Method and system for skin color estimation from an image

• US 20070058860A1
• Filed: 09/09/2005
• Published: 03/15/2007
• Est. Priority Date: 09/09/2005
• Status: Active Grant

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Harville, Michael, Susstrunk, Sabine, Bhatti, Nina, Baker, Harlyn
• Granted Patent: US 7,522,769 B2
• Time in Patent Office: Days
• US Class Current: 382/167
• CPC Class Codes: G06V 40/162 using pixel segmentation or colour matching

Abstract

Embodiments of the present invention recite a computer implemented method and system for estimating skin color from an image. In embodiments of the present invention, an image of a subject and of an imaged reference color set is accessed. In embodiments of the present invention, a controlled infrastructure is not required when capturing the image. At least one skin pixel of the subject is located in the image and a color description of the at least one skin pixel is determined. Then, a color correction function to be applied to the color description of the at least one skin pixel is determined. The color correction function is applied to the color description of the at least one skin pixel to create a modified color description. Finally, the skin color of the subject is estimated based upon an analysis of the modified color description.

Claims

1. A computer implemented method for estimating skin color from at least one image, said method comprising:

accessing at least one image of a subject and of an imaged reference color set;

locating at least one skin pixel of said subject in said at least one image and determining a color description thereof;

determining a color correction function to be applied to said color description of said at least one skin pixel;

applying said color correction function to said color description of said at least one skin pixel to create a modified color description; and

generating a skin color estimate of said subject based upon an analysis of said modified color description.

2. The computer implemented method as recited in claim 1 wherein capturing said at least one image does not require a controlled infrastructure.

3. The computer implemented method as recited in claim 1 further comprising:

identifying said imaged reference color set within said at least one image; and

sampling at least one reference color from a known location relative to said imaged reference color set.

4. The computer implemented method as recited in claim 3 further comprising:

printing said imaged reference color set using a fading ink wherein the fading of said ink renders said imaged reference color set unusable when it is no longer valid.

5. The computer implemented method as recited in claim 3 wherein said determining further comprises:

calculating a first description corresponding to the observed characteristics of said at least one reference color; and

comparing said first description with a second description corresponding with the control characteristics of said at least one color;

discovering a discrepancy between said first description and said second description; and

determining said color correction function which is applied to said first description to substantially eliminate said discrepancy.

6. The computer implemented method as recited in claim 3 wherein said at least one reference color is representative of a human skin color.

7. The computer implemented method as recited in claim 1 further comprising:

processing said at least one image using an automatic face detection algorithm;

identifying said at least one skin pixel which is representative of the skin color of said subject;

applying said color correction function to said color description of said at least one skin pixel to create a modified skin color description; and

analyzing said modified skin color description to determine said skin color estimate.

8. The computer implemented method as recited in claim 1 wherein locating said at least one skin pixel further comprises:

clustering said at least one skin pixel in a reference color space.

9. The computer implemented method as recited in claim 1 wherein said locating further comprises:

generating a probability that said at least one skin pixel comprises human skin; and

using said probability when determining said skin color estimate.

10. The computer implemented method as recited in claim 1 wherein said generating further comprises:

outputting an estimate of the skin color of said subject in the form of a single vector value.

11. An image analysis system comprising:

an input for accessing at least one image of a subject and of an imaged reference color set;

a skin pixel selection component for locating a plurality of skin pixels of said subject in said at least one image and determining a color description thereof;

a color correction determiner for determining a color correction function to be applied to said color description of said plurality of skin pixels;

a color correction applicator for applying said color correction function to said color description of said plurality of skin pixels to create a modified color description; and

a skin color estimator for generating a skin color estimate of said subject based upon an analysis of said modified color description.

12. The image analysis system of claim 11 wherein the capturing of said at least one image does not require a controlled infrastructure.

13. The image analysis system of claim 11 wherein said color correction determiner further identifies said imaged reference color set within said at least one image and samples at least one reference color from a known location relative to said imaged reference color set.

14. The image analysis system of claim 13 wherein said color correction determiner is further for comparing a first description corresponding to the observed characteristics of said at least one color with a second description corresponding with the control characteristics of said at least one color to discover a discrepancy between said first description and said second description and is further for determining said color correction function which, when applied to said first description, substantially eliminates said discrepancy.

15. The image analysis system of claim 13 wherein said at least one reference color is representative of a human skin color.

16. The image analysis system of claim 13 wherein said color correction applicator is for applying said color correction function to said color description of said at least one skin pixel to create a corresponding modified skin color description and wherein said skin color estimator analyzes said modified skin color description to determine said skin color estimate.

17. The image analysis system of claim 11 wherein said skin pixel selection component uses a skin color detection algorithm to detect a color representative of human skin in said at least one image.

18. The image analysis system of claim 11 wherein said skin pixel selection component is for clustering said plurality of skin pixels in a reference color space.

19. The image analysis system of claim 11 wherein said skin pixel selection component is further for generating a probability that said at least one skin pixel comprises human skin and for using said probability when determining said skin color estimate.

20. The image analysis system of claim 11 wherein said skin color estimator outputs an estimate of the skin color of said subject in the form a single vector value.

21. A computer-usable medium having computer-readable program code embodied therein for causing a computer system to perform a method for estimating skin color from at least one image, said method comprising:

accessing at least one image of a subject and of an imaged reference color set

locating at least one skin pixel of said subject in said at least one image and determining a color description thereof;

determining a color correction function to be applied to said color description of said at least one skin pixel;

applying said color correction function to said color description of said at least one skin pixel to create a modified color description; and

generating a skin color estimate of said subject based upon an analysis of said modified color description.

22. The computer-usable medium of claim 21 wherein the capturing said at least one image does not require a controlled infrastructure.

23. The computer-usable medium of claim 21 further comprising:

identifying said imaged reference color set within said at least one image; and

sampling at least one reference color from a known location relative to said imaged reference color set.

24. The computer-usable medium of claim 23 wherein said determining further comprises:

calculating a first description corresponding to the observed characteristics of said at least one color; and

comparing said first description with a second description corresponding with the control characteristics of said at least one color;

discovering a discrepancy between said first description and said second description; and

determining said color correction function which is applied to said first description to substantially eliminate said discrepancy.

25. The computer-usable medium of claim 23 wherein said at least one reference color is representative of a human skin color.

26. The computer-usable medium of claim 23 wherein said method further comprises:

applying said color correction function to said color description of said at least one skin pixel to create a modified skin color description; and

analyzing said modified skin color description to determine said skin color estimate.

27. The computer-usable medium of claim 21 wherein said locating further comprises:

using a skin color detection algorithm to detect a color representative of human skin in said at least one image.

28. The computer-usable medium of claim 21 wherein locating said at least one skin pixel further comprises:

clustering said at least one skin pixel in a reference color space.

29. The computer-usable medium of claim 21 wherein said locating further comprises:

generating a probability that said at least one skin pixel comprises human skin; and

using said probability when determining said skin color estimate.

30. The computer-usable medium of claim 21 wherein said generating further comprises:

generating an estimate of the skin color of said subject in the form of a single vector value.