Search Images Maps Play YouTube News Gmail Drive More »
Sign in
Screen reader users: click this link for accessible mode. Accessible mode has the same essential features but works better with your reader.

Patents

  1. Advanced Patent Search
Publication numberUS20050018923 A1
Publication typeApplication
Application numberUS 10/848,815
Publication dateJan 27, 2005
Filing dateMay 18, 2004
Priority dateMay 19, 2003
Also published asDE60314851D1, EP1482724A1, EP1482724B1, US7778483, US20080089583
Publication number10848815, 848815, US 2005/0018923 A1, US 2005/018923 A1, US 20050018923 A1, US 20050018923A1, US 2005018923 A1, US 2005018923A1, US-A1-20050018923, US-A1-2005018923, US2005/0018923A1, US2005/018923A1, US20050018923 A1, US20050018923A1, US2005018923 A1, US2005018923A1
InventorsGiuseppe Messina, Sebastiano Battiato, Alfio Castorina, Laurent Plaza
Original AssigneeStmicroelectronics S.A., Stmicroelectronics S.R.I.
Export CitationBiBTeX, EndNote, RefMan
External Links: USPTO, USPTO Assignment, Espacenet
Digital image processing method having an exposure correction based on recognition of areas corresponding to the skin of the photographed subject
US 20050018923 A1
Abstract
A digital image processing method includes extracting chromatic information of an image taken by an image taking device and related to a human subject; detecting visually interesting regions; and exposure correcting of the taken image by normalizing a grey scale of the taken image based on the visually interesting regions. Advantageously, the method includes recognizing areas corresponding to the skin of the subject, these areas being used as the visually interesting regions for the exposure correction step.
Images(11)
Previous page
Next page
Claims(23)
1. A digital image processing method, comprising the steps of:
extracting chromatic information of an image taken by an image taking device and related to a human subject;
detecting visually interesting regions in the taken image by recognizing areas corresponding to skin of the subject, wherein the recognized areas are the visually interesting regions; and
correcting an exposure of said taken image by normalizing a grey scale of said taken image based on said visually interesting regions.
2. A digital image processing method according to claim 1, wherein said recognizing step comprises constructing a probabilistic slicing of said image taken in a YcrCb format to evaluate if pixels of said image must be classified as belonging to said areas corresponding to the skin of said subject.
3. A digital image processing method according to claim 2, wherein pixels with higher grey values are classified as belonging to said areas corresponding to the skin of said photographed subject.
4. A digital image processing method according to claim 1, wherein said recognizing step comprises applying a threshold area of said image taken in an RGB format to evaluate if pixels of said image must be classified as belonging to said areas corresponding to the skin of said subject.
5. A digital image processing method according to claim 4, wherein applying said threshold area comprises constructing a chrominance slicing histogram of said taken image.
6. A digital image processing method according to claim 5, wherein constructing said chrominance slicing histogram uses normalized channels r and g of the type:

r=R/(R+G+B)
g=G/(R+G+B)
R, G and B being red, green and blue values of each pixel of said taken RGB image.
7. A digital image processing method according to claim 5, wherein said recognizing step uses said chrominance slicing histogram to detect said areas corresponding to the skin of said subject formed by the pixels of said taken image belonging in said chrominance slicing histogram to said threshold area.
8. A digital image processing method according to claim 5, wherein said taken image is of a Bayer type, the method further comprising:
sub-sampling the image according to G*=(G1+G2)/2, wherein G1 and G2 are first and second green channels of the image, and said step of constructing said chrominance slicing histogram uses normalized channels r and g of the type:

r=R/(R+G*+B)
g=G*/(R+G*+B),
R, G and B being red, green and blue values of each pixel of said taken RGB image.
9. A digital image processing method according to claim 1, wherein said taken image is of a Bayer type.
10. A digital image processing method according to claim 9, wherein said exposure correction of said taken image uses:
a simulated response function of a type:
f ( q ) = 255 ( 1 + - ( Aq ) ) C
A and C being predetermined control parameters and q being a light quantity value expressed in base 2 logarithmic units; and
a grey average level (avg) calculated on said visually interesting regions, in order to calculate a distance Δ of an ideal exposure situation using:

Δ=ƒ−1(128)−ƒ−1(avg)
and in order to change a luminance value Y(x, y) of a pixel with position (x, y) in:

Y′(x,y)=ƒ(ƒ−1(Y(x,y))+Δ).
11. A digital image processing method according to claim 1, further comprising a final color reconstruction step.
12. A digital image processing method according to claim 11, said image being taken in an RGB format, wherein said final color reconstruction step comprises the relations:
R = 0.5 · ( Y Y · ( R + Y ) + R - Y ) G = 0.5 · ( Y Y · ( G + Y ) + G - Y ) B = 0.5 · ( Y Y · ( B + Y ) + B - Y )
R, G, B, and Y being respective red, green, blue, and luminance values of said taken image, Y′ being a desired luminance value, and R′, G′, and B′ being respective red, green, and blue values of the image after said final color reconstruction step.
13. A digital image processing method according to claim 11, said image being taken in a Bayer Pattern format, wherein said final color reconstruction step provides that a grey value I(x, y) of a pixel with position (x, y) is changed in:

I′(x,y)=ƒ(ƒ−1(I(x,y))+Δ),
where Δ is a distance of an ideal exposure situation.
14. A digital image processing method, comprising:
extracting chromatic information from an image of a human subject;
detecting, based on the extracted chromatic information, which areas of the image correspond to skin of the subject; and
normalizing grey scale values of the image based on the areas of the image that are detected as corresponding to the skin of the subject.
