Device for reducing a color cast of a facial image
Patent Information
- Application Number
- DE502017016987
- Authority / Receiving Office
- DE · DE
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2016-12-22
- Filing Date
- 2017-12-12
- Publication Date
- 2025-08-14
- Estimated Expiration
- 2037-12-12
AI Technical Summary
Facial recognition algorithms do not effectively utilize color information due to unreliable color correction of facial images, particularly in identification documents and access control systems, where color casts distort skin colors and reduce recognition performance.
Incorporating a reference color strip in the facial image capture process, which is used to detect and correct color deviations through a processor, allowing for color transformation and correction within a CIE L*a*b* color space.
Enables improved facial recognition performance by reducing color casts, ensuring compliance with identification document standards and enhancing classification accuracy.
Description
[0001] The present invention relates to the field of color correction of facial images of persons, in particular for use in identification documents and access control arrangements.
[0002] Today's facial recognition algorithms typically do not use color information, but rather only location, contrast, and texture information from captured facial images to identify individuals. This is primarily due to the fact that, in practice, there is currently no reliable color correction of facial images, especially passport photos. This also applies to captured facial images for verification purposes, such as in access control systems or ABC gates, where the captured facial image is compared with a stored facial image. In such environments, even uniform facial illumination typically proves to be a challenge.
[0003] In countries where printed and re-scanned facial images are used to apply for identification documents, such as currently in Germany, the unknown lighting of the face during the facial image capture process is compounded by further changes during the printing and subsequent scanning of the printed facial image. Furthermore, in certain cases, the color capture behavior of the camera's sensor may be unknown. Consequently, there are typically at least three locations where the colors of the facial image can be significantly altered, potentially causing a color cast in the facial image.
[0004] For this reason, facial recognition algorithms typically do not evaluate color information, which adversely affects recognition performance. Furthermore, facial images in identification documents often exhibit a color cast, which, for example, distorts the person's actual skin color.
[0005] In JF Dlugos, JL Taylor, "Field Application Report: Materials Characterization: UV / Vis / NIR Spectroscopy; Visible Reflectance Spectroscopy of Human Skin: the use of CIE L*a*b* Color Analysis for In Vivo Ethnic Skin Characterization", PerkinElmer Inc., 2012, typical skin and hair colors of individuals are given in a CIE L*a*b* color space.
[0006] It is therefore an object of the present invention to provide an efficient concept for reducing a color cast of a facial image of a person.
[0007] This problem is solved by the features of the independent claims. Advantageous further developments are the subject of the dependent patent claims, the description, and the drawings.
[0008] The invention is based on and includes the finding that the above object can be achieved using a reference color strip which is visible in addition to a person's face in a captured facial image. The use of a reference color strip is known, for example, from documents US 2006 / 129411 A1 and US 2010 / 214421 A1. The reference color strip comprises at least one reference color field with a predetermined reference color. The reference color strip is typically arranged in an environment of the person when the facial image is captured, such that it is visible in the facial image. The facial image can then undergo a number of analog and / or digital processing steps, for example, be printed out and scanned again, which can cause an undesirable color cast in the facial image.The color cast of the facial image can then be automatically reduced using the reference color stripe in the facial image. The reference color stripe is also automatically removed from the corrected facial image.
[0009] In particular, embodiments enable compliance with future standards for capturing facial images for identification documents, allowing established procedures for applying for identification documents to be maintained. Such standards are set, for example, by the International Civil Aviation Organization (ICAO), which requires compliance with ISO / IEC 19794-5 or ISO / IEC 39794-5 when capturing facial images. Furthermore, the additional use of color information by access control arrangements can achieve improved recognition performance in terms of smaller classification errors.
[0010] According to a first aspect, the invention relates to a device for reducing a color cast of a facial image of a person, wherein the facial image comprises an image of a face of the person and a reference color strip. The reference color strip comprises a reference color field, wherein a predetermined reference color is assigned to the reference color field. The device comprises a memory configured to store the facial image. The device further comprises a processor configured to detect the reference color strip in the facial image, detect the reference color field in the detected reference color strip, extract a color from the detected reference color field, compare the extracted color with the predetermined reference color to obtain a color deviation, and reduce the color cast of the facial image based on the color deviation to obtain a corrected facial image.The color cast can be reduced, for example, using a color transformation, which involves shifting the colors of the facial image within a color space. This allows a corrected facial image to be obtained in which the colors of all pixels belonging to the person's face are corrected—where necessary—based on the color deviation.
