49, wherein the correlation correction comprises template matching.

51. The method according to claim 50, wherein template matching comprises calculating a differential image, the first individual image used for forming the differential image being shifted in the calculation of the differential image with respect to the second individual image used for forming the differential image in such a manner that a correlation function between the two individual images is maximum.

52. The method according to claim 12, wherein two consecutive individual images are only used for determining the intrinsic movements if a change between the two consecutive individual images Is within a predetermined range.

53. The method according to claim 52, wherein a motional intensity is calculated for determining the change between two consecutive individual images.

54. The method according to claim 53, wherein the motional intensity is substantially calculated through the sum of the gray level of the differential image.

55. The method according to claim 53, wherein the motional intensity is substantially calculated through the sum of the 1 pixel or 0 pixel of the binarized differential image.
CA002323867A 1998-03-12 1999-03-11 Method for verifying the authenticity of an image recorded in a person identifying process Abandoned CA2323867A1 (en)

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
DE19810792.7 1998-03-12
DE19810792A DE19810792A1 (en) 1998-03-12 1998-03-12 Personal identity verification method for access control e.g. for automatic banking machine
PCT/EP1999/001600 WO1999046737A1 (en) 1998-03-12 1999-03-11 Method for verifying the authenticity of an image recorded during a personal identification process

Publications (1)

Publication Number Publication Date
CA2323867A1 true CA2323867A1 (en) 1999-09-16

Family

ID=7860678

Family Applications (1)

Application Number Title Priority Date Filing Date
CA002323867A Abandoned CA2323867A1 (en) 1998-03-12 1999-03-11 Method for verifying the authenticity of an image recorded in a person identifying process

Country Status (6)

Country Link
US (1) US6922478B1 (en)
EP (1) EP1062640B1 (en)
JP (1) JP2002507035A (en)
CA (1) CA2323867A1 (en)
DE (2) DE19810792A1 (en)
WO (1) WO1999046737A1 (en)

Families Citing this family (28)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2001088857A1 (en) 2000-05-16 2001-11-22 Swisscom Mobile Ag Biometric method for identification and authorisation
GB0118020D0 (en) * 2001-07-24 2001-09-19 Memco Ltd Door or access control system
US7039224B2 (en) * 2002-04-29 2006-05-02 Activcard Ireland Limited Method and device for preventing false acceptance of latent fingerprint images
JP4013684B2 (en) * 2002-07-23 2007-11-28 オムロン株式会社 Unauthorized registration prevention device in personal authentication system
KR100480781B1 (en) * 2002-12-28 2005-04-06 삼성전자주식회사 Method of extracting teeth area from teeth image and personal identification method and apparatus using teeth image
EP1671257A1 (en) * 2003-09-30 2006-06-21 Koninklijke Philips Electronics N.V. System and method for adaptively setting biometric measurement thresholds
US7503488B2 (en) * 2003-10-17 2009-03-17 Davis Bruce L Fraud prevention in issuance of identification credentials
US7225977B2 (en) * 2003-10-17 2007-06-05 Digimarc Corporation Fraud deterrence in connection with identity documents
JP4085959B2 (en) * 2003-11-14 2008-05-14 コニカミノルタホールディングス株式会社 Object detection device, object detection method, and recording medium
US7356165B2 (en) * 2004-03-09 2008-04-08 Matsushita Electric Industrial Co., Ltd. Imaging apparatus performing a high reliability data analysis
US7386369B1 (en) * 2004-08-09 2008-06-10 Graftech International Holdings Inc. Digital electrode observation
JP4516516B2 (en) * 2005-12-07 2010-08-04 本田技研工業株式会社 Person detection device, person detection method, and person detection program
WO2007099834A1 (en) * 2006-03-01 2007-09-07 Nec Corporation Face authentication device, face authentication method, and program
JP4816321B2 (en) 2006-08-14 2011-11-16 ソニー株式会社 Authentication apparatus, authentication method, and program
JP2009088801A (en) * 2007-09-28 2009-04-23 Hitachi Software Eng Co Ltd Motionlessness deciding method and device for photograph object
JP5020920B2 (en) * 2008-10-29 2012-09-05 株式会社豊田中央研究所 Pedestrian detection device and program
DE102008055884A1 (en) 2008-11-03 2010-05-06 Cross Match Technologies Gmbh Method for detecting two-dimensional representation of face of person for generating identification document, involves generating signal for displaying representation when intensity is greater than multiple of algorithm and black value
JP5159950B2 (en) * 2009-05-28 2013-03-13 株式会社東芝 Image processing apparatus, method, and program
JP4609805B2 (en) * 2009-05-28 2011-01-12 Necカシオモバイルコミュニケーションズ株式会社 Terminal device and program
US20120076368A1 (en) * 2010-09-27 2012-03-29 David Staudacher Face identification based on facial feature changes
US9582705B2 (en) 2014-08-31 2017-02-28 Qualcomm Incorporated Layered filtering for biometric sensors
US9195879B1 (en) 2014-08-31 2015-11-24 Qualcomm Incorporated Air/object determination for biometric sensors
US9665763B2 (en) * 2014-08-31 2017-05-30 Qualcomm Incorporated Finger/non-finger determination for biometric sensors
US9619723B1 (en) 2016-02-17 2017-04-11 Hong Kong Applied Science and Technology Research Institute Company Limited Method and system of identification and authentication using facial expression
WO2019005825A1 (en) 2017-06-26 2019-01-03 Rank One Computing Corporation Method and system for implementing image authentication for authenticating persons or items
US10061996B1 (en) * 2017-10-09 2018-08-28 Hampen Technology Corporation Limited Face recognition method and system for personal identification and authentication
US11166077B2 (en) 2018-12-20 2021-11-02 Rovi Guides, Inc. Systems and methods for displaying subjects of a video portion of content
JPWO2021075198A1 (en) * 2019-10-17 2021-04-22