15. The method of claim 14, wherein the detecting step comprises constructing a probabilistic slicing of the image taken in a YcrCb format to evaluate if pixels of the image belong to the areas corresponding to the skin of the subject.
16. The method of claim 14, wherein the detecting step comprises applying a threshold area of the image taken in an RGB format to evaluate if pixels of the image belong to the areas corresponding to the skin of the subject.
17. The method of claim 16, wherein applying the threshold area comprises constructing a chrominance slicing histogram of the image and using the chrominance slicing histogram to detect the areas corresponding to the skin of the subject formed by the pixels of the image belonging in the chrominance slicing histogram to the threshold area.
18. The method of claim 16, wherein applying the threshold area comprises constructing a chrominance slicing histogram of the image using normalized channels r and g of the type:

r=R/(R+G+B)
g=G/(R+G+B)
R, G and B being red, green and blue values of each pixel of the image.
19. The method of claim 16, wherein applying the threshold area comprises constructing a chrominance slicing histogram of the image and the image is of a Bayer type, the method further comprising:
sub-sampling the image according to G*=(G1+G2)/2, wherein G1 and G2 are first and second green channels of the image, and said step of constructing said chrominance slicing histogram uses normalized channels r and g of the type:

r=R/(R+G*+B)
g=G*/(R+G*+B),
R, G and B being red, green and blue values of each pixel of said taken RGB image.
20. The method of claim 14, wherein the normalizing step performs exposure correction of the image that includes:
using a simulated response function of a type:
f ( q ) = 255 ( 1 + - ( Aq ) ) C
A and C being predetermined control parameters and q being a light quantity value expressed in base 2 logarithmic units; and
calculating a grey average level (avg) of the areas corresponding to the skin;
calculating a distance Δ of an ideal exposure situation using:

Δ=ƒ−1(128)−ƒ−1(avg); and
changing a luminance value Y(x, y) of a pixel with position (x, y) in:

Y′(x, y)=ƒ(ƒ−1(Y(x, y))+Δ).
21. The method of claim 14, further comprising a final color reconstruction step using the relations:
R = 0.5 · ( Y Y · ( R + Y ) + R - Y ) G = 0.5 · ( Y Y · ( G + Y ) + G - Y ) B = 0.5 · ( Y Y · ( B + Y ) + B - Y )
R, G, B, and Y being respective red, green, blue, and luminance values of the image, Y′ being a desired luminance value, and R′, G′, and B′ being respective red, green, and blue values of the image after the final color reconstruction step.
22. The method of claim 14, further comprising a final color reconstruction step that changes a grey value I(x, y) of a pixel with position (x, y) using:

I′(x,y)=ƒ(ƒ−1(I(x,y))+Δ),
where Δ is a distance of an ideal exposure situation.
23. A digital image processor, comprising:
means for extracting chromatic information from an image of a human subject;
means for detecting, based on the extracted chromatic information, which areas of the image correspond to skin of the subject; and
means for normalizing grey scale values of the image based on the areas of the image that are detected as corresponding to the skin of the subject.
Description
    BACKGROUND OF THE INVENTION
  • [0001]
    1. Field of the Invention
  • [0002]
    The present invention relates to a digital image processing method. The invention relates particularly, but not exclusively, to an image processing method of human subjects being photographed by portable image taking devices, particularly of backlit subjects and the following description is made with reference to this field of application for convenience of illustration only.
  • [0003]
    2. Description of the Related Art
  • [0004]
    As is well known, one of the main problems limiting photographic image quality involves the generation of sub-optimal photographs due to the wrong exposure to light of the photographed subject.
  • [0005]
    This problem is particularly suffered in portable devices such as mobile phones, wherein several factors concur in obtaining photographs that are wrongly exposed: the smallness of the available optical device, the lack of a flash device and the like. Moreover, the portable device nature and the traditional use of the photographs produced therefrom, particularly linked to the so-called multimedia messaging services or MMS, cause the acquisition of photographs of the type shown in FIG. 3.
  • [0006]
    Although it is impossible to provide a precise definition of a correct exposure, since the exposure depends on the photographed subject as well as on the personal taste of the person looking at the photograph, it is however possible to state that, for “normal” subjects (and thus not considering extreme cases, like a snow-covered landscape whose correct acquisition would involve an intentional photograph overexposure), a correct exposure is obtained when the main features of the photographic image are reproduced by using an intermediate grey level.
  • [0007]
    In the image processing field several techniques for improving the tone quality of photographic images are well known, such as histogram equalization, grey-level slicing, and histogram stretching.
  • [0008]
    Although advantageous under many aspects, these prior art techniques have several drawbacks mainly linked to the fact of being independent from the visual content of the photographed images.
  • [0009]
    The article entitled “Automated Global Enhancement of Digitized Photographs” by Bhukhanwale et al., published on the IEEE Transaction on Consumer Electronics, vol. 40, no. 1, 1994, which is hereby incorporated by reference in its entirety, describes instead an algorithm being capable to identify visually important regions in a photographic image, by adjusting the image exposure so that these regions occupy intermediate tone levels.
  • [0010]
    Moreover, the European patent application no. EP 01830803.1 filed in the name of STMicroelectronics, the assignee of the present application, which is hereby incorporated by reference in its entirety, describes an algorithm being similarly capable to identify visually important regions in a photographic image in order to replace them at intermediate tone levels. This algorithm directly processes images of the Bayer Pattern type and simplifies the statistical measures used to detect regions in the image having a high information content, i.e., visually important regions.
  • [0011]
    The algorithms provided in this document directly operate on the image in the Bayer Pattern format and they comprise the following steps:
      • extraction of the Bayer Pattern green plane or channel G: this plane provides a good approximation of the luminance Y.
        • visual analysis: once the channel G has been extracted, the visually interesting regions are identified on this channel. For this purpose, the green plane is split into N blocks having the same size and the following statistical values are calculated for each block:
        • focus: it characterizes the block sharpness and it is used for identifying the regions comprising high-frequency components, corresponding to details of the photographed image;
        • contrast: it is related to the image tone range—the higher the contrast, the higher the insulation of the so-called clusters of points in the block, i.e., the higher the block visual impact.
  • [0016]
    In order to obtain important visual features, independently from the lighting conditions of the photographed image, the visual analysis is performed on an image having an intermediate luminosity produced by making a temporary correction only based on the average value of the channel G calculated on the whole plane. The algorithms further perform exposure adjustment: once the visually interesting regions have been detected, the exposure adjustment is performed by using the average grey levels of these regions as reference values. In greater detail, the photographed image is changed so to bring the average value of these regions to a target value T by changing all the pixels belonging to the Bayer Pattern. This target value T should be a value ranging around 128 and it should take into consideration a possible correction range performed after the color reconstruction of the corrected Bayer Pattern. This means that, in certain cases, the target value T could be substantially lower than 128.
  • [0017]
    To this aim, a simulated response curve of a digital image taking device or camera is used, schematically shown in FIG. 1.
  • [0018]
    This curve gives an evaluation of how the light values picked up by the camera are turned into pixel values, i.e., it represents the function:
    ƒ(q)=I  (1)
    q being the light amount and I the final pixel value.
  • [0020]
    This simulated response function (1) of a camera can be expressed in a parametric way: f ( q ) = 255 ( 1 + - ( Aq ) ) C ( 2 )
  • [0021]
    A and C being the control parameters of the curve shape and the value q being expressed in base 2 logarithmic units (also known with the name “stops”. It is possible to evaluate these control parameters A and C by using the information comprised in the article by Mann et al. entitled “Comparametric Equations with Practical Applications in Quantigraphic Image Processing”, IEEE Transactions on Image Processing, Vol. 9, no. 8, 2000, which is hereby incorporated by reference in its entirety.
  • [0022]
    It is also possible to obtain experimentally the values of these parameters A and C or to set them in order to realize a particular final effect (for example, a more or less marked improvement of the contrast). In particular, FIG. 1 shows the trend of the simulated response curve expressed by the formula (2) with A=7 and C=0.13.
  • [0023]
    By using this simulated response curve f and an average grey level avg for the visually important regions, the distance Δ of an ideal exposure situation is expressed as:
    Δ=ƒ−1(128)−ƒ−1(avg)  (3)
    and the grey value I(x, y) of a pixel with position (x, y) is thus changed in:
    I′(x,y)=ƒ(ƒ−1(I(x,y))+Δ)  (4)
    It is worth noting that all the grey values of the pixels are corrected.
  • [0026]
    In particular, the above-mentioned changes are substantially a look-up table (LUT) transformation (i.e., they can be put in a table. in order to be then referred to) and FIGS. 2A and 2B show two different transformations (the curves LUT1 and LUT2) generated from a first simulated response curve f1 with values A=7 and C=0.13 and a second simulated response curve f2 with values A=0.85 and C=1.
  • [0027]
    It is worth noting that the distance or offset of the value 128 is 1.24 for f1 and 0.62 for f2 respectively (starting from a same input value equal to 72).
  • [0028]
    From the FIGS. 2A and 2B it is evident that the first curve LUT1 has a more linear trend, while the second curve LUT2 has a so-called range trend.
  • [0029]
    Although advantageous under several aspects, these prior art techniques are not very effective in the case of portable devices like mobile phones for which the photographic images are often backlit and they are mainly focused on human figures, when the user uses the image transmission for videophony, as shown in FIG. 3.
  • BRIEF SUMMARY OF THE INVENTION
  • [0030]
    One embodiment of the present invention provides an image processing method having such features as to overcome the limits still affecting prior art techniques.
  • [0031]
    One embodiment of the present invention detects the features in the photograph of the skin of the subject being photographed in order to select and find convenient interesting regions on whose base an exposure adjustment/correction is applied.
  • [0032]
    One embodiment of the present invention is directed to a digital image processing method that includes: extracting chromatic information of an image taken by an image taking device and related to a human subject; detecting visually interesting regions in the taken image by recognizing areas corresponding to skin of the subject, wherein the recognized areas are the visually interesting regions; and correcting exposure of the taken image by normalizing a grey scale of the taken image based on the visually interesting regions.
  • [0033]
    The features and advantages of the method according to the invention will be apparent from the following description of an embodiment thereof given by way of non-limiting example with reference to the attached drawings.
  • BRIEF DESCRIPTION OF THE DRAWINGS
  • [0034]
    In the drawings:
  • [0035]
    FIG. 1 shows the trend of a simulated response curve of a known image taking device;
  • [0036]
    FIGS. 2A and 2B show LUT transformations related to different curves like the one in FIG. 1;
  • [0037]
    FIG. 3 shows an image of a backlit subject taken by a known image taking device;
  • [0038]
    FIG. 4 shows an illustrative diagram of a step of the image processing method according to one embodiment of the invention;
  • [0039]
    FIGS. 5A-5C and 6A-6C show following image processings for detecting important areas which are used in a step of the image processing method according to one embodiment of the invention;
  • [0040]
    FIGS. 7A-7D schematically shows the method according to one embodiment of the invention by means of following image processings;
  • [0041]
    FIG. 8A shows an image of a subject; FIG. 8B shows an image of the subject of FIG. 8A with areas highlighted corresponding to the skin of the subject; and FIG. 8C shows a detection histogram of the image of FIG. 8A;
  • [0042]
    FIGS. 9A-12B show processed images obtained by the method according to alternate embodiments of the invention.
