Robust feature extraction for color and grayscale images

• US 20080144932A1
• Filed: 12/13/2006
• Published: 06/19/2008
• Est. Priority Date: 12/13/2006
• Status: Active Grant

• Current Assignee: Adobe Inc.
• Original Assignee: Adobe Systems Incorporated (Adobe Inc.)
• Inventors: Chien, Jen-Chan, Jin, Hailin
• Granted Patent: US 8,103,102 B2
• Time in Patent Office: Days
• US Class Current: 382/169
• CPC Class Codes: G06T 5/50 using two or more images, e.g. averaging or subtraction
• CPC Class Codes: G06T 7/33 using feature-based methods
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]

Abstract

Systems, methods and program products for converting a first image to an intensity image using principal components analysis where the intensity image is based on a first principal component. The intensity image is analyzed to identify a plurality of scale invariant features in the intensity image, each identified feature associated with a score. An adaptive threshold is applied to the identified features to establish a set of features for the first image.

Claims

1. A computer-implemented method, comprising:

converting a first image to an intensity image using principal components analysis where the intensity image is based on a first principal component;

analyzing the intensity image to identify a plurality of scale invariant features in the intensity image, each identified feature associated with a score; and

applying an adaptive threshold to the identified features to establish a set of features for the first image.

2. The method of claim 1 where the intensity image includes a plurality of pixels and each pixel is associated with a respective intensity value, the method further comprising:

determining a number of pixels in the intensity image that represent underexposed or overexposed areas of the first image based on the respective intensity values; and

if the number of pixels is above a threshold, scaling the respective intensity values of the plurality of pixels to a middle intensity value.

3. The method of claim 1 where applying the threshold includes:

discarding identified features whose respective scores are below a threshold.

4. The method of claim 1 where applying the threshold includes:

determining a desired number of features based on a size of the first image; and

keeping the desired number of identified features whose respective scores are highest.

5. The method of claim 1, where:

the first image is a lower resolution version of a second image; and

mapping the set of features from the first image to the second image.

6. The method of claim 1, further comprising:

identifying a non-constant tile in a second image; and

using the non-constant tile as the first image.

7. The method of claim 1, where:

the first image is a non-constant tile in a second image, the second imaging being larger than the first image.

8. The method of claim 1 where the first image is in one of the following color spaces:

RGB, CIE L*a*b*, CMYK, HSL, HVL, grayscale, or black and white.

9. The method of claim 1 where:

the first image includes a plurality of tiles; and

performing the method of claim 1 on each tile in parallel.

10. The method of claim 9 where:

the adaptive threshold is applied to each tile independently.

11. The method of claim 1, further comprising:

accepting user selection of one or more regions of the first image.

12. The method of claim 11 where:

features detected in the one or more regions are associated with a priority;

the one or more regions indicate one or more areas of the first image where feature detection will or will not be performed;

or

the one or more regions indicate one or more areas of the first image where features will be given priority when applying the adaptive threshold.

13. A computer-implemented method, comprising:

converting a first CIE L*a*b* image to an intensity image using an L* component of the image;

analyzing the intensity image to identify a plurality of scale invariant features in the intensity image, each identified feature associated with a score; and

applying an adaptive threshold to the identified features to establish a set of features for the first image.

14. The method of claim 13 where the intensity image includes a plurality of pixels and each pixel is associated with a respective intensity value, the method further comprising:

determining a number of pixels in the intensity image that represent underexposed or overexposed areas of the first image based on the respective intensity values; and

if the number of pixels is above a threshold, scaling the respective intensity values of the plurality of pixels to a middle intensity value.

15. The method of claim 13 where applying the threshold includes:

discarding identified features whose respective scores are below a threshold.

16. The method of claim 13 where applying the threshold includes:

determining a desired number of features based on a size of the first image; and

keeping the desired number of identified features whose respective scores are highest.

17. The method of claim 13, where:

the first image is a lower resolution version of a second image; and

mapping the set of features from the first image to the second image.

18. The method of claim 13, further comprising:

identifying a non-constant tile in a second image; and

using the non-constant tile as the first image.

19. The method of claim 13, where:

the first image is a non-constant tile in a second image, the second imaging being larger than the first image.

20. The method of claim 13 where:

the first image includes a plurality of tiles; and

performing the method of claim 13 on each tile in parallel.

21. The method of claim 20 where:

the adaptive threshold is applied to each tile independently.

22. The method of claim 13, further comprising:

accepting user selection of one or more regions of the first image.

23. The method of claim 22 where:

features detected in the one or more regions are associated with a priority;

the one or more regions indicate one or more areas of the first image where feature detection will or will not be performed;

or

the one or more regions indicate one or more areas of the first image where features will be given priority when applying the adaptive threshold.

