METHOD AND APPARATUS FOR OBJECT DETECTION

• US 20140112535A1
• Filed: 10/18/2013
• Published: 04/24/2014
• Est. Priority Date: 10/19/2012
• Status: Active Grant

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: XU, LIFENG, CHEN, ZENG, ZHU, FUGUO
• Granted Patent: US 9,135,506 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06V 40/11 Hand-related biometrics; Hand pose recognition
• CPC Class Codes: G06V 40/19 Sensors therefor

Abstract

A method for detecting a plurality of object regions in an image, wherein the plurality of object regions having similar specific structural features, comprises: an estimation step for estimating a common initial value for the specific structural features of the plurality of object regions; and a determination step for determining, for each of the plurality of object regions, a final value for the specific structural feature of the object region and a final position thereof separately based on the estimated common initial value.

Claims

1. A method for detecting a plurality of object regions in an image, the plurality of object regions having similar specific structural features, the method comprising:

an estimation step for estimating a common initial value for the specific structural features of the plurality of object regions; and

a determination step for determining, for each of the plurality of object regions, a final value for the specific structural feature of the object region and a final position thereof separately based on the estimated common initial value.

2. The method according to claim 1, wherein the estimation step comprises:

setting an estimation range for the common initial value of the specific structural features of the plurality of object regions;

for each value in the estimation range, calculating a confidence score of each of the plurality of object regions, and summing the calculated confidence scores of respective object regions in the plurality of object regions; and

selecting the value in the estimation range corresponding to maximum sum of the calculated confidence scores of respective object regions in the plurality of object regions as the common initial value.

3. The method according to claim 1, wherein the determination step comprises:

setting a determination range for the final value of the specific structural features based on the common initial value;

for each of the plurality of object regions, calculating a confidence score for each value in the determination range for the object region; and

for each of the plurality of object regions, selecting a value in the determination range which makes the confidence score of the object region maximum as the final value of the specific structural feature of the object region, and designating a position in the object region where the confidence score of the object region is maximum as the final position of the specific structural feature of the object region.

4. The method according to claim 1, wherein the specific structural feature of an object region is a structural feature defining the shape of the object region.

5. The method according to claim 2, wherein, calculating a confidence score of an object region for a value of a specific structural feature comprising:

at each position in the object region, calculating a confidence score of the object region for the value of the specific structural feature; and

selecting a maximum among the confidence scores of the object region at all positions in the object region as the confidence score of the object region for the value of the specific structural feature.

6. The method according to claim 1, further comprising:

an object region pre-processing step, for each of the plurality of object regions, pre-processing the object region to obtain a partial region in which the final position likely appears in the object region, and

wherein the estimation step and determination step are performed on the partial region.

7. A method for detecting two eye pupil regions in a face image, the two eye pupil regions having similar radii, the method comprising:

an estimation step for estimating a common initial value for the radii of the two eye pupil regions; and

a determination step for determining, for each of the two eye pupil regions, a final value for the radius of the eye pupil region and a final center of the eye pupil region separately based on the estimated common initial value.

8. The method according to claim 7, wherein the estimation step comprises:

setting an estimation range for the common initial value of the radii of the two eye pupil regions;

for each value in the estimation range, calculating a confidence score of each of the two eye pupil regions, and summing the calculated confidence scores of respective eye pupil region in the two eye pupil regions; and

selecting the value in the estimation range corresponding to maximum sum of the calculated confidence scores of respective eye pupil region in the two eye pupil regions as the common initial value.

9. The method according to claim 7, wherein the determination step comprises:

setting a determination range for the final value of the radii based on the common initial value;

for each of the two eye pupil regions, calculating a confidence score for each value in the determination range for the eye pupil region; and

for each of the two eye pupil regions, selecting a value in the determination range which makes the confidence score of the eye pupil region maximum as the final value of the radius of the eye pupil region, and designating a position in the eye pupil region where the confidence score of the eye pupil region is maximum as the final center of the eye pupil region.

10. The method according to claim 8, wherein, calculating a confidence score of an eye pupil region for a value of a radius comprising:

at each position in the eye pupil region, calculating a confidence score of the eye pupil region for the value of the radius; and

selecting a maximum among the confidence scores of the eye pupil region at all positions in the eye pupil region as the confidence score of the eye pupil region for the value of the radius.

11. The method according to claim 7, further comprising:

a pre-processing step, for each of the two eye pupil regions, pre-processing the eye pupil region to obtain a partial region in which the final center of the eye pupil region likely appears in the eye pupil region, and

wherein the estimation step and determination step are performed on the partial region.