Family Cites Families (23)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE2950660A1 (en) 1979-12-15 1981-07-02 Alfred Teves Gmbh, 6000 Frankfurt Disc brake calliper assembly - has heavily-stressed portions made of material with high modulus of elasticity
DE3633360A1 (en) * 1986-10-01 1988-04-14 Heinrich Prof Dr Ing Reents Method together with the associated devices for safeguarding against the unauthorised use of payment accounts, premises and appliances
US4843377A (en) * 1987-04-21 1989-06-27 Guardian Technologies, Inc. Remote confinement system
DE3807727A1 (en) * 1988-03-09 1989-09-21 Robot Foto Electr Kg MONITORING DEVICE FOR SECURING AUTOMATES
DE4009051A1 (en) 1990-03-21 1991-09-26 Diehl Gmbh & Co Biometric identification and access monitoring system - compares acquired, processed thermal image of person's face with stored reference images
US5150432A (en) * 1990-03-26 1992-09-22 Kabushiki Kaisha Toshiba Apparatus for encoding/decoding video signals to improve quality of a specific region
JPH0411466A (en) * 1990-04-28 1992-01-16 Sony Corp Noise reducer
DE69131350T2 (en) 1990-07-17 1999-12-30 British Telecomm METHOD AND DEVICE FOR IMAGE PROCESSING
DE4028191A1 (en) * 1990-09-05 1992-03-12 Philips Patentverwaltung CIRCUIT ARRANGEMENT FOR DETECTING A HUMAN FACE
JPH0546743A (en) * 1991-08-09 1993-02-26 Matsushita Electric Ind Co Ltd Personal identification device
IT1257073B (en) 1992-08-11 1996-01-05 Ist Trentino Di Cultura RECOGNITION SYSTEM, ESPECIALLY FOR THE RECOGNITION OF PEOPLE.
US6181805B1 (en) * 1993-08-11 2001-01-30 Nippon Telegraph & Telephone Corporation Object image detecting method and system
DE4406020C1 (en) * 1994-02-24 1995-06-29 Zentrum Fuer Neuroinformatik G Automatic digital image recognition system
US6101264A (en) 1994-03-15 2000-08-08 Fraunhofer Gesellschaft Fuer Angewandte Forschung E.V. Et Al Person identification based on movement information
DE4413788C1 (en) * 1994-03-15 1995-10-12 Fraunhofer Ges Forschung Personal identification with movement information
JPH09510636A (en) 1994-03-24 1997-10-28 ミネソタ マイニング アンド マニュファクチャリング カンパニー Biometric personal identification system
US5625704A (en) * 1994-11-10 1997-04-29 Ricoh Corporation Speaker recognition using spatiotemporal cues
US5995153A (en) * 1995-11-02 1999-11-30 Prime Image, Inc. Video processing system with real time program duration compression and expansion
US5761329A (en) * 1995-12-15 1998-06-02 Chen; Tsuhan Method and apparatus employing audio and video data from an individual for authentication purposes
DE19610066C1 (en) * 1996-03-14 1997-09-18 Siemens Nixdorf Advanced Techn Process for the collection of face-related personal data and their use for the identification or verification of persons
DE19641000B8 (en) * 1996-10-04 2005-08-25 Viisage Technology Ag Method and arrangement for the automatic detection of the number of persons in a security gate
US6678393B1 (en) * 1997-12-23 2004-01-13 Intel Corporation Image selection based on image content
US6421453B1 (en) * 1998-05-15 2002-07-16 International Business Machines Corporation Apparatus and methods for user recognition employing behavioral passwords

Also Published As

Publication number Publication date
EP1062640A1 (en) 2000-12-27
EP1062640B1 (en) 2005-03-02
DE19810792A1 (en) 1999-09-16
JP2002507035A (en) 2002-03-05
WO1999046737A1 (en) 1999-09-16
US6922478B1 (en) 2005-07-26
DE59911699D1 (en) 2005-04-07

Similar Documents

Publication Publication Date Title
CA2323867A1 (en) Method for verifying the authenticity of an image recorded in a person identifying process
KR100831122B1 (en) Face authentication apparatus, face authentication method, and entrance and exit management apparatus
KR101546137B1 (en) Person recognizing device and method
EP1357520A2 (en) Method for verifying users and updating database, and face verification system using the same
KR101159164B1 (en) Fake video detecting apparatus and method
JP6601513B2 (en) Information processing device
JP5955031B2 (en) Face image authentication device
US20230015771A1 (en) Methods for detecting phantom projection attacks against computer vision algorithms
KR20030091345A (en) Method and system for facial liveness test
CN111066023A (en) Detection system, detection device and method thereof
JP2019053674A (en) Collation system
EP2299388A2 (en) Apparatus and method for detecting face
JP4788319B2 (en) Opening and closing eye determination device and method
JP7006668B2 (en) Information processing equipment
RU2175148C1 (en) Method for recognizing person identity
KR102211903B1 (en) Method And Apparatus for Photographing for Detecting Vehicle Occupancy
KR100293897B1 (en) A method for recognizing a face in a bank transaction system
KR100920937B1 (en) Apparatus and method for detecting motion, and storing video within security system
JP2019194875A (en) Collation system
JP6915673B2 (en) Information processing system
CN115396658A (en) Camera tampering detection
JP2015046126A (en) Vehicle detector
CN106447862A (en) Intelligent gate ticket-checking method based on computer vision technique
JP5361014B2 (en) Traffic monitoring system
KR20080028700A (en) Method for preserving of a public peace by means of a face recognition, and a face recognition apparatus

Legal Events

Date Code Title Description
EEER Examination request
FZDE Discontinued