  • DETAILED DESCRIPTION OF THE INVENTION
  • [0043]
    An image processing method according to one embodiment of the invention performs an exposure correction of a digital photographed image taken by an image taking device on the basis of a recognition algorithm of the skin of the photographed subject, thus improving the final photographic image quality, in a decisive way in the case of backlit subjects as in FIG. 3.
  • [0044]
    In particular, the method comprises the following steps:
      • 1) a first extraction step of chromatic information from the photographic image;
      • 2) a second visual analysis step using a recognition method of the areas corresponding to the skin of the subject photographed in the photographic image; and
      • 3) a third exposure adjustment step of the obtained photographic image.
  • [0048]
    1) Extraction Step
  • [0049]
    As has been seen with reference to known image processing methods, the method provides the extraction of the green channel G of the image taken when the images are in the Bayer format.
  • [0050]
    On the contrary, it provides the extraction of the luminance channel Y for images of the YcbCr type obtained from RGB images.
  • [0051]
    2) Visual Analysis Step
  • [0052]
    This analysis can be performed on:
      • 1. images in the RGB format;
      • 2. images in the Bayer Pattern format generating, from an initial image, a conveniently sub-sampled RGB copy.
  • [0055]
    In particular, by using this skin recognition method, a plurality of visually interesting regions corresponding to the skin of the photographed subject are detected.
  • [0056]
    The chromatic information obtained during the first extraction step is thus used.
  • [0057]
    In particular, using Bayer data, it is possible to operate on three color planes and on sub-samples having a size corresponding to a quarter of the initial data, as schematically shown in FIG. 4, thus considerably reducing the calculation efforts of the method.
  • [0058]
    3) Third Exposure Adjustment Step
  • [0059]
    This adjustment can be performed in two ways:
      • 1. correction of images in the RGB format;
      • 2. correction of images in the Bayer Pattern format before a following color interpolation algorithm.
  • [0062]
    In the case of the correction of images in the RGB format, once the visually important pixels have been detected as above-mentioned (i.e., the pixels belonging to the area corresponding to the skin of the photographed subject), a known exposure correction algorithm is used, wherein the average grey level of the known pixel clusters is considered as belonging to the skin of the photographed subject.
  • [0063]
    In other words, the pixels belonging to the subject skin are placed at the intermediate level of the image grey scale and all the remaining image pixels are placed once again based on this average level.
  • [0064]
    In particular, once the luminance value has been corrected from an original value Y to a revised value Y′ that reflects the average grey level of the know pixel clusters corresponding to the skin of the subject, according to the above-mentioned steps (2), (3) and thus using the information comprised in the article by Sakaue et al. entitled “Adaptive Gamma Processing of the Video Cameras for the Expansion of the Dynamic Range”, IEEE Transaction on Consumer Electronics, Vol. 41, n. 3, August 1995, which is hereby incorporated by reference in its entirety, starting from a curve of the type shown in FIG. 1, the pixel chromatic values can be reconstructed according to the formulas: R = 0.5 · ( Y Y · ( R + Y ) + R - Y ) ( 5 ) G = 0.5 · ( Y Y · ( G + Y ) + G - Y ) ( 6 ) B = 0.5 · ( Y Y · ( B + Y ) + B - Y ) ( 7 )
    R, G, B being the color values of the input pixels.
  • [0066]
    In the case of the correction of images in the Bayer format the formulas (5), (6) and (7) cannot be used and the output product will be obtained by simply applying the relation (4) to all the pixels of the pattern.
  • [0067]
    The recognition method of the areas corresponding to the skin of the subject photographed in the photographic image will be now described in greater detail.
  • [0068]
    Several recognition methods of the color of the skin of the photographed subject are known, substantially based on the application of a threshold to a color probability measure for the skin.
  • [0069]
    In fact, the colors of the human skin belong to a particular color category, different from the colors of most natural objects. In particular, in the article by Zarti et al. entitled “Comparison of Five Color Models in Skin Pixel Classification”, Proc. Of Int. Workshop on Recognition, Analysis and Tracking of Faces and Gestures in Real-Time Systems, IEEE Computer Society, Corfu, Greece, pages 58-63, 1999, which is hereby incorporated by reference in its entirety, it has been shown that the colors of the human skin are clustered and that the skin changes between each subject are substantially due to a different intensity and they can thus be reduced by using only the chrominance component of the subject image.
  • [0070]
    Moreover, in the article by Yang et al. entitled “Skin-Color Modeling and Adaptation”, Technical Report CMU-CS-97-146, School of Computer Science, Carnegie Mellon University, 1997, which is hereby incorporated by reference in its entirety, it has been shown that the human skin color slicing can be represented by a bidimensional Gaussian function in the chrominance plane. The center of this slicing is determined by the mean vector {right arrow over (μ)} and the amplitude of the bell thereof is determined by the covariance matrix Σ, these two values being evaluated starting from a convenient group of test data.