12. The method according to claim 11, wherein the pre-processing step further comprises:

for each of the two eye pupil regions,

scanning the eye pupil region by means of a rectangular sliding window with a predetermined size, and

designating an area under the rectangular sliding window in the eye pupil region with a lowest grey intensity as the partial region.

13. The method according to claim 12, wherein the predetermined size is related to the size of the eye pupil region.

14. The method according to claim 8, wherein the estimation range is determined based on the distance between the two eye pupil regions.

15. An apparatus for detecting a plurality of object regions in an image, the plurality of object regions having similar specific structural features, the apparatus comprising:

an estimation unit configured to estimate a common initial value for the specific structural features of the plurality of object regions; and

a determination unit configured to determine, for each of the plurality of object regions, a final value for the specific structural feature of the object region and a final position thereof separately based on the estimated common initial value.

16. The apparatus according to claim 15, wherein the estimation unit comprises:

a unit configured to set an estimation range for the common initial value of the specific structural features of the plurality of object regions;

a unit configured to, for each value in the estimation range, calculate a confidence score of each of the plurality of object regions, and sum the calculated confidence scores of respective object regions in the plurality of object regions; and

a unit configured to select the value in the estimation range corresponding to maximum sum of the calculated confidence scores of respective object regions in the plurality of object regions as the common initial value.

17. The apparatus according to claim 15, wherein the determination unit comprises:

a unit configured to set a determination range for the final value of the specific structural features based on the common initial value;

a unit configured to, for each of the plurality of object regions, calculate a confidence score for each value in the determination range for the object region; and

a unit configured to, for each of the plurality of object regions, select a value in the determination range which makes the confidence score of the object region maximum as the final value of the specific structural feature of the object region, and designate a position in the object region where the confidence score of the object region is maximum as the final position of the specific structural feature of the object region.

18. The apparatus according to claim 15, wherein the specific structural feature of an object region is a structural feature defining the shape of the object region.

19. The apparatus according to claim 16, wherein, the unit configured to calculate a confidence score of an object region for a value of a specific structural feature comprise:

a unit configured to, at each position in the object region, calculate a confidence score of the object region for the value of the specific structural feature; and

a unit configured to select a maximum among the confidence scores of the object region at all positions in the object region as the confidence score of the object region for the value of the specific structural feature.

20. The apparatus according to claim 15, further comprising:

an object region pre-processing unit configured to, for each of the plurality of object regions, pre-process the object region to obtain a partial region in which the final position likely appears in the object region,

wherein the estimation unit and determination unit operate on the partial region.

21. An apparatus for detecting two eye pupil regions in a face image, the two eye pupil regions having similar radii, the apparatus comprising:

an estimation unit configured to estimate a common initial value for the radii of the two eye pupil regions; and

a determination unit configured to determine, for each of the two eye pupil regions, a final value for the radius of the eye pupil region and a final center of the eye pupil region separately based on the estimated common initial value.

22. The apparatus according to claim 21, wherein the estimation unit comprises:

a unit configured to set an estimation range for the common initial value of the radii of the two eye pupil regions;

a unit configured to, for each value in the estimation range, calculate a confidence score of each of the two eye pupil regions, and sum the calculated confidence scores of respective eye pupil region in the two eye pupil regions; and

a unit configured to select the value in the estimation range corresponding to maximum sum of the calculated confidence scores of respective eye pupil region in the two eye pupil regions as the common initial value.

23. The apparatus according to claim 21, wherein the determination unit comprises:

a unit configured to set a determination range for the final value of the radii based on the common initial value;

a unit configured to, for each of the two eye pupil regions, calculate a confidence score for each value in the determination range for the eye pupil region; and

a unit configured to, for each of the two eye pupil regions, select a value in the determination range which makes the confidence score of the eye pupil region maximum as the final value of the radius of the eye pupil region, and designate a position in the eye pupil region where the confidence score of the eye pupil region is maximum as the final center of the eye pupil region.

24. The apparatus according to claim 22, wherein, the unit configured to calculate a confidence score of an eye pupil region for a value of a radius comprises:

a unit configured to, at each position in the eye pupil region, calculate a confidence score of the eye pupil region for the value of the radius; and

a unit configured to select a maximum among the confidence scores of the eye pupil region at all positions in the eye pupil region as the confidence score of the eye pupil region for the value of the radius.

25. The apparatus according to claim 21, further comprising;

a pre-processing unit configured to, for each of the two eye pupil regions, pre-process the eