Foreground subject detection

• US 8,509,545 B2
• Filed: 11/29/2011
• Issued: 08/13/2013
• Est. Priority Date: 11/29/2011
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Dedhia, Bhaven, Leyvand, Tommer
• Primary Examiner(s): Do, Anh
• Application Number:
  US13/306,876
• Publication Number:
  US 20130136358A1
• Time in Patent Office: 623 Days
• Field of Search: 382/103, 382/173, 382/118, 382/191, 382/181, 382/224, 345/156, 345/158, 345/157, 348/580, 348/218.1, 348/208.13, 359/463, 273/348.1, 273/409
• US Class Current: 382/224
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/10024 Color image
• CPC Class Codes: G06T 2207/10028 Range image; Depth image; 3D point clouds
• CPC Class Codes: G06T 2207/20072 Graph-based image processing
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/162 involving graph-based methods
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation

Abstract

Classifying pixels in a digital image includes receiving a primary image from one or more image sensors. The primary image includes a plurality of primary pixels. A depth image from one or more depth sensors is also received. The depth image includes a plurality of depth pixels, each depth pixel registered to one or more primary pixels. The depth image and the primary image are cooperatively used to identify whether a primary pixel images a foreground subject or a background subject.

Claims

1. A method of classifying pixels in a digital image, the method comprising:

receiving a primary image from one or more image sensors, the primary image including a plurality of primary pixels;

receiving a depth image from one or more depth sensors, the depth image including a plurality of depth pixels, each depth pixel registered to one or more primary pixels; and

cooperatively using the depth image and the primary image to identify whether a primary pixel images a foreground subject or a background subject.

2. A method of classifying pixels in a digital image, the method comprising:

receiving a primary image from one or more image sensors, the primary image including a plurality of primary pixels;

receiving a depth image from one or more depth sensors, the depth image including a plurality of depth pixels, each depth pixel registered to one or more primary pixels;

producing a segmentation map from the depth image, the segmentation map including a plurality of estimated foreground pixels and a plurality of estimated background pixels;

producing a trimap from the segmentation map, the trimap matching the segmentation map but including a plurality of test pixels at one or more boundaries between estimated foreground pixels and estimated background pixels; and

refining each of the test pixels to either a confirmed foreground pixel or a confirmed background pixel based on an analysis of primary pixels registered to test pixels.

3. The method of claim 2, further comprising refining each of the estimated foreground pixels not identified as test pixels to confirmed foreground pixels, and refining each of the estimated background pixels not identified as test pixels to confirmed background pixels.

4. The method of claim 3, where the primary pixels registered to the confirmed foreground pixels are useable without the primary pixels registered to the confirmed background pixels.

5. The method of claim 3, further comprising producing a segmented image including the primary pixels registered to the confirmed foreground pixels but without the primary pixels registered to the confirmed background pixels.

6. The method of claim 5, where pixels of the segmented image registered to the confirmed background pixels are transparent pixels.

7. The method of claim 5, where pixels of the segmented image registered to the confirmed background pixels are from another image.

8. The method of claim 5, where one or more pixels of the segmented image registered to the confirmed foreground pixels are smoothed.

9. The method of claim 8, where the one or more pixels of the segmented image registered to the confirmed foreground pixels are smoothed via anti-aliasing.

10. The method of claim 8, where the one or more pixels of the segmented image registered to the confirmed foreground pixels are smoothed via alpha-blending.

11. The method of claim 8, where the one or more pixels of the segmented image registered to the confirmed foreground pixels include boundary pixels.

12. The method of claim 2, where refining each of the test pixels comprises executing a graph-cut algorithm.

13. The method of claim 2, where refining each of the test pixels comprises utilizing temporal feedback.

14. The method of claim 2, where the image sensor is a visible light sensor.

15. The method of claim 2, where the image sensor is an infrared light sensor.

16. A method of classifying pixels in a digital image, the method comprising:

receiving a primary image from one or more image sensors, the primary image including a plurality of primary pixels;

receiving a depth image from one or more depth sensors, the depth image including a plurality of depth pixels, each depth pixel registered to one or more primary pixels;

recognizing a plurality of estimated background pixels derived from the depth image, each estimated background pixel registered to one or more primary pixels and identifying subjects imaged by those one or more primary pixels as background subjects;

recognizing a plurality of estimated foreground pixels derived from the depth image, each estimated foreground pixel registered to one or more primary pixels and identifying subjects imaged by those one or more primary pixels as foreground subjects;

identifying estimated foreground pixels near estimated background pixels and/or estimated background pixels near estimated foreground pixels as test pixels; and

refining each of the test pixels to either a confirmed foreground pixel or a confirmed background pixel based on a graph cut analysis of primary pixels registered to test pixels.

17. The method of claim 16, further comprising refining each of the estimated foreground pixels not identified as test pixels to confirmed foreground pixels, and refining each of the estimated background pixels not identified as test pixels to confirmed background pixels.

18. The method of claim 17, where the primary pixels registered to the confirmed foreground pixels are useable without the primary pixels registered to the confirmed background pixels.

19. The method of claim 17, further comprising producing a segmented image including the primary pixels registered to the confirmed foreground pixels but without the primary pixels registered to the confirmed background pixels.

20. The method of claim 16, where refining each of the test pixels comprises utilizing temporal feedback.