Method for detecting sky in images

• US 6,504,951 B1
• Filed: 11/29/1999
• Issued: 01/07/2003
• Est. Priority Date: 11/29/1999
• Status: Expired due to Fees

• Current Assignee: Eastman Kodak Company
• Original Assignee: Eastman Kodak Company
• Inventors: Etz, Stephen, Luo, Jiebo
• Primary Examiner(s): Wu, Jingge
• Application Number:
  US09/450,190
• Time in Patent Office: 1,135 Days
• Field of Search: 382/162-167, 382/305, 382/224-228, 358/515-527, 345/597-618
• US Class Current: 382/165
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/187 involving region growing; involving region merging; involving connected component labelling
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation
• CPC Class Codes: G06T 7/90 Determination of colour characteristics
• CPC Class Codes: G06V 20/10 Terrestrial scenes scenes under surveillance with static cameras G06V20/52; scenes perceived from the exterior of a vehicle G06V20/56; scenes perceived from the interior of a vehicle G06V20/59

Abstract

A method, image recognition system, computer program, etc., for detecting sky regions in an image comprise classifying potential sky pixels in the image by color, extracting connected components of the potential sky pixels, eliminating ones of the connected components that have a texture above a predetermined texture threshold, computing desaturation gradients of the connected components, and comparing the desaturation gradients of the connected components with a predetermined desaturation gradient for sky to identify true sky regions in the image.

Claims

1. A method of detecting sky regions in an image comprising:

classifying potential sky pixels in said image by color;

extracting connected components of said potential sky pixels;

eliminating ones of said connected components that have a texture above a predetermined texture threshold;

computing desaturation gradients of said connected components; and

comparing said desaturation gradients of said connected components with a predetermined desaturation gradient for sky to identify true sky regions in said image.

2. The method in claim 1, wherein said desaturation gradients comprise desaturation gradients for red, green and blue trace components of said image.

3. The method in claim 1, wherein said predetermined desaturation gradient for sky comprises, from horizon to zenith, a decrease in red and green light trace components and a substantially constant blue light trace component.

4. The method in claim 1, wherein said classifying comprises;

forming a belief map of pixels in said image using a pixel classifier;

computing an adaptive threshold of sky color; and

classifying ones of said pixels that exceed said threshold as said potential sky pixels.

5. The method in claim 4, wherein said computing of said adaptive threshold comprises identifying a first valley in a belief histogram derived from said belief map.

6. The method in claim 5, wherein said belief map and said belief histogram are unique to said image.

7. The method in claim 1, further comprising determining a horizontal direction of a scene within said image by:

identifying a first gradient parallel to a width direction of said image;

identifying a second gradient perpendicular to said width direction of said image; and

comparing said first gradient and said second gradient, said horizontal direction of said scene being identified by the smaller of said first gradient and said second gradient.

8. A method of detecting sky regions in an image comprising:

computing desaturation gradients of regions in said image; and

comparing said desaturation gradients of said regions with a predetermined desaturation gradient for sky to identify true sky regions in said image.

9. The method in claim 8, further comprising classifying potential sky regions in said image by color.

10. The method in claim 8, further comprising eliminating ones of said regions that have a texture above a predetermined texture threshold.

11. The method in claim 8, wherein said desaturation gradients comprise desaturation gradients for red, green and blue trace components of said image.

12. The method in claim 8, wherein said predetermined desaturation gradient for sky comprises, from horizon to zenith, a decrease in red and green light trace components and a substantially constant blue light trace component.

13. The method in claim 9, wherein said classifying comprises:

forming a belief map of pixels in said image using a pixel classifier;

computing an adaptive threshold of sky color; and

classifying ones of said pixels that exceed said threshold as said potential sky pixels.

14. The method in claim 13, wherein said computing of said adaptive threshold comprises identifying a first valley in a belief histogram derived from said belief map.

15. The method in claim 14, wherein said belief map and said belief histogram are unique to said image.

16. The method in claim 8, further comprising determining a horizontal direction of a scene within said image by:

identifying a first gradient parallel to a width direction of said image;

identifying a second gradient perpendicular to said width direction of said image; and

comparing said first gradient and said second gradient, said horizontal direction of said scene being identified by the smaller of said first gradient and said second gradient.

17. An image identification system comprising:

a color classifier classifying potential sky pixels in said image by color;

an extractor identifying connected components of said potential sky pixels;

a first comparator eliminating ones of said connected components that have a texture above a predetermined texture threshold;

a logic unit computing desaturation gradients of said connected components; and

a second comparator comparing said desaturation gradients of said connected components with a predetermined desaturation gradient for sky to identify true sky regions in said image.

18. The image identification system in claim 17, wherein said desaturation gradients comprise desaturation gradients for red, green and blue trace components of said image.

19. The image identification system in claim 17, wherein said predetermined desaturation gradient for sky comprises, from horizon to zenith, a decrease in red and green light trace components and a substantial constant blue light trace component.

20. The image identification system in claim 17, wherein said color classifier includes:

a pixel classifier forming a belief map of pixels in said image;

a second logic unit computing an adaptive threshold of sky color; and

a third comparator classifying ones of said pixels that exceed said threshold as said potential sky pixels.

21. The image identification system in claim 20, wherein said second logic unit identifies a first valley on a belief histogram derived from said belief map to computer said adaptive threshold.

22. The image identification system in claim 21, wherein said belief map and said belief histogram are unique to said image.

23. The image identification system in claim 17, further comprising:

a horizontal identification unit that identifies a first gradient parallel to a width direction of said image and a second gradient perpendicular to said width direction of said image,

a third comparator comparing said first gradient and said second gradient, said horizontal direction of said comparator identifying said scene by identifying the smaller of said first gradient and said second gradient.

24. A method of using and a computer program running on a computer system to detect sky regions in an image, said method comprising:

using said computer program to classify potential sky pixels in said image by color;

using said computer program to extract connected components of said potential sky pixels;

using said computer program to eliminate ones of said connect components that have a texture above a predetermined texture threshold;

using said computer program to computer desaturation gradients of said connected components; and

using said computer program to compare said desaturation gradients of said connected components with a predetermined desaturation gradient for sky to identify true sky regions in said image.

25. The method in claim 24, wherein said desaturation gradients comprise desaturation gradients for red, green and blue trace components of said image.

26. The method in claim 24, wherein said predetermined desaturation gradient for sky comprises, from horizon to zenith, a decease in red and green light trace components and a substantially constant blue light trace component.

27. The method in claim 24, wherein said using of said computer program to classify comprises:

using said computer program to form a belief map of pixels in said image using a pixel classifier;

using said computer program to compute an adaptive threshold of sky color; and

using said computer program to classify ones of said pixels that exceed said threshold as said potential sky pixels.

28. The method in claim 27, wherein said using of said computer program to computer said adaptive threshold comprises using said computer program to identify a first valley in a belief histogram derived from said belief map.

29. The method in claim 28, wherein said belief map and said belief histogram are unique to said image.

30. The method in claim 24, further comprising using said computer program to determine a horizontal direction of a scene within said image by:

using said computer program to identify a first gradient parallel to a width direction of said image;

using said computer program to identify a second gradient perpendicular to said width direction of said image; and

using said computer program to comparing said first gradient and said second gradient, said horizontal direction of said scene being identified by the smaller of said first gradient and said second gradient.