  • [0071]
    The conditional probability p({right arrow over (x)}|s) of a pixel block to belong to a human skin color class starting from the chrominance vector thereof {right arrow over (x)} is thus given by: p ( x | s ) = 1 2 π Σ - 1 2 exp { - [ d ( x ) ] 2 2 } ( 8 )
    d({right arrow over (x)}) being the so-called Mahalonobis distance of the chrominance vector {right arrow over (x)} of the mean vector {right arrow over (μ)}, defined as:
    [d({right arrow over (x)})] 2=({right arrow over (x)}−{right arrow over (μ)})′Σ−1({right arrow over (x)}−{right arrow over (μ)})  (9)
    In other words, the value of the Mahalonobis distance d({right arrow over (x)}) of a pixel block with chrominance vector {right arrow over (x)} determines the probability of this block to belong to a predetermined human skin color class. The higher the Mahalonobis distance d({right arrow over (x)}) is, the lower the probability of the block belonging to this human skin color class.
  • [0074]
    Given the considerable amount of color types, of distance measures and of bidimensional slicings, a considerable variety of image processing algorithms can be considered. Moreover, the lighting conditions and the color models can change according to the image taking modes.
  • [0075]
    Advantageously, the method comprises a step of recognizing a portion of the photographic image corresponding to the subject skin.
  • [0076]
    In a first embodiment, this recognition step is substantially based on a probabilistic function.
  • [0077]
    In particular, for each pixel of an image taken in the YcrCb format a probabilistic slicing is prepared to evaluate if this pixel must be classified as belonging to the subject skin. Based on this slicing, a new image is thus processed with a normalized grey scale, wherein the subject skin is highlighted as indicated in FIGS. 5A-6C on two different photographic images which depict human subjects.
  • [0078]
    Based on this first embodiment of the recognition step of the photographed subject skin, the image pixels with higher grey values are considered as belonging to the skin of the photographed subject.
  • [0079]
    The areas being detected by using this first embodiment of the recognition step of the skin of the photographed subject on an image shown in FIG. 5A are shown in Figure and 5C. Similarly, FIG. 6C highlights areas of the image of FIG. 6A corresponding to the skin of another photographed subject, according to the first embodiment of the recognition step.
  • [0080]
    In a second embodiment, the recognition step of the image areas corresponding to the skin of the photographed subject is substantially based on a single threshold area and it provides an RGB-format image processing in order to produce a chrominance slicing histogram starting from normalized channels r and g as described in the article by Soriano et al. entitled “Skin Color Modeling Under Varying Illumination Conditions Using the Skin Locus for Selecting Training Pixels”, Real-time Image Sequence Analysis (RISA2000, August 31-September 1, Finland), which is hereby incorporated by reference in its entirety. In particular, the normalized channels r and g are defined as:
    r=R/(R+G+B)  (10)
    g=G/(R+G+B)  (11)
    The resulting bidimensional histogram shows the chrominance slicing in the image and the areas having the right human skin chrominance slicing are thus detected by applying a single threshold area. In particular, the pixels of the processed image belonging to the threshold area are classified as belonging to the subject skin.
  • [0082]
    Similarly, FIG. 5B shows the areas, detected by using this second embodiment of the recognition step, corresponding to the skin of the photographed subject from an image shown in FIG. 5A of a human subject. Also, FIG. 6B highlights areas of the image of FIG. 6A corresponding to the skin of another photographed subject, according to the second embodiment of the recognition step.
  • [0083]
    FIG. 7A-D schematically show the following processings of an image concerning a human subject, particularly backlit, after the different steps of the method according to alternate embodiments of the invention.
  • [0084]
    In particular, on a taken image (FIG. 7A) the recognition step of the areas corresponding to the photographed subject skin is performed with a probabilistic (FIG. 7B) or threshold (FIG. 7C) method.
  • [0085]
    An exposure correction step can thus be performed by using the areas detected as belonging to the skin of the photographed subject in order to normalize the grey levels obtaining a final processed image (FIG. 7D), the best image quality being immediately evident by comparing it with the starting image (FIG. 7A).
  • [0086]
    Advantageously, the regions being detected in the recognition step as belonging to the subject's skin are used as visually important images for the following exposure adjustment step of the photographic image.
  • [0087]
    In a preferred embodiment of the image processing method according to the invention, the recognition step of the areas belonging to the photographed subject's skin processes a 8-bit image of the Bayer type, constructing a color image of sub-samples with size corresponding to a quarter of the initial data, obtained as previously described and schematically shown in FIG. 4.
  • [0088]
    Starting from this color image of sub-samples, a recognition step of the areas belonging to the photographed subject's skin is performed using a chrominance slicing histogram according to the first probabilistic embodiment or the normalized channels r and g according to the second threshold embodiment.
  • [0089]
    However, in this second case, the normalized channels r and g are defined as:
    r=R/(R+G*+B)  (12)
    g=G*/(R+G*+B)  (13)
    being
    G*=(G1+G2)/2  (14)
    The resulting bidimensional histogram shows the chrominance slicing of the processed image, therefore the areas corresponding to the photographed subject skin, as schematically shown in FIGS. 8A-C, showing in series a taken image of the Bayer type (FIG. 8A), the image (FIG. 8B) processed to detect the areas corresponding to the photographed subject skin and a detection histogram (FIG. 8C) r-g of these areas.
  • [0091]
    The method finally comprises a reconstruction step of the color of the image taken according to the relations (5) to (7), already shown with reference to the prior art, R, G, B and R′, G′, B′ being the red, green and blue values of the images being respectively taken and processed.
  • [0092]
    The step sequence being described is suitable for a simple change allowing a correction to be performed directly on images in the Bayer Pattern format in favor of a further simplification from the calculation point of view. In fact, once the image for the skin detection according to the diagram of FIG. 4 has been constructed, the average value calculated for the regions concerned can be used to directly perform the Bayer Pattern correction, using for example the modes described in the above-mentioned European patent application no. 01830803.1.
