Image background replacement method

• US 6,912,313 B2
• Filed: 05/31/2001
• Issued: 06/28/2005
• Est. Priority Date: 05/31/2001
• Status: Expired due to Term

• Current Assignee: Rakuten Group, Inc.
• Original Assignee: Sharp Laboratories of America Incorporated (Hon Hai Precision Industry Co., Ltd.)
• Inventors: Li, Baoxin
• Primary Examiner(s): Johns, Andrew W.
• Assistant Examiner(s): Alavi, Amir
• Application Number:
  US09/872,205
• Publication Number:
  US 20020186881A1
• Time in Patent Office: 1,489 Days
• Field of Search: 382/162, 382/164, 382/172, 382/218, 382/219, 382/224, 382/236, 382/260, 382/268, 382/272, 348/143, 348/154, 348/586, 345/619, 345/626, 345/628
• US Class Current: 382/219
• CPC Class Codes: G06T 11/00 2D [Two Dimensional] image generation
• CPC Class Codes: H04N 1/3871 the composed originals being of different kinds, e.g. low- and high-resolution originals

Abstract

An image'"'"'s pixels are initially segmented into pixels of the image foreground and background by comparing the pixels of the image to updated models of background reference pixels. Pixel structures are identified in the image. The probability that a pixel is in the background or foreground is refined by considering the initial segmentation of the pixel'"'"'s neighbors and the pixel'"'"'s membership in a pixel structure. If a pixel is identified as a background pixel it is replaced by a pixel of a new background.

Claims

1. A method of image processing comprising the steps of:

(a) determining a probability that an input pixel is one of a background and a foreground pixel of said image by comparing said input pixel to a background reference pixel model; and

(b) revising said probability for said input pixel according to a probability that a neighboring pixel is a pixel of one of said background and said foreground and a relationship of said input pixel and a pixel structure included in said image.

2. The method of claim 1 wherein the step of determining a probability that an input pixel is one of a background and a foreground pixel of said image by comparing said input pixel to a model of a background reference pixel comprises the steps of:

(a) comparing a difference between said input pixel and said background reference pixel model to a threshold difference;

(b) determining said probability that said input pixel is one of said background and said foreground pixels from a first probability function if said difference at least equals said threshold difference; and

(c) determining said probability that said input pixel is one of said background and said foreground pixels from a second probability function if said difference does not at least equal said threshold difference.

3. The method of claim 2 wherein said difference between said input pixel and said background reference pixel model comprises a difference in a chromatic value of said input pixel and said background reference pixel model.

4. The method of claim 2 wherein said threshold comprises a chromatic value.

5. The method of claim 1 wherein the step of revising said probability for said input pixel according to a probability that a neighboring pixel is a pixel of one of said background and said foreground and a relationship of said input pixel and a pixel structure included in said image comprises the steps of:

(a) filtering said image to identify a pixel structure included in said image;

(b) assigning a first probability that a pixel neighboring said input pixel is a pixel of one of said background and said foreground as said revised probability for said input pixel, if said input pixel is included in said structure; and

(c) if said input pixel is not included in said structure, assigning a second probability that a pixel neighboring said input pixel is a pixel of one of said background and said foreground as said revised probability for said input pixel.

6. The method of claim 1 wherein said background reference pixel model is a mean of a pixel of a background reference image and at least one input pixel of an image background.

7. The method of claim 1 further comprising the step of revising said background reference pixel model if said input pixel is a pixel of said background.

8. The method of claim 1 further comprising the step of substituting a new pixel for said input pixel in said image if said revised probability determines that said input pixel is more probably a pixel of said background.

9. A method of processing an image comprising the steps of:

(a) capturing an image comprising a plurality of input pixels representing a background and a foreground of a scene;

(b) capturing a background reference image comprising a plurality of background reference pixels representing a background of said scene;

(c) determining a probability that an input pixel is a pixel of one of said background and said foreground according to a relationship of said input pixel and a background reference pixel;

(d) identifying a pixel structure in said image;

(f) revising said probability for said input pixel according to a membership of said input pixel in said pixel structure and a probability that a neighboring pixel is a pixel of one of said background and said foreground; and

(g) substituting a new background pixel for a spatially corresponding input pixel classified as a background pixel.

10. The method claim 9 wherein the step of determining a probability that an input pixel is a pixel of one of said background and said foreground according to a relationship of said input pixel and a background reference pixel comprises the steps of:

(a) determining a first probability that said input pixel is a pixel of one of said background and said foreground if a relationship of said input pixel and said background reference pixel at least equals a threshold; and

(b) determining a second probability that said input pixel is a pixel of one of said background and said foreground if said relationship of said input pixel and said background reference pixel does not at least equal to said threshold.

11. The method of claim 10 wherein at least one of said first and said second probabilities is a function of a difference between said input pixel and said background reference pixel.