[0011] According to one embodiment, the reference color strip comprises a further reference color field, wherein a further predetermined reference color is assigned to the further reference color field, and wherein the processor is configured to detect the further reference color field in the detected reference color strip, extract a further color from the detected further reference color field, compare the extracted further color with the further predetermined reference color to obtain a further color deviation, and further reduce the color cast of the facial image based on the further color deviation. The further predetermined reference color of the further reference color field differs from the reference color of the reference color field. This achieves the advantage that an improved color transformation can be performed to reduce the color cast of the facial image.In particular, the colors of the facial image can be scaled in a color space.
[0012] According to one embodiment, the reference color stripe has a predetermined geometric shape, wherein the processor is configured to detect the reference color stripe in the facial image based on the predetermined geometric shape. This provides the advantage that the reference color stripe can be efficiently detected, for example, using pattern recognition or edge detection. The detected reference color stripe can also be perspectively corrected.
[0013] According to one embodiment, the reference color stripe comprises a texture field, wherein the texture field has a predetermined geometric texture, and wherein the processor is configured to detect the reference color stripe in the facial image based on the predetermined geometric texture. This provides the advantage that the reference color stripe can be efficiently detected, for example, using pattern recognition or edge detection. The detected reference color stripe is further perspectively corrected.
[0014] According to one embodiment, the processor is configured to detect the person's face in the facial image, determine a section of the facial image, wherein the section includes the detected face of the person and wherein the section does not include the detected reference color strip, and limit the corrected facial image to the determined section. This provides the advantage that the corrected facial image can be used directly for applications in which the reference color strip should not be visible. The section can be rectangular, for example. The section can be determined in the original detected facial image or in the corrected facial image.
[0015] According to one embodiment, the processor is configured to determine a position of the captured reference color stripe in the facial image, in particular the originally captured facial image, or the corrected facial image, and to replace the reference color stripe at the determined position in the corrected facial image with a predetermined color. This provides the advantage that the corrected facial image can be used directly for applications in which the reference color stripe should not be visible.
[0016] According to one embodiment, the processor is configured to determine a position of the captured reference color stripe in the facial image, in particular the originally captured facial image, or the corrected facial image, and to replace the reference color stripe at the determined position in the corrected facial image with a background color of the corrected facial image. This provides the advantage that the corrected facial image can be used directly for applications in which the reference color stripe should not be visible.
[0017] According to one embodiment, the processor is configured to determine the background color of the corrected facial image. This can be done, for example, by determining the largest contiguous area of the corrected facial image in which all pixels lie within a predetermined color interval, wherein an average value or a most frequent color value within the color interval can then be selected as the background color. This provides the advantage that the background color can be determined dynamically for each corrected facial image.
[0018] According to one embodiment, the color deviation is formed by a Euclidean distance between the color coordinates of the extracted color and the color coordinates of the predetermined reference color. This provides the advantage that the color deviation can be efficiently determined by the processor.
[0019] According to one embodiment, the further color deviation is formed by a further Euclidean distance between the color coordinates of the extracted further color and the color coordinates of the further predetermined reference color. This provides the advantage that the further color deviation can be efficiently determined by the processor.
[0020] According to one embodiment, the processor is configured to convert the facial image into a CIE L*a*b* color space. This provides the advantage that the facial image can be efficiently further processed. For example, the facial image may originally be in an sRGB color space or AdobeRGB color space.
[0021] According to one embodiment, the processor is configured to convert the corrected facial image into a CIE L*a*b* color space. This provides the advantage that the corrected facial image can be efficiently further processed.
[0022] According to one embodiment, the predetermined reference color can be represented in a CIE L*a*b* color space, wherein the predetermined reference color has an a* color coordinate between 5 and 14 and a b* color coordinate between 10 and 24. This achieves the advantage that a color cast of a skin color can be efficiently reduced.
[0023] According to one embodiment, the further predetermined reference color can be represented in a CIE L*a*b* color space, wherein the further predetermined reference color has an a* color coordinate between 5 and 14 and a b* color coordinate between 10 and 24. This achieves the advantage that a color cast of a skin color can be efficiently reduced.