  • [0093]
    It is however worth noting that the color reconstruction formulas described in the equations (5), (6), (7) cannot be used in this case and the output product of the corrected Bayer Pattern will be obtained by simply applying the relation (4) to all the model pixels.
  • [0094]
    In other words, the grey value I(x, y) of a pixel with position (x, y) is modified in:
    I′(x,y)=ƒ(ƒ−1(I(x,y))+Δ),  (15)
    where Δ is the distance of the ideal exposure situation as expressed in relation (3). The image processing of a backlit subject being performed by using a CMOS-VGA sensor and an evaluation kit on the Windows® platform is shown in FIGS. 9A-9B, wherein in the panel V the areas detected as belonging to the photographed subject's skin have been indicated on a black background.
  • [0096]
    Similarly, FIGS. 10A-12B show the results of a simulation of the method performed by the inventors starting from images taken by a common VGA sensor in the compressed jpeg format (FIGS. 10A, 11A) and by a 4.1 Mpixel CCD sensor of a traditional average-band DSC (Digital Still Camera) (12A) and the images processed with the method 10B, 11B, 12B respectively have been indicated, wherein the image qualitative improvement is completely evident.
  • [0097]
    All of the above U.S. patents, U.S. patent application publications, U.S. patent applications, foreign patents, foreign patent applications and non-patent publications referred to in this specification and/or listed in the Application Data Sheetare incorporated herein by reference, in their entirety.
  • [0098]
    From the foregoing it will be appreciated that, although specific embodiments of the invention have been described herein for purposes of illustration, various modifications may be made without deviating from the spirit and scope of the invention. Accordingly, the invention is not limited except as by the appended claims.
Patent Citations
Cited PatentFiling datePublication dateApplicantTitle
US5585860 *Apr 12, 1995Dec 17, 1996Matsushita Electric Industrial Co., Ltd.Reproduction circuit for skin color in video signals
Referenced by
Citing PatentFiling datePublication dateApplicantTitle
US7403643May 24, 2007Jul 22, 2008Fotonation Vision LimitedReal-time face tracking in a digital image acquisition device
US7460694Jun 19, 2007Dec 2, 2008Fotonation Vision LimitedReal-time face tracking in a digital image acquisition device
US7460695Jun 21, 2007Dec 2, 2008Fotonation Vision LimitedReal-time face tracking in a digital image acquisition device
US7469055Jun 19, 2007Dec 23, 2008Fotonation Vision LimitedReal-time face tracking in a digital image acquisition device
US7684630Dec 9, 2008Mar 23, 2010Fotonation Vision LimitedDigital image adjustable compression and resolution using face detection information
US7693311Jul 5, 2007Apr 6, 2010Fotonation Vision LimitedPerfecting the effect of flash within an image acquisition devices using face detection
US7702136Jul 5, 2007Apr 20, 2010Fotonation Vision LimitedPerfecting the effect of flash within an image acquisition devices using face detection
US7809162Oct 30, 2008Oct 5, 2010Fotonation Vision LimitedDigital image processing using face detection information
US7844076Oct 30, 2006Nov 30, 2010Fotonation Vision LimitedDigital image processing using face detection and skin tone information
US7844135Jun 10, 2009Nov 30, 2010Tessera Technologies Ireland LimitedDetecting orientation of digital images using face detection information
US7848549Oct 30, 2008Dec 7, 2010Fotonation Vision LimitedDigital image processing using face detection information
US7853043Dec 14, 2009Dec 14, 2010Tessera Technologies Ireland LimitedDigital image processing using face detection information
US7860274Oct 30, 2008Dec 28, 2010Fotonation Vision LimitedDigital image processing using face detection information
US7864990Dec 11, 2008Jan 4, 2011Tessera Technologies Ireland LimitedReal-time face tracking in a digital image acquisition device
US7912245Jun 20, 2007Mar 22, 2011Tessera Technologies Ireland LimitedMethod of improving orientation and color balance of digital images using face detection information
US7916897Jun 5, 2009Mar 29, 2011Tessera Technologies Ireland LimitedFace tracking for controlling imaging parameters
US7953251Nov 16, 2010May 31, 2011Tessera Technologies Ireland LimitedMethod and apparatus for detection and correction of flash-induced eye defects within digital images using preview or other reference images
US7962629Sep 6, 2010Jun 14, 2011Tessera Technologies Ireland LimitedMethod for establishing a paired connection between media devices
US7965875Jun 12, 2007Jun 21, 2011Tessera Technologies Ireland LimitedAdvances in extending the AAM techniques from grayscale to color images
US8005265Sep 8, 2008Aug 23, 2011Tessera Technologies Ireland LimitedDigital image processing using face detection information
US8050465Jul 3, 2008Nov 1, 2011DigitalOptics Corporation Europe LimitedReal-time face tracking in a digital image acquisition device
US8055029Jun 18, 2007Nov 8, 2011DigitalOptics Corporation Europe LimitedReal-time face tracking in a digital image acquisition device
US8055067Jan 18, 2007Nov 8, 2011DigitalOptics Corporation Europe LimitedColor segmentation
US8055090Sep 14, 2010Nov 8, 2011DigitalOptics Corporation Europe LimitedDigital image processing using face detection information
