Blob representation in video processing

• US 8,498,444 B2
• Filed: 12/13/2010
• Issued: 07/30/2013
• Est. Priority Date: 12/13/2010
• Status: Active Grant

• Current Assignee: Texas Instruments, Inc.
• Original Assignee: Texas Instruments, Inc.
• Inventors: Sharma, Vinay
• Primary Examiner(s): Lu, Tom Y
• Application Number:
  US12/966,798
• Publication Number:
  US 20120148093A1
• Time in Patent Office: 960 Days
• Field of Search: 382/103
• US Class Current: 382/103
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/20041 Distance transform
• CPC Class Codes: G06T 2207/30232 Surveillance
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation
• CPC Class Codes: G06T 7/251 involving models
• CPC Class Codes: G06T 7/62 of area, perimeter, diameter or volume
• CPC Class Codes: G06T 7/66 of image moments or centre of gravity
• CPC Class Codes: G06V 20/52 Surveillance or monitoring of activities, e.g. for recognising suspicious objects recognising microscopic objects G06V20/69

Abstract

A method of processing a video sequence is provided that includes receiving a frame of the video sequence, identifying a plurality of blobs in the frame, computing at least one interior point of each blob of the plurality of blobs, and using the interior points in further processing of the video sequence. The interior points may be used, for example, in object tracking.

Claims

1. A method of processing a video sequence, the method comprising:

receiving a frame of the video sequence;

identifying a plurality of blobs in the frame;

computing at least one interior point of each blob of the plurality of blobs, wherein computing at least one interior point comprises applying a distance transform to each foreground pixel in a blob to compute a distance value for the foreground pixel, wherein the distance value is a shortest distance from the foreground pixel to a background pixel, and selecting a foreground pixel in the blob as the at least one interior point based on a comparison of the distance value of the foreground pixel to a threshold; and

using the interior points in further processing of the video sequence.

2. The method of claim 1, further comprising:

applying non-maximum suppression to the pixel distance values to determine local maxima prior to selecting a foreground pixel.

3. The method of claim 1, wherein using the interior points comprises tracking objects using the interior points.

4. The method of claim 3, wherein tracking objects comprises identifying a blob in the plurality of blobs as corresponding to an object in a previous frame based on proximity between a location of the object and an interior point of the blob.

5. The method of claim 4, further comprising:

computing a centroid for each blob of the plurality of blobs; and

wherein tracking objects further comprises;

selecting a candidate blob in the plurality of blobs as corresponding to the object based on distance between the location of the object and the centroid of the blob;

when the distance is acceptable, identifying the candidate blob as corresponding the object; and

when the distance is not acceptable, identifying a blob in the plurality of blobs having an interior point closest to the location of the object as corresponding to the object.

6. The method of claim 5, wherein tracking objects further comprises selecting the candidate blob based on percentage size difference between the object and a bounding box of the blob.

7. A digital system comprising:

a video input component configured to provide a video sequence; and

an object detection component configured to

receive a frame of the video sequence;

identify a plurality of blobs in the frame;

compute at least one interior point of each blob of the plurality of blobs and to compute at least one interior point by applying a distance transform to each foreground pixel in a blob to compute a distance value for the foreground pixel, wherein the distance value is a shortest distance from the foreground pixel to a background pixel and selecting a foreground pixel in the blob as the at least one interior point based on a comparison of the distance value of the foreground pixel to a threshold; and

provide the interior points for use in further processing of the video sequence.

8. The digital system of claim 7, wherein the object detection component is further configured to compute at least one interior point by:

applying non-maximum suppression to the pixel distance values to determine local maxima prior to selecting a foreground pixel.

9. The digital system of claim 7, further comprising an object tracking component configured to track objects using the interior points.

10. The digital system of claim 9, wherein the object tracking component is configured to identify a blob in the plurality of blobs as corresponding to an object in a previous frame based on proximity between a location of the object and an interior point of the blob.

11. The digital system of claim 10,

wherein the object detection component is further configured to compute a centroid for each blob of the plurality of blobs, and

wherein the object tracing component is further configured to:

select a candidate blob in the plurality of blobs as corresponding to the object based on distance between the location of the object and the centroid of the blob;

when the distance is acceptable, to identify the candidate blob as corresponding the object; and

when the distance is not acceptable, to identify a blob in the plurality of blobs having an interior point closest to the location of the object as corresponding to the object.

12. The digital system of claim 11, wherein the object tracking component is further configured to select the candidate blob based on percentage size difference between the object and a bounding box of the blob.

13. A non-transitory computer readable medium comprising executable instructions to cause a digital system to perform a method processing a video sequence, the method comprising:

receiving a frame of the video sequence;

identifying a plurality of blobs in the frame;

computing at least one interior point of the blob, wherein computing at least one interior point comprises applying a distance transform to each foreground pixel in a blob to compute a distance value for the foreground pixel, wherein the distance value is a shortest distance from the foreground pixel to a background pixel and selecting a foreground pixel in the blob as the at least one interior point based on a comparison of the distance value of the foreground pixel to a threshold; and

using the interior points in further processing of the video sequence.

14. The computer readable medium of claim 13, wherein the method further comprises:

applying non-maximum suppression to the pixel distance values to determine local maxima prior to selecting a foreground pixel.

15. The computer readable medium of claim 13, wherein using the interior points comprises tracking objects using the interior points.

16. The computer readable medium of claim 15, wherein tracking objects comprises identifying a blob in the plurality of blobs as corresponding to an object in a previous frame based on proximity between a location of the object and an interior point of the blob.

17. The computer readable medium of claim 16, wherein the method further comprises:

computing a centroid for each blob of the plurality of blobs; and

wherein tracking objects further comprises;

selecting a candidate blob in the plurality of blobs as corresponding to the object based on distance between the location of the object and the centroid of the blob;

when the distance is acceptable, identifying the candidate blob as corresponding the object; and

when the distance is not acceptable, identifying a blob in the plurality of blobs having an interior point closest to the location of the object as corresponding to the object.