[0024] Where reference is made in this document to the CIE L*a*b* colour space, this refers in particular to the colour space standardised in EN ISO 11664-4 in the version dated July 2011.
[0025] For both the predetermined reference color and the further predetermined reference color, reference colors with the following color coordinates {a*, b*} can be used, for example: {5, 10}, {5, 15}, {5, 20}, {5, 24}, {10, 10}, {10, 15}, {10, 20}, {10, 24}, {14, 10}, {14, 15}, {14, 20}, {14, 24}. In particular, the standardized X-Rite or Macbeth ColorChecker Classic reference colors with the numbers 1, 2, 4, 7, and / or 10 can be used.
[0026] According to one embodiment, the device comprises a communication interface configured to receive the facial image via a communication network and to store the facial image in the memory. This provides the advantage that the facial image, including the person's face and the reference color strip, can be captured and preprocessed at any location, and a color cast of the facial image can be reduced centrally by the device. The communication network can be the Internet, for example.
[0027] According to a second aspect, the invention relates to an access control arrangement. The access control arrangement comprises an image camera configured to capture a facial image of a person, wherein the facial image comprises an image of the person's face and a reference color stripe, and a device according to the first aspect of the invention, wherein the device is configured to reduce a color cast of the facial image of the person in order to obtain a corrected facial image. For example, the reference color stripe can be applied to a service or membership card of the person who presents it and holds it within a detection range of the image camera in order to gain access.
[0028] According to one embodiment, the reference color strip is part of the access control arrangement and is arranged in a detection area of the image camera, for example, on the access control arrangement. This provides the advantage that the position of the reference color strip in the facial image can be known in advance.
[0029] According to one embodiment, the access control arrangement comprises an identification device configured to compare the corrected facial image of the person with a pre-stored reference facial image in order to identify the person. This provides the advantage that efficient identification of the person can be achieved using a facial recognition algorithm. The corrected facial image and the pre-stored reference facial image can be color images.
[0030] According to one embodiment, the access control arrangement comprises an illumination device configured to illuminate the person's face and the reference color strip with light of a predetermined color temperature. This provides the advantage that the color cast of the captured facial image can be reduced at its source. The person's face and the reference color strip can be illuminated identically. The predetermined color temperature can be, for example, 6504 K.
[0031] According to one embodiment, the light from the illumination device has a predetermined emission spectrum. This provides the advantage that the reduction of the color cast can be carried out efficiently. The predetermined emission spectrum can, for example, be a D65 emission spectrum.
[0032] According to a third aspect, the invention relates to a method for reducing a color cast of a facial image of a person using a device, wherein the facial image comprises an image of the person's face and a reference color strip, wherein the reference color strip comprises a reference color field, and wherein a predetermined reference color is assigned to the reference color field. The device comprises a memory and a processor, wherein the memory is configured to store the facial image.The method includes detecting the reference color stripe in the face image by the processor, detecting the reference color patch in the detected reference color stripe by the processor, extracting a color from the detected reference color patch by the processor, comparing the extracted color with the predetermined reference color to obtain a color deviation by the processor, and reducing the color cast of the face image based on the color deviation by the processor to obtain a corrected face image.
[0033] The method can be carried out by the device and / or the access control arrangement. Further features of the method result directly from the features or functionality of the device and / or the access control arrangement.
[0034] According to a fourth aspect, the invention relates to a computer program with program code for executing the method according to the third aspect of the invention. The device and / or the access control arrangement can be configured to execute the program code.
[0035] The invention can be implemented in hardware and / or software.
[0036] Further embodiments are explained in more detail with reference to the accompanying drawings. They show: Fig. 1 a schematic diagram of an apparatus for reducing a color cast of a person's facial image; Fig. 2 a schematic diagram of an access control arrangement comprising a device, an image camera, an identification device and a lighting device; Fig. 3a schematic diagram of a method for reducing a color cast of a facial image of a person using an apparatus; Fig. 4 a schematic diagram of a face image comprising an image of a person's face and a reference color stripe; Fig. 5 a schematic diagram of a reference color strip with a plurality of reference color fields; Fig. 6 a schematic diagram of a reference color strip with a plurality of reference color fields; Fig. 7 a schematic diagram of a section of a CIE L*a*b* color space with an a* color coordinate and a b* color coordinate; and Fig. 8 a schematic diagram of a section of a CIE L*a*b* color space with an a* color coordinate and a b* color coordinate.
