Digital image color correction device and method employing fuzzy logic

• US 6,272,239 B1
• Filed: 12/28/1998
• Issued: 08/07/2001
• Est. Priority Date: 12/30/1997
• Status: Expired due to Term

• Current Assignee: STMicroelectronics SRL (STMicroelectronics NV)
• Original Assignee: STMicroelectronics SRL (STMicroelectronics NV)
• Inventors: Colla, Federica, Mancuso, Massimo, Poluzzi, Rinaldo
• Primary Examiner(s): Tran, Phuoc
• Assistant Examiner(s): Alavi, Amir
• Application Number:
  US09/222,247
• Time in Patent Office: 953 Days
• Field of Search: 382/162, 382/163, 382/164, 382/165, 382/166, 382/167, 382/168, 382/169, 382/170, 382/171, 382/172, 382/232, 382/233, 382/234
• US Class Current: 382/167
• CPC Class Codes: G06T 7/90 Determination of colour characteristics
• CPC Class Codes: H04N 1/628 Memory colours, e.g. skin or sky
• CPC Class Codes: H04N 9/643 Hue control means, e.g. flesh tone control

Abstract

A digital image color correction device and method employing fuzzy logic, for correcting a facial tone image portion of a digital video image is provided. The device is a pixel fuzzifier unit (1) receiving in input a stream of pixels belonging to a sequence of correlated frames of a digital video image and computing a multilevel value representing a membership of each pixel to a skin color class; a global parameter estimator (2) receiving in input each of said pixel and the relative membership value, and computing a first and a second parameter which define the characteristics of a portion of said image that belongs to said skin color class; a processing unit (3) connected downstream to said global parameter estimator and to said pixel fuzzifier unit and adapted to correct each of the pixels of said portion of the image that belongs to said skin color class, according to said first global parameter (300), to obtain corrected pixels; and a processing switch (4) for outputting said pixels or said corrected pixels according to said second global parameter (400).

Claims

1. A digital image color correction device for correcting a facial tone image portion of a digital video image, the device comprising:

a pixel fuzzifier unit receiving in input a stream of pixels belonging to a sequence of correlated frames of a digital video image and computing a multilevel value representing a membership of each pixel to a skin color class;

a global parameter estimator receiving in input each of said pixels and the relative membership value, and computing a first and second parameter which define the characteristics of pixels in a portion of said image that belongs to said skin color class;

a processing unit connected downstream to said global parameter estimator and to said pixel fuzzifier unit and adapted to correct each of the pixels of said portion of the image that belongs to said skin color class, according to said first global parameter, to obtain corrected pixels, so that each pixel of said portion is corrected based upon other pixels in said portion; and

a processing switch for outputting said pixels or said corrected pixels according to said second global parameter.

2. The device according to claim 1, wherein said pixel fuzzifier unit comprises:

a contextual processing unit adapted to perform a spatial filtering on the pixels input to said pixel fuzzifier unit;

a conversion unit connected downstream to said contextual processing unit and adapted to perform a transformation of each pixel from a RGB color space to a HSV color space; and

a membership function evaluation unit connected to the output of said conversion unit and adapted to compute the membership value of each of said pixel with respect to said skin color class.

3. The device according to claim 1, wherein said global parameter estimator comprises:

an area estimation unit receiving in input said pixels and determining the areas of said skin color class inside said digital image; and

a coefficient estimation unit receiving in input the membership value of each of said pixels and adapted to compute, by means of fuzzy logic inference, correction coefficients for said pixels which belong to said skin color class.

4. The device according to claim 2, wherein said processing unit receives in input said pixels, said first global parameter, and the memberships of each of said pixels.

5. The device according to claim 4, wherein said processing unit comprises a conversion unit connected to a correction unit which is in turn connected to a back-conversion unit cascade connected to a soft-switch unit.

6. The device according to claim 5, wherein said conversion unit receives in input the pixels of the digital image and is the same as the conversion unit of the pixel fuzzifier unit.

7. The device according to claim 6, wherein said correction unit receives in unit the pixels color space-converted by said conversion unit and said second global parameter computed by the global parameter estimator and outputs corrected pixel which are then input to said back-conversion unit.

8. The device according to claim 7, wherein said softswitch unit receives in input the output of said back-conversion unit, said pixels and the membership values for each of said pixels, said soft-switch unit performing a weighted average to output furtherly corrected pixels which are then input to said processing switch unit.

9. The device according to claim 8, wherein said processing switch unit comprises a selector which received in input said pixels, said first global parameter and the output of said soft-switch unit, to output said pixels or said furtherly corrected pixels.

