IMAGE PREPROCESSING

• US 20120300991A1
• Filed: 07/27/2012
• Published: 11/29/2012
• Est. Priority Date: 01/25/2010
• Status: Active Grant

• Current Assignee: Apple Inc.
• Original Assignee: Apple Inc.
• Inventors: Free, Robert Mikio
• Granted Patent: US 8,358,813 B2
• Time in Patent Office: Days
• US Class Current: 382/118
• CPC Class Codes: G06V 40/162 using pixel segmentation or colour matching

Abstract

A method, performed by a software process executing on a computer system, includes accessing a digital image comprising a plurality of pixels. The method also includes determining whether one or more pixels bounding a first rectangular sub-region of a predetermined size within the digital image satisfy a specified criterion. If a predetermined percentage of bounding pixels satisfy the specified criterion, the method assumes that all pixels within the first rectangular sub-region also satisfy the specified criterion. The method further includes selectively executing an image analysis algorithm on the digital image using the assumption that all pixels within the rectangular sub-region also satisfy the specified criterion.

Claims

1. A method performed by a software process executing on a computer system, the method comprising:

accessing a digital image comprising a plurality of pixels;

determining, by a hardware processor, whether one or more pixels bounding a first rectangular sub-region of a predetermined size within the digital image satisfy a specified criterion;

if a predetermined percentage of only the bounding pixels satisfy the specified criterion, assuming that all pixels within the first rectangular sub-region also satisfy the specified criterion; and

selectively performing, by the hardware processor, an image analysis algorithm on the digital image using the assumption that all pixels within the first rectangular sub-region also satisfy the specified criterion.

2. The method of claim 1, wherein the first rectangular sub-region is selected by using a filter kernel a first predetermined number of pixels high and a second predetermined number of pixels wide, where the filter kernel includes one or more other pixels different from the bounding pixels.

3. The method of claim 2, wherein the filter kernel is used to select a plurality of rectangular sub-regions in the digital image such that each pixel of the digital image is included in at least one of the plurality of rectangular sub-regions.

4. The method of claim 2, wherein the filter kernel is used to select a second rectangular sub-region if the predetermined percentage of only the bounding pixels satisfy the specified criterion, wherein the second rectangular sub-region is non-overlapping with the first rectangular sub-region.

5. The method of claim 2, wherein the filter kernel is used to select a third rectangular sub-region if the predetermined percentage of only the bounding pixels do not satisfy the specified criterion, wherein the third rectangular sub-region partially overlaps the first rectangular sub-region.

6. The method of claim 2, wherein the first and second predetermined numbers are chosen based on a hardware related data word size that is optimized to speed execution.

7. The method of claim 2, further comprising performing, by the hardware processor, a second pass wherein another rectangular sub-region is selected using a second filter kernel a third predetermined number of pixels high and a fourth predetermined number of pixels wide.

8. The method of claim 7, wherein at least one of the third and fourth predetermined numbers is relatively prime to the first and second predetermined numbers, respectively.

9. The method of claim 1, wherein the specified criterion is whether the one or more pixels represent orange hue.

10. The method of claim 1, wherein the image analysis algorithm comprises a face detection algorithm.

11. The method of claim 1, wherein the pixels within the first rectangular sub-region are masked.

12. The method of claim 11, wherein the masked pixels are ignored in performing the image analysis algorithm.

13. The method of claim 1, wherein the specified criterion is configured to remove noise from the digital image.

14. A method performed by a software process executing on a computer system, the method comprising:

accessing a digital image comprising a plurality of pixels;

iteratively moving, by a hardware processor, a logical kernel across the digital image to identify pixels within the digital image that can be assumed to satisfy a specified condition, such that the logical kernel is moved between consecutive positions based on determining whether a predetermined percentage of only bounding pixels of the logical kernel satisfy the specified condition; and

executing, by a hardware processor, an image analysis algorithm on fewer than all pixels in the digital image based on the identified pixels.

15. The method of claim 14, wherein the image analysis algorithm comprises a face detection algorithm.

16. The method of claim 14, wherein the logical kernel is a first predetermined number of pixels high and a second predetermined number of pixels wide.

17. The method of claim 16, wherein the first and second predetermined numbers are chosen based on a hardware related data word size that is optimized to speed execution.