24. A computer program product, encoded on a computer-readable medium, operable to cause data processing apparatus to perform operations comprising:

converting a first image to an intensity image using principal components analysis where the intensity image is based on a first principal component;

analyzing the intensity image to identify a plurality of scale invariant features in the intensity image, each identified feature associated with a score; and

applying an adaptive threshold to the identified features to establish a set of features for the first image.

25. The program product of claim 24 where the intensity image includes a plurality of pixels and each pixel is associated with a respective intensity value, the program product further comprising:

determining a number of pixels in the intensity image that represent underexposed or overexposed areas of the first image based on the respective intensity values; and

if the number of pixels is above a threshold, scaling the respective intensity values of the plurality of pixels to a middle intensity value.

26. The program product of claim 24 where applying the threshold includes:

determining a desired number of features based on a size of the first image; and

keeping the desired number of identified features whose respective scores are highest.

27. The program product of claim 24, where:

the first image is a lower resolution version of a second image; and

mapping the set of features from the first image to the second image.

28. The program product of claim 24, further comprising:

accepting user selection of one or more regions of the first image.

29. The program product of claim 28 where:

features detected in the one or more regions are associated with a priority;

the one or more regions indicate one or more areas of the first image where feature detection will or will not be performed;

or

the one or more regions indicate one or more areas of the first image where features will be given priority when applying the adaptive threshold.

30. A computer program product, encoded on a computer-readable medium, operable to cause data processing apparatus to perform operations comprising:

converting a first CIE L*a*b* image to an intensity image using an L* component of the image;

analyzing the intensity image to identify a plurality of scale invariant features in the intensity image, each identified feature associated with a score; and

applying an adaptive threshold to the identified features to establish a set of features for the first image.

31. The program product of claim 30 where the intensity image includes a plurality of pixels and each pixel is associated with a respective intensity value, the program product further comprising:

determining a number of pixels in the intensity image that represent underexposed or overexposed areas of the first image based on the respective intensity values; and

if the number of pixels is above a threshold, scaling the respective intensity values of the plurality of pixels to a middle intensity value.

32. The program product of claim 30 where applying the threshold includes:

determining a desired number of features based on a size of the first image; and

keeping the desired number of identified features whose respective scores are highest.

33. The program product of claim 30, where:

the first image is a lower resolution version of a second image; and

mapping the set of features from the first image to the second image.

34. The program product of claim 30, further comprising:

accepting user selection of one or more regions of the first image.

35. The program product of claim 34 where:

features detected in the one or more regions are associated with a priority;

the one or more regions indicate one or more areas of the first image where feature detection will or will not be performed;

or

the one or more regions indicate one or more areas of the first image where features will be given priority when applying the adaptive threshold.

36. A system for detecting image features, the system comprising:

a display for presenting a target display window to a user;

one or more control interfaces configured to receive input from the user; and

processor electronics configured to perform operations comprising;

converting a first image to an intensity image using principal components analysis where the intensity image is based on a first principal component;

analyzing the intensity image to identify a plurality of scale invariant features in the intensity image, each identified feature associated with a score; and

applying an adaptive threshold to the identified features to establish a set of features for the first image.

37. The system of claim 36 where the intensity image includes a plurality of pixels and each pixel is associated with a respective intensity value, the system further comprising:

determining a number of pixels in the intensity image that represent underexposed or overexposed areas of the first image based on the respective intensity values; and

if the number of pixels is above a threshold, scaling the respective intensity values of the plurality of pixels to a middle intensity value.

38. The system of claim 36 where applying the threshold includes:

discarding identified features whose respective scores are below a threshold.

39. The system of claim 36 where applying the threshold includes:

determining a desired number of features based on a size of the first image; and

keeping the desired number of identified features whose respective scores are highest.

40. The system of claim 36, where:

the first image is a lower resolution version of a second image; and

mapping the set of features from the first image to the second image.

41. The system of claim 36, further comprising:

identifying a non-constant tile in a second image; and

using the non-constant tile as the first image.

42. The system of claim 36, where:

the first image is a non-constant tile in a second image, the second imaging being larger than the first image.

43. The system of claim 36 where the first image is in one of the following color spaces:

RGB, CIE L*a*b*, CMYK, HSL, HVL, grayscale, or black and white.

44. The system of claim 36 where:

the first image includes a plurality of tiles; and

the adaptive threshold is applied to each tile independently.

45. The system of claim 36, further comprising:

accepting user selection of one or more regions of the first image.

46. The system of claim 45 where:

features detected in the one or more regions are associated with a priority;

the one or more regions indicate one or more areas of the first image where feature detection will or will not be performed;

or

the one or more regions indicate one or more areas of the first image where features will be given priority when applying the adaptive threshold.