[0037] Fig. 1shows a schematic diagram of a device 100 for reducing a color cast of a facial image of a person, wherein the facial image comprises an image of a face of the person and a reference color strip. The reference color strip comprises a reference color field, wherein a predetermined reference color is assigned to the reference color field. The device 100 comprises a memory 101 configured to store the facial image. The device 100 further comprises a processor 103 configured to detect the reference color strip in the facial image, detect the reference color field in the detected reference color strip, extract a color from the detected reference color field, compare the extracted color with the predetermined reference color to obtain a color deviation, and reduce the color cast of the facial image based on the color deviation to obtain a corrected facial image.The device 100 further comprises a communication interface 105, which is configured to receive the facial image via a communication network and to store the facial image in the memory 101.
[0038] Fig. 2shows a schematic diagram of an access control arrangement 200 with a device 100, an image camera 201, an identification device 203, and an illumination device 205. The image camera 201 is configured to capture a facial image of a person, wherein the facial image comprises an image of the person's face and a reference color strip. The device 100 is configured to reduce a color cast of the person's facial image to obtain a corrected facial image. The identification device 203 is configured to compare the corrected facial image of the person with a pre-stored reference facial image to identify the person. The illumination device 205 is configured to illuminate the person's face and the reference color strip with light of a predetermined color temperature. The light can further have a predetermined emission spectrum.
[0039] Fig. 3shows a schematic diagram of a method 300 for reducing a color cast of a facial image of a person using a device, wherein the facial image comprises an image of a face of the person and a reference color strip, wherein the reference color strip comprises a reference color field, and wherein a predetermined reference color is assigned to the reference color field. The device comprises a memory and a processor, wherein the memory is configured to store the facial image.The method 300 includes the processor detecting 301 the reference color stripe in the facial image, the processor detecting 303 the reference color field in the detected reference color stripe, the processor extracting 305 a color from the detected reference color field, the processor comparing 307 the extracted color with the predetermined reference color to obtain a color deviation, and the processor reducing 309 the color cast of the facial image based on the color deviation to obtain a corrected facial image. The method 300 can be performed by the device and / or the access control arrangement.
[0040] Fig. 4shows a schematic diagram of a facial image 401 with an image of a person's face 403 and a reference color strip 405. The reference color strip 405 comprises a reference color field 407, which is assigned a predetermined reference color, and another reference color field 409, which is assigned another predetermined reference color. The reference color strip 405 can have additional reference color fields. Furthermore, the facial image 401 can comprise an image of another such reference strip.
[0041] The reference color stripe 405 and / or the further reference color stripe may further comprise a texture field, wherein the texture field may have a predetermined geometric texture. The predetermined geometric texture may, for example, be a two-dimensional pattern with a plurality of points or lines. The predetermined geometric texture may further indicate an orientation of the reference color stripe 405 or the further reference color stripe. The processor may be configured to detect the reference color stripe 405 or the further reference color stripe in the facial image 401 based on the predetermined geometric texture, for example, using pattern recognition or edge detection.
[0042] The reference color stripe 405 and / or the further reference color stripe are typically arranged in an environment of the person when the facial image 401 is captured, so that they are visible in the facial image 401.
[0043] Fig. 5 shows a schematic diagram of a reference color strip 405 with a plurality of reference color patches. The reference color strip 405 is formed by a standardized X-Rite or Macbeth ColorChecker Classic. A plurality of predetermined reference colors are assigned to the plurality of reference color patches.
