Parallelizing cascaded face detection

• US 8,649,612 B1
• Filed: 01/25/2010
• Issued: 02/11/2014
• Est. Priority Date: 01/06/2010
• Status: Active Grant

• Current Assignee: Apple Inc.
• Original Assignee: Apple Inc.
• Inventors: Brunner, Ralph
• Primary Examiner(s): CHAWAN, SHEELA C
• Application Number:
  US12/693,204
• Time in Patent Office: 1,478 Days
• Field of Search: 382/103, 382/115, 382/224, 382/118, 382/254, 382/274, 382/154, 382/203, 382/276, 382/172, 382/304, 382/218, 382/283, 382/284, 348/16, 348/152, 348/169, 348/207.99, 348/68, 348 77- 78, 348/143, 348/155, 348/158, 348/220.1, 348/229.1, 345/419, 345/426, 345/473, 703/2
• US Class Current: 382/224
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data
• CPC Class Codes: G06V 10/809 of classification results, e.g. where the classifiers operate on the same input data
• CPC Class Codes: G06V 40/162 using pixel segmentation or colour matching

Abstract

One or more portions of a cascaded face detection process, such as a process using a number of stages in which each stage includes classifiers having one or more associated features, can be parallelized. For instance, multiple pixels associated with an image can be evaluated in parallel, including simultaneously evaluating one or more features for each of the selected pixels. As a result, the number of operations required to perform the face detection processing can be reduced. Further, pre-processing can be performed to identify one or more image regions for which face detection process need not be performed. For instance, either or both of a skin-tone measure and a variance measure can be used to identify image regions in which a face is unlikely to be depicted.

Claims

1. A computer-implemented method of performing face detection using parallel analysis, the method comprising:

selecting, using a processor, a plurality of locations within an image;

determining, for each selected location in parallel, a classifier score for each of a plurality of classifiers associated with an analysis stage;

combining, for each selected location, the classifier scores for the plurality of classifiers to generate a total classifier score associated with the analysis stage;

comparing, for each selected location, the total classifier score to a threshold; and

generating a classifier mask based on the comparison of the total classifier score and the threshold for each of the selected locations.

2. The computer-implemented method of claim 1, wherein the image comprises an integral image.

3. The computer-implemented method of claim 1, wherein at least one of the plurality of classifiers includes two or more features.

4. The computer-implemented method of claim 1, further comprising:

providing the classifier mask as an input to a subsequent analysis stage.

5. A computer-readable storage device comprising non-transitory storage medium encoding instructions that, when executed by a processing apparatus, cause the processing apparatus to perform operations comprising:

selecting a plurality of locations within an image;

determining, for each selected location in parallel, a classifier score for each of a plurality of classifiers associated with an analysis stage;

combining, for each selected location, the classifier scores for the plurality of classifiers to generate a total classifier score associated with the analysis stage;

comparing, for each selected location, the total classifier score to a threshold; and

generating a classifier mask based on the comparison of the total classifier score and the threshold for each of the selected locations.

6. The computer-readable storage device of claim 5, further comprising instructions that, when executed by a processing apparatus, cause the processing apparatus to perform operations comprising:

foregoing evaluation of a particular location in a subsequent analysis stage when the classifier mask includes a first value corresponding to the particular location.

7. The computer-readable storage device of claim 5, further comprising instructions that, when executed by a processing apparatus, cause the processing apparatus to perform operations comprising:

determining that the classifier mask includes at least one second value; and

providing the classifier mask as an input to a subsequent analysis stage.

8. The computer-readable storage device of claim 7, further comprising instructions that, when executed by a processing apparatus, cause the processing apparatus to perform operations comprising:

initiating the subsequent analysis stage in response to the provided classifier mask;

generating, for each location having a corresponding value in the provided classifier mask equal to the second value, a second total classifier score for a plurality of classifiers associated with the subsequent analysis stage;

comparing the generated second total classifier score for each of the corresponding locations to a second threshold associated with the subsequent analysis stage; and

modifying the provided classifier mask for locations having a corresponding value in the provided classifier mask equal to the second value based on the comparison of the generated second total classifier score to the second threshold.

9. The computer-readable storage device of claim 5, further comprising instructions that, when executed by a processing apparatus, cause the processing apparatus to perform operations comprising:

selecting the plurality of locations such that neighboring locations are separated by at least one intervening pixel location.

10. The computer-readable storage device of claim 5, wherein the threshold is determined through training.

11. The computer-readable storage device of claim 5, wherein each of the plurality of locations corresponds to a pixel of the image.

12. The computer-readable storage device of claim 5, wherein each of the plurality of locations represents an origin of a corresponding two-dimensional region.

13. The computer-readable storage device of claim 12, wherein each of the plurality of two-dimensional regions is at least partially overlapping with at least one other two-dimensional region.

14. A system comprising:

a storage device tangibly storing an image; and

processor electronics configured to perform operations comprising;

selecting a plurality of locations within an image;

determining, for each selected location in parallel, a classifier score for each of a plurality of classifiers associated with an analysis stage;

combining, for each selected location, the classifier scores for the plurality of classifiers to generate a total classifier score associated with the analysis stage;

comparing, for each selected location, the total classifier score to a threshold; and

generating a classifier mask based on the comparison of the total classifier score and the threshold for each of the selected locations.

15. The system of claim 14, wherein the image comprises an integral image.

16. The system of claim 14, wherein each of the plurality of classifiers associated with the analysis stage includes at least two features.

17. The system of claim 14, wherein the processor electronics are further configured to perform operations comprising:

determining that the classifier mask includes at least one first value; and

providing the classifier mask as an input to a subsequent analysis stage.

18. The system of claim 14, wherein the processor electronics are further configured to perform operations comprising:

selecting the plurality of locations such that each of the locations is aligned horizontally and neighboring locations are separated by at least one intervening pixel location.