10. A method for color correcting a facial tone image portion of a digital video image composed by a sequence of correlated frames, the method comprising the steps of:

defining one or more fuzzy logic-defined sets to be used to identify a skin color class of a facial tone image portion of a digital video signal;

examining the digital video image to determine if a portion thereof is a facial tone image;

in case of positive response, dividing said facial tone image into spatial subregions and for each subregion calculating the mean saturation and the mean value thereof;

carrying out a color correction of the pixels of the facial tone image portion, a color correction of each pixel of the facial tone image portion being based upon the mean saturation and mean value of the spatial subregion corresponding to the pixel; and

outputting a corrected facial tone image portion.

11. The method according to claim 10, wherein the step of examining the image comprises applying fuzzy logic on the pixels belonging to the video signal image to determine the membership of each pixel to the skin color class.

12. The method according to claim 11, wherein said step of applying fuzzy logic comprises the steps of:

spatially filtering the video digital signal;

converting the pixel representation of the video digital signal from RGB color space representation to HSV color space representation; and

determining membership value of each pixel of the image to the skin color class.

13. The method according to claim 10, wherein said step of carrying out a color correction includes the step of computing parameters for correction of the pixels belonging to said facial tone image portion, by means of said mean saturation and mean value and further by means of prototypes of skin saturation and skin value.

14. The method according to claim 13, wherein the step of computing parameters for correction of the pixels belonging to said facial tone image portion comprises the steps of:

estimating the area of the skin color class inside the facial tone image portion; and

computing, by means of fuzzy inference, coefficients for the correction of the pixel of the facial tone image portion.

15. The method according to claim 14, wherein the area of the skin color class is evaluated on the basis of the percentage of pixels belonging to the class.

16. The method according to claim 14, wherein the coefficients for correction are computed by using fuzzy logic inference.

17. The method according to claim 14, wherein the step of correcting the pixels of said facial tone image portion comprises the steps of:

adding the coefficients to the coordinates of pixels of the facial tone image portion, the sign of said coefficients depending from predetermined fuzzy logic rules; and

back-converting the corrected pixels from HSV color space representation in RGB color space representation.

18. The method according to claim 17, wherein the step of correcting the pixels of the facial tone image portion further comprises the steps of:

making, for each pixel of the facial tone image portion, a weighted average of the pixel, of the corrected pixel and of the membership value to the skin class.

19. The method according to claim 18, further comprising the steps of:

determining if the skin color class has an area larger than a predetermined threshold;

in case of positive response, outputting the corrected pixels; and

in case of negative response, outputting the pixels without any correction.

20. The method of claim 10, wherein each spatial subregion is defined using fuzzy logic.

21. An apparatus for correcting a digital color image, comprising:

a first unit for determining whether a portion of the digital color image includes a predetermined tone;

a second unit for selectively dividing the predetermined tone image portion into spatial subregions;

a third unit for calculating the mean saturation and the mean value for each spatial subregion;

a fourth unit for determining a color correction of the pixels of the predetermined tone image portion, a color correction of each pixel of the predetermined tone image portion being based upon the mean saturation and mean value of the spatial subregion corresponding to the pixel; and

a fifth unit for selectively incorporating the color correction of the predetermined tone image portion into the digital color image.

22. The apparatus according to claim 21, wherein the first unit operates to perform fuzzy logic on the pixels belonging to the video signal image to determine the membership of each pixel to the class of the predetermined tone.

23. The method according to claim 21, wherein the second unit operates to compute parameters for correction of the pixels belonging to said predetermined tone image portion, based upon said mean saturation, said mean value and predetermined prototypes of predetermined saturation and value.

24. The apparatus according to claim 21, wherein the second unit operates to:

spatially filter the video digital signal;

convert the pixel representation of the video digital signal from RGB color space representation to HSV color space representation; and

determine membership value of each pixel of the image to the class of the predetermined tone.

25. The apparatus according to claim 21, wherein the second fourth unit operates to:

estimate the area of the class of the predetermined tone inside the image; and

compute, by means of fuzzy inference, coefficients for the correction of the pixel of the predetermined tone image portion.

26. The apparatus according to claim 21, wherein the area of the class of the predetermined tone is evaluated on the basis of the percentage of pixels belonging to the class.

27. The apparatus according to claim 21, wherein:

the second unit operates to determine if the class of the predetermined tone has an area larger than a predetermined threshold; and

the fifth unit outputs the corrected pixels in case of positive response, and outputs uncorrected pixels in case of negative response.

28. The apparatus of claim 21, wherein:

the second unit divides the predetermined tone image portion into a plurality of fuzzy logic subregions.

29. The apparatus of claim 21, wherein:

the second unit divides the predetermined tone image portion into a plurality of subregions using fuzzy logic.