[0044] According to one embodiment (before November 2014), the predetermined reference colors of the X-Rite or Macbeth ColorChecker Classic have the following values for the color coordinates {L*, a*, b*} in a CIE L*a*b* color space: Field name L* a* b* A1 37,986 13,555 14,059 A2 62,661 36,067 57,096 A3 28,778 14,179 -50,297 A4 96,539 -0,425 1,186 B1 65,711 18,13 17,81 B2 40,02 10,41 -45,964 B3 55,261 -38,342 31,37 B4 81,257 -0,638 -0,335 C1 49,927 -4,88 -21,905 C2 51,124 48,239 16,248 C3 42,101 53,378 28,19 C4 66,766 -0,734 -0,504 D1 43,139 -13,095 21,905 D2 30,325 22,976 -21,587 D3 81,733 4,039 79,819 D4 50,867 -0,153 -0,27 E1 55,112 8,844 -25,399 E2 72,532 -23,709 57,255 E3 51,935 49,986 -14,574 E4 35,656 -0,421 -1,231 F1 70,719 -33,397 -0,199 F2 71,941 19,363 67,857 F3 51,038 -28,631 -28,638 F4 20,461 -0,079 -0,973
[0045] According to a further embodiment (as of November 2014), the predetermined reference colors of the X-Rite or Macbeth ColorChecker Classic have the following values for the color coordinates {L*, a*, b*} in a CIE L*a*b* color space: Field name L* a* b* A1 37,54 14,37 14,92 A2 62,73 35,83 56,5 A3 28,37 15,42 -49,8 A4 95,19 -1,03 2,93 B1 64,66 19,27 17,5 B2 39,43 10,75 -45,17 B3 54,38 -39,72 32,27 B4 81,29 -0,57 0,44 C1 49,32 -3,82 -22,54 C2 50,57 48,64 16,67 C3 42,43 51,05 28,62 C4 66,89 -0,75 -0,06 D1 43,46 -12,74 22,72 D2 30,1 22,54 -20,87 D3 81,8 2,67 80,41 D4 50,76 -0,13 0,14 E1 54,94 9,61 -24,79 E2 71,77 -24,13 58,19 E3 50,63 51,28 -14,12 E4 35,63 -0,46 -0,48 F1 70,48 -32,26 -0,37 F2 71,51 18,24 67,37 F3 49,57 -29,71 -28,32 F4 20,64 0,07 -0,46
[0046] The X-Rite or Macbeth ColorChecker Classic reference color with the number 1 has the patch designation A1, the reference color with the number 2 has the patch designation B1, the reference color with the number 4 has the patch designation D1, the reference color with the number 7 has the patch designation A2 and the reference color with the number 10 has the patch designation D2.
[0047] Fig. 6 shows a schematic diagram of a reference color strip 405 with a plurality of reference color patches. The reference color strip 405 is formed by a standardized reference color strip according to IEC 61966-8. A plurality of predetermined reference colors are assigned to the plurality of reference color patches.
[0048] Fig. 7 shows a schematic diagram of a section of a CIE L*a*b* color space 701 with an a* color coordinate 703 and a b* color coordinate 705. Typical human skin colors, hair colors, and iris colors are highlighted in the diagram. Furthermore, sRGB color ranges for a luminance L* of 50 and 75 are shown.
[0049] The CIE L*a*b* color space 701 is based on the principle that complementary colors exist, such as red / green (for a*) or yellow / blue (for b*). The "*" denotes a dual vector space. L* denotes the luminance, which represents the perceived brightness of a pixel independent of its color value.
[0050] Skin colors typically fall in a range with an a* color coordinate between 5 and 14 and a b* color coordinate between 10 and 24. Hair colors typically fall in a range with an a* color coordinate and a b* color coordinate between -5 and 15. Iris colors typically fall in a range with an a* color coordinate and a b* color coordinate between -10 and 30.
[0051] Fig. 8shows a schematic diagram of a section of a CIE L*a*b* color space 701 with an a* color coordinate 703 and a b* color coordinate 705. The squares each represent a predetermined reference color. The circles each represent a corresponding extracted color from a face image.
[0052] The processor can be configured to determine a color deviation, which can be formed by a Euclidean distance between the color coordinates of the respective extracted color and the color coordinates of the associated predetermined reference color. By reducing the color cast of the facial image by the processor, the extracted colors, indicated by circles, can be aligned with the predetermined reference colors, indicated by squares. This can be achieved, for example, by means of a color transformation, in particular a shift, scaling, and / or rotation in the color space.
[0053] The concept enables a facial image 401 to be captured in a straightforward manner and with little effort, for example during registration, but also for verification purposes, for example by means of an access control arrangement 200, in such a way that realistic colors, in particular of skin and hair, can be restored even after printing and scanning.
[0054] When capturing precise colors, lighting conditions prove particularly challenging. Theoretically, a light emission spectrum such as D65 should be used as a reference image when capturing a facial image 401. In practice, however, this is difficult to achieve in heterogeneous systems, such as when applying for an identification document. Light typically has properties that are difficult to predict. Even achieving uniform illumination and avoiding excessive shadows in the eye sockets or reflections on glasses are a challenge for passport authorities, for example.