12. The method of claim 11 wherein said difference between said input pixel and said background reference pixel comprises a difference between a chromatic value of said input pixel and a chromatic value of said background reference pixel.

13. The method of claim 12 wherein at least one of said functions of said chromatic values of said input pixel and said background reference pixel comprises a sum of a plurality of chromatic components for said pixel.

14. The method of claim 10 wherein said relationship of said input pixel and said background reference pixel comprises a difference between a chromatic value of said input pixel and a chromatic value of said background reference pixel.

15. The method of claim 14 wherein at least one said chromatic values of said input pixel and said background reference pixel comprises a sum of a plurality of chromatic components for said pixel.

16. The method of claim 14 wherein said chromatic value of said background reference pixel comprises a mean of a plurality of chromatic values for a plurality of background pixels included in a plurality of images.

17. The method of claim 10 wherein said threshold comprises a function of a chromatic value for said background reference pixel.

18. The method of claim 17 wherein said function of said chromatic value of said background reference pixel comprises a function of a standard deviation of a chromatic value for a plurality of background pixels included in a plurality of images.

19. The method of claim 9 wherein the step of identifying a structure in said image comprises the step of filtering said image with a morphological filter to identify a plurality of connected input pixels in said image.

20. The method of claim 9 wherein the step of revising said probability for said input pixel according to a membership of said input pixel in said pixel structure and a probability that a neighboring pixel is a pixel of one of said background and said foreground comprises the steps of:

(a) revising one of said first and said second probabilities according to a first relationship if said input pixel is a pixel included in said structure; and

(b) revising one of said first and said second probabilities according to a second relationship if said input pixel is not a pixel included in said structure.

21. The method of claim 20 wherein at least one of said first and said second relationships comprises a probability that a pixel neighboring said input pixel is a pixel of one of said background and said foreground.

22. The method of claim 20 wherein at least one of said first and said second relationship comprises one of a maximum and a minimum of a probability that a pixel neighboring said input pixel is a pixel of one of said foreground and said background.

23. The method of claim 9 further comprising the step of revising a background reference pixel according to a relationship of said background reference pixel and an input pixel.

24. The method of claim 23 wherein the step of revising said background reference pixel according to a relationship of said background reference pixel and an input pixel comprises the step of determining a mean of a chromatic value of said background reference pixel and a chromatic value of said input pixel if said input pixel is classified as a pixel of said background.

25. A method of replacing an image background in a video sequence comprising the steps of:

(a) capturing an image comprising a plurality of input pixels representing a background and a foreground of a scene;

(b) capturing a background reference image comprising a plurality of background reference pixels representing a background of said scene;

(c) classifying an input pixel as a pixel of said background if a difference between said input pixel and a background reference pixel is less than a threshold difference;

(d) filtering said image to identify a pixel structure included in said image;

(e) revising said classification of said input pixel according to a first probability relationship if said input pixel is a pixel of said pixel structure;

(f) revising said classification of said input pixel according to a second probability relationship if said input pixel is not a pixel of said pixel structure;

(g) revising a background reference pixel if said revised classification of a spatially corresponding input pixel classifies said input pixel as a background pixel;

(h) substituting a new background pixel for said input pixel if said input pixel is classified as a pixel of said background; and

(i) capturing another image comprising a plurality of input pixels representing a background and a foreground of said scene.

26. The method of claim 25 wherein the step of classifying an input pixel as a pixel of said background if a difference between said input pixel and a background reference pixel is less than a threshold difference comprises the steps of:

(a) determining a difference between a chromatic value for said input pixel and a chromatic value for said background reference pixel; and

(b) comparing said difference to a threshold chromatic value.

27. The method of claim 26 wherein at least one of said chromatic values of said input pixel, said background reference pixel and said threshold comprises a sum of a plurality of chromatic components.

28. The method of claim 26 wherein said threshold is a function of a standard deviation of a chromatic value of said background reference pixel.

29. The method of claim 26 wherein said chromatic value of said background reference pixel comprises a mean of a chromatic value of a plurality of background pixels from a plurality images.

30. The method of claim 25 wherein the step of revising said classification of said input pixel according to a first probability relationship if said input pixel is a pixel of said pixel structure comprises the step of assigning to said input pixel a probability that one of a plurality of neighboring pixels is a background pixel if said input pixel is a pixel of said structure.

31. The method of claim 25 wherein the step of revising said classification of said input pixel according to a second probability relationship if said input pixel is a pixel of said pixel structure comprises the step of assigning to said input pixel a probability that one of a plurality of neighboring pixels is a background pixel if said input pixel is a pixel of said structure.

32. The method of claim 25 wherein the step of revising a background reference pixel if said revised classification of a spatially corresponding input pixel classifies said input pixel as a background pixel comprises the step of calculating the mean of a chromatic value of said background reference pixel and said spatially corresponding input pixel of said background.