US8126208Dec 3, 2010Feb 28, 2012DigitalOptics Corporation Europe LimitedDigital image processing using face detection information
US8131016Dec 3, 2010Mar 6, 2012DigitalOptics Corporation Europe LimitedDigital image processing using face detection information
US8135184May 23, 2011Mar 13, 2012DigitalOptics Corporation Europe LimitedMethod and apparatus for detection and correction of multiple image defects within digital images using preview or other reference images
US8155397Sep 26, 2007Apr 10, 2012DigitalOptics Corporation Europe LimitedFace tracking in a camera processor
US8213737Jun 20, 2008Jul 3, 2012DigitalOptics Corporation Europe LimitedDigital image enhancement with reference images
US8224108Dec 4, 2010Jul 17, 2012DigitalOptics Corporation Europe LimitedDigital image processing using face detection information
US8270674Jan 3, 2011Sep 18, 2012DigitalOptics Corporation Europe LimitedReal-time face tracking in a digital image acquisition device
US8320641Jun 19, 2008Nov 27, 2012DigitalOptics Corporation Europe LimitedMethod and apparatus for red-eye detection using preview or other reference images
US8326066Mar 8, 2010Dec 4, 2012DigitalOptics Corporation Europe LimitedDigital image adjustable compression and resolution using face detection information
US8330831Jun 16, 2008Dec 11, 2012DigitalOptics Corporation Europe LimitedMethod of gathering visual meta data using a reference image
US8345114Jul 30, 2009Jan 1, 2013DigitalOptics Corporation Europe LimitedAutomatic face and skin beautification using face detection
US8374425 *Dec 17, 2007Feb 12, 2013Stmicroelectronics, S.R.L.Method of chromatic classification of pixels and method of adaptive enhancement of a color image
US8379917Oct 2, 2009Feb 19, 2013DigitalOptics Corporation Europe LimitedFace recognition performance using additional image features
US8384793Jul 30, 2009Feb 26, 2013DigitalOptics Corporation Europe LimitedAutomatic face and skin beautification using face detection
US8385610Jun 11, 2010Feb 26, 2013DigitalOptics Corporation Europe LimitedFace tracking for controlling imaging parameters
US8422739Sep 15, 2011Apr 16, 2013DigitalOptics Corporation Europe LimitedReal-time face tracking in a digital image acquisition device
US8498452Aug 26, 2008Jul 30, 2013DigitalOptics Corporation Europe LimitedDigital image processing using face detection information
US8509496Nov 16, 2009Aug 13, 2013DigitalOptics Corporation Europe LimitedReal-time face tracking with reference images
US8509498Sep 26, 2011Aug 13, 2013DigitalOptics Corporation Europe LimitedReal-time face tracking in a digital image acquisition device
US8593542Jun 17, 2008Nov 26, 2013DigitalOptics Corporation Europe LimitedForeground/background separation using reference images
US8666124Mar 12, 2013Mar 4, 2014DigitalOptics Corporation Europe LimitedReal-time face tracking in a digital image acquisition device
US8666125Mar 12, 2013Mar 4, 2014DigitalOptics Corporation European LimitedReal-time face tracking in a digital image acquisition device
US8675991Jun 2, 2006Mar 18, 2014DigitalOptics Corporation Europe LimitedModification of post-viewing parameters for digital images using region or feature information
US8682097Jun 16, 2008Mar 25, 2014DigitalOptics Corporation Europe LimitedDigital image enhancement with reference images
US8744145Mar 12, 2013Jun 3, 2014DigitalOptics Corporation Europe LimitedReal-time face tracking in a digital image acquisition device
US8811733 *Jan 14, 2013Aug 19, 2014Stmicroelectronics S.R.L.Method of chromatic classification of pixels and method of adaptive enhancement of a color image
US8824827 *Jan 5, 2012Sep 2, 2014Qualcomm IncorporatedSystems, methods, and apparatus for exposure control
US8896725Jun 17, 2008Nov 25, 2014Fotonation LimitedImage capture device with contemporaneous reference image capture mechanism
US8948468Jun 26, 2003Feb 3, 2015Fotonation LimitedModification of viewing parameters for digital images using face detection information
US8989453Aug 26, 2008Mar 24, 2015Fotonation LimitedDigital image processing using face detection information
US9007480Jul 30, 2009Apr 14, 2015Fotonation LimitedAutomatic face and skin beautification using face detection
US9053545Mar 19, 2007Jun 9, 2015Fotonation LimitedModification of viewing parameters for digital images using face detection information
US9129381Jun 17, 2008Sep 8, 2015Fotonation LimitedModification of post-viewing parameters for digital images using image region or feature information
US20050270948 *Jun 1, 2005Dec 8, 2005Funai Electric Co., Ltd.DVD recorder and recording and reproducing device
US20060204034 *Jun 26, 2003Sep 14, 2006Eran SteinbergModification of viewing parameters for digital images using face detection information
US20070110305 *Oct 30, 2006May 17, 2007Fotonation Vision LimitedDigital Image Processing Using Face Detection and Skin Tone Information
US20070160307 *Mar 19, 2007Jul 12, 2007Fotonation Vision LimitedModification of Viewing Parameters for Digital Images Using Face Detection Information
US20080037838 *May 24, 2007Feb 14, 2008Fotonation Vision LimitedReal-Time Face Tracking in a Digital Image Acquisition Device
US20080037839 *Jun 19, 2007Feb 14, 2008Fotonation Vision LimitedReal-Time Face Tracking in a Digital Image Acquisition Device
US20080037840 *Jun 21, 2007Feb 14, 2008Fotonation Vision LimitedReal-Time Face Tracking in a Digital Image Acquisition Device