[0055] In the CIE L*a*b* color space 701, a person's skin, hair, and iris colors typically lie within a comparatively narrow range. Simply reducing a color cast by using an X-Rite or Macbeth ColorChecker Classic as the 405 reference color strip, particularly the reference color patches numbered 1 and 2, significantly improves the color quality of skin tones. Alternatively, a different 405 reference color strip can be used, for example, a reference color strip according to IEC 61966-8 in the February 2001 version. If the three reference color patches numbered 4, 7, and 10 of the X-Rite or Macbeth ColorChecker Classic are used in addition to the reference color patches numbered 1 and 2, the color quality can be further improved. The evaluation of additional reference color patches may be omitted.
[0056] Consequently, a reference color strip 405 with at least one reference color field 407, 409, for example, two or five reference color fields, can be arranged at a non-disruptive location, which is illuminated like the person's face 403. For example, the reference color strip 405 can be held next to the person's face 403 when capturing the facial image 401. Furthermore, the reference color strip 405 can be mounted on a rear wall, but far enough away from the face 403 to maintain a uniform background, if required by a standard. The rear wall and the face 403 can be illuminated similarly.
[0057] However, in certain environments, such as certain access control systems, a rear wall may not be practical. However, two image cameras can be used there, for example, with one image camera capturing the entire surroundings with a wide angle and locating the person's face 403, and another image camera subsequently capturing only the person's face 403 in high resolution. Therefore, the reference color stripe 405 can be visible in the field of view of an image camera 201 or of image cameras and illuminated like the person's face 403. If a singling device is used in an access control system 200, the reference color stripe 405 can be arranged, for example, at its edge or on an archway at the entrance to the singling device. If multiple image cameras are used, they can each have the same color calibration.
[0058] If the reference color stripe 405 is visible in the facial image 401, it can be printed out, for example. This typically distorts the colors. Upon subsequent scanning of the printed facial image 401, these colors are further distorted. However, at least one predetermined reference color with original color coordinates is now known in the scanned facial image 401. Consequently, color correction, in particular, a reduction of a color cast, can be performed.
[0059] If not just one reference color field 407 with a predetermined reference color is used, but rather several reference color fields with the predetermined reference color, the respectively extracted colors can be averaged. Furthermore, facial images of people of different sizes can be captured without retooling. In addition to intensity, texture, and contrast information of the facial image 401, color information can thus be used for facial recognition or identification of a person. Furthermore, it is easily achieved that the facial image 401 can be represented in an identification document with realistic colors, even if colors were falsified during the capture or preprocessing process. In particular, the use of color-calibrated equipment in the capture or preprocessing process can be dispensed with.
[0060] In summary, for example, a reference color strip 405 comprising reference color fields 407, 409 with predetermined reference colors can be arranged in the field of view of an image camera 201. The reference color strip 405 can be arranged such that color corrections in the facial image 401 are possible regardless of the lighting and preprocessing steps used.
[0061] For example, in a photo studio, the reference color strip 405 can be applied to a uniformly colored background at a position that is far enough away from the face 403, but is still visible in the facial image 401. The reference color strip 405 can be an X-Rite or Macbeth ColorChecker Classic reference color strip. Furthermore, another such reference color strip can be arranged on a different side of the face 403. The reference color strip 405 can also be held next to the face 403 of the person when the facial image 401 is captured. When used in an access control arrangement 200, the reference color strip 405 can be arranged, for example, on a side of an entrance gate of a separation device of the access control arrangement 200 facing the image camera 201. Consequently, additional color information for face recognition orused to identify the person. LIST OF REFERENCE SYMBOLS
[0062] 100Device 101Memory 103Processor 105Communication interface 200Access control device 201Image camera 203Identification device 205Lighting device 300Method for reducing a color cast of a facial image 301Detecting the reference color stripe 303Detecting the reference color patch 305Extracting a color 307Comparing the extracted color 309Reducing the color cast of the facial image 401Face image 403Face 405Reference color strip 407Reference color patch 409Additional reference color patch 701CIE L*a*b* color space 703a* color coordinate 705b* color coordinate
Claims
1. A device (100) for reducing a color cast of a facial image (401) of a person, wherein the facial image (401) comprises an image of a face (403) of the person and a reference color strip (405), wherein the reference color strip (405) comprises a reference color field (407), and wherein a predetermined reference color is assigned to the reference color field (407), the device comprising: a memory (101), which is configured to store the facial image (401); and a processor (103), which is configured to detect the reference color strip (405) in the facial image (401), detect the reference color field (407) in the detected reference color strip (405), extract a color from the detected reference color field (407), compare the extracted color with the predetermined reference color to obtain a color deviation, and reduce the color cast of the facial image (401) based on the color deviation in order to obtain a corrected facial image, characterized in that the processor (103) is configured to automatically remove the reference color strip (405) from the corrected facial image.