18. The method of claim 14, wherein the logical kernel is used to select a plurality of rectangular sub-regions in the digital image such that each pixel of the digital image is included in at least one of the plurality of rectangular sub-regions.

19. The method of claim 14, wherein the logical kernel is moved from a first position to a non overlapping second position if the predetermined percentage of only the bounding pixels of the logical kernel satisfy the specified condition.

20. The method of claim 14, wherein the logical kernel is moved from a first position to a partially overlapping second position if the predetermined percentage of only the bounding pixels do not satisfy the specified condition.

21. The method of claim 14, further comprising performing a second pass using a second logical kernel a third predetermined number of pixels high and a fourth predetermined number of pixels wide.

22. The method of claim 21, wherein at least one of the third and fourth predetermined numbers is relatively prime to the first and second predetermined numbers, respectively.

23. A system comprising:

a data processing apparatus; and

a storage device storing instructions that when executed by the data processing apparatus cause the system to perform operations comprising;

accessing a digital image comprising a plurality of pixels;

determining whether one or more pixels bounding a first rectangular sub-region of a predetermined size within the digital image satisfy a specified criterion;

if a predetermined percentage of only the bounding pixels satisfy the specified criterion, assuming that all pixels within the first rectangular sub-region also satisfy the specified criterion; and

selectively applying an image analysis algorithm on the digital image using the assumption that all pixels within the first rectangular sub-region also satisfy the specified criterion.

24. The system of claim 23 wherein the specified criterion is whether the one or more pixels represent orange hue.

25. The system of claim 23, wherein the image analysis algorithm comprises a face detection algorithm.

26. The system of claim 23, wherein the first rectangular sub-region is selected by using a filter kernel a first predetermined number of pixels high and a second predetermined number of pixels wide.

27. The system of claim 26, wherein the filter kernel is used to select a second rectangular sub-region if the predetermined percentage of only the bounding pixels satisfy the specified criterion, wherein the second rectangular sub-region is non-overlapping with the first rectangular sub-region.

28. The system of claim 26, wherein the filter kernel is used to select a third rectangular sub-region if the predetermined percentage of bounding pixels do not satisfy the specified criterion, wherein the third rectangular sub-region partially overlaps the first rectangular sub-region.

29. The system of claim 23, wherein the pixels within the first rectangular sub-region are masked.

30. The system of claim 29, wherein the masked pixels are ignored in applying the image analysis algorithm.

31. A non-transitory computer readable medium having encoded thereon instructions that, when executed by a data processing apparatus, cause the data processing apparatus to perform operations comprising:

accessing a digital image comprising a plurality of pixels;

determining whether one or more pixels bounding a first rectangular sub-region of a predetermined size within the digital image satisfy a specified criterion;

if a predetermined percentage of only the bounding pixels satisfy the specified criterion, assuming that all pixels within the first rectangular sub-region also satisfy the specified criterion; and

selectively executing an image analysis algorithm on the digital image using the assumption that all pixels within the first rectangular sub-region also satisfy the specified criterion.

32. The non-transitory computer readable medium of claim 31, wherein the specified criterion is whether the one or more pixels represent orange hue.

33. The non-transitory computer readable medium of claim 31, wherein the image analysis algorithm comprises a face detection algorithm.

34. The non-transitory computer readable medium of claim 31, wherein the first rectangular sub-region is selected by using a filter kernel a first predetermined number of pixels high and a second predetermined number of pixels wide.

35. The non-transitory computer readable medium of claim 34, wherein the filter kernel is used to select a third rectangular sub-region if the predetermined percentage of only the bounding pixels do not satisfy the specified criterion, wherein the third rectangular sub-region partially overlaps the first rectangular sub-region.

36. The non-transitory computer readable medium of claim 34, wherein the filter kernel is used to select a third rectangular sub-region if the predetermined percentage of only the bounding pixels do not satisfy the specified criterion, wherein the third rectangular sub-region partially overlaps the first rectangular sub-region.

37. The non-transitory computer readable medium of claim 31, wherein the pixels within the first rectangular sub-region are masked.

38. The non-transitory computer readable medium of claim 37, wherein the masked pixels are ignored in performing the image analysis algorithm.