US20080043122 *Jul 5, 2007Feb 21, 2008Fotonation Vision LimitedPerfecting the Effect of Flash within an Image Acquisition Devices Using Face Detection
US20080143854 *Nov 18, 2007Jun 19, 2008Fotonation Vision LimitedPerfecting the optics within a digital image acquisition device using face detection
US20080144946 *Dec 17, 2007Jun 19, 2008Stmicroelectronics S.R.L.Method of chromatic classification of pixels and method of adaptive enhancement of a color image
US20080175481 *Jan 18, 2007Jul 24, 2008Stefan PetrescuColor Segmentation
US20080316328 *Jun 17, 2008Dec 25, 2008Fotonation Ireland LimitedForeground/background separation using reference images
US20080317357 *Jun 16, 2008Dec 25, 2008Fotonation Ireland LimitedMethod of gathering visual meta data using a reference image
US20080317378 *Jun 16, 2008Dec 25, 2008Fotonation Ireland LimitedDigital image enhancement with reference images
US20080317379 *Jun 20, 2008Dec 25, 2008Fotonation Ireland LimitedDigital image enhancement with reference images
US20090003652 *Jun 17, 2008Jan 1, 2009Fotonation Ireland LimitedReal-time face tracking with reference images
US20090003708 *Jun 17, 2008Jan 1, 2009Fotonation Ireland LimitedModification of post-viewing parameters for digital images using image region or feature information
US20090052749 *Oct 30, 2008Feb 26, 2009Fotonation Vision LimitedDigital Image Processing Using Face Detection Information
US20090052750 *Oct 30, 2008Feb 26, 2009Fotonation Vision LimitedDigital Image Processing Using Face Detection Information
US20090080713 *Sep 26, 2007Mar 26, 2009Fotonation Vision LimitedFace tracking in a camera processor
US20090102949 *Jul 5, 2007Apr 23, 2009Fotonation Vision LimitedPerfecting the Effect of Flash within an Image Acquisition Devices using Face Detection
US20090141144 *Dec 9, 2008Jun 4, 2009Fotonation Vision LimitedDigital Image Adjustable Compression and Resolution Using Face Detection Information
US20090208056 *Dec 11, 2008Aug 20, 2009Fotonation Vision LimitedReal-time face tracking in a digital image acquisition device
US20100026831 *Jul 30, 2009Feb 4, 2010Fotonation Ireland LimitedAutomatic face and skin beautification using face detection
US20100026832 *Jul 30, 2009Feb 4, 2010Mihai CiucAutomatic face and skin beautification using face detection
US20100039525 *Oct 20, 2009Feb 18, 2010Fotonation Ireland LimitedPerfecting of Digital Image Capture Parameters Within Acquisition Devices Using Face Detection
US20100054533 *Aug 26, 2008Mar 4, 2010Fotonation Vision LimitedDigital Image Processing Using Face Detection Information
US20100054549 *Aug 26, 2008Mar 4, 2010Fotonation Vision LimitedDigital Image Processing Using Face Detection Information
US20100060727 *Nov 16, 2009Mar 11, 2010Eran SteinbergReal-time face tracking with reference images
US20100092039 *Dec 14, 2009Apr 15, 2010Eran SteinbergDigital Image Processing Using Face Detection Information
US20100165140 *Mar 8, 2010Jul 1, 2010Fotonation Vision LimitedDigital image adjustable compression and resolution using face detection information
US20100271499 *Oct 20, 2009Oct 28, 2010Fotonation Ireland LimitedPerfecting of Digital Image Capture Parameters Within Acquisition Devices Using Face Detection
US20100321520 *Aug 24, 2010Dec 23, 2010Texas Instruments IncorporatedDigital camera and method
US20110026780 *Jun 11, 2010Feb 3, 2011Tessera Technologies Ireland LimitedFace tracking for controlling imaging parameters
US20110060836 *Sep 6, 2010Mar 10, 2011Tessera Technologies Ireland LimitedMethod for Establishing a Paired Connection Between Media Devices
US20110075894 *Dec 3, 2010Mar 31, 2011Tessera Technologies Ireland LimitedDigital Image Processing Using Face Detection Information
US20110081052 *Oct 2, 2009Apr 7, 2011Fotonation Ireland LimitedFace recognition performance using additional image features
US20110129121 *Jan 3, 2011Jun 2, 2011Tessera Technologies Ireland LimitedReal-time face tracking in a digital image acquisition device
US20110221936 *May 23, 2011Sep 15, 2011Tessera Technologies Ireland LimitedMethod and Apparatus for Detection and Correction of Multiple Image Defects Within Digital Images Using Preview or Other Reference Images
US20120105675 *Jan 5, 2012May 3, 2012Qualcomm IncorporatedSystems, methods, and apparatus for exposure control
WO2008017343A1 *Jun 18, 2007Feb 14, 2008Fotonation Vision LimitedReal-time face tracking in a digital image acquisition device
Classifications
U.S. Classification382/274, 382/162
International ClassificationH04N1/46, G06T5/40, H04N1/407, G06K9/00, H04N9/64, H04N1/60
Cooperative ClassificationG06T5/40, H04N1/6027, G06K9/00234, G06T5/007
European ClassificationG06T5/00M, G06K9/00F1C, G06T5/40, H04N1/60E
Legal Events
DateCodeEventDescription
Apr 20, 2005ASAssignment
Owner name: STMICROELECTRONICS S.R.L., ITALY
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:STMICROELECTRONICS S.R.L.;REEL/FRAME:015924/0310
Effective date: 20041115
Owner name: STMICROELECTRONICS SA, FRANCE
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:STMICROELECTRONICS S.R.L.;REEL/FRAME:015924/0310
Effective date: 20041115
Aug 2, 2012ASAssignment
Owner name: STMICROELECTRONICS S.R.L., ITALY
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:MESSINA, GIUSEPPE;BATTIATO, SEBASTIANO;CASTORINA, ALFIO;AND OTHERS;REEL/FRAME:028710/0574
Effective date: 20040907