2. The device (100) of claim 1, wherein the reference color strip (405) comprises a further reference color field (409), wherein the further reference color field (409) is assigned a further predetermined reference color, and wherein the processor (103) is configured to detect the further reference color field (409) in the detected reference color strip (405), extract a further color from the detected further reference color field (409), compare the extracted further color with the further predetermined reference color in order to obtain a further color deviation, and further reduce the color cast of the facial image (401) based on the further color deviation.
3. The device (100) according to one of the preceding claims, wherein the reference color strip (405) has a predetermined geometric shape, and wherein the processor (103) is configured to detect the reference color strip (405) in the facial image (401) on the basis of the predetermined geometric shape.
4. The device (100) according to one of the preceding claims, wherein the processor (103) is configured to detect the face (403) of the person in the facial image (401), to determine a section of the facial image (401), wherein the section includes the detected face (403) of the person, and wherein the section does not include the detected reference color strip (405), and to limit the corrected facial image to the determined section.
5. The device (100) according to one of the preceding claims, wherein the processor (103) is configured to determine a position of the detected reference color strip (405) in the facial image (401) or the corrected facial image, and to replace the reference color strip (405) at the determined position in the corrected facial image with a predetermined color.
6. The device (100) according to one of the preceding claims, wherein the color deviation is formed by a Euclidean distance between color coordinates (703, 705) of the extracted color and color coordinates (703, 705) of the predetermined reference color.
7. The device (100) according to one of the preceding claims, wherein the processor (103) is configured to convert the facial image (401) into a CIE L*a*b* color space (701).
8. The device (100) according to one of the preceding claims, wherein the predetermined reference color can be represented in a CIE L*a*b* color space (701), and wherein the predetermined reference color has an a* color coordinate (703) between 5 and 14 and a b* color coordinate (705) between 10 and 24.
9. The device (100) according to one of the preceding claims, comprising: a communication interface (105), which is configured to receive the facial image (401) via a communication network and to store the facial image (401) in the memory (101).
10. An access control arrangement (200), comprising: an image camera (201), which is configured to capture a facial image (401) of a person, wherein the facial image (401) comprises an image of a face (403) of the person and a reference color strip (405); and a device (100) according to one of claims 1 to 9, wherein the device (100) is configured to reduce a color cast of the facial image (401) of the person in order to obtain a corrected facial image.
11. The access control arrangement (200) according to claim 10, comprising: an identification device (203), which is configured to compare the corrected facial image of the person with a pre-stored reference facial image in order to identify the person.
12. The access control arrangement (200) according to one of claims 10 or 11, comprising: an illumination device (205), which is configured to illuminate the face (403) of the person and the reference color strip (405) with light of a predetermined color temperature.
13. The access control arrangement (200) according to claim 12, wherein the light of the illumination device (205) has a predetermined emission spectrum.
14. A method (300) for reducing a color cast of a facial image (401) of a person using a device (100), wherein the facial image (401) comprises an image of a face (403) of the person and a reference color strip (405), wherein the reference color strip (405) comprises a reference color field (407), wherein a predetermined reference color is assigned to the reference color field (407), wherein the device (100) comprises a memory (101) and a processor (103), and wherein the memory (101) is configured to store the facial image (401), the method comprising: detecting (301) the reference color strip (405) in the facial image (401) by the processor (103); detecting (303) the reference color field (407) in the detected reference color strip (405) by the processor (103); extracting (305) a color from the detected reference color field (407) by the processor (103); comparing (307) the extracted color with the predetermined reference color by the processor (103) to obtain a color deviation; and reducing (309) the color cast of the facial image (401) based on the color deviation by the processor (103) in order to obtain a corrected facial image, characterized in automated removal of the reference color strip (405) from the corrected facial image by the processor (103).
15. A computer program having program code for executing the method (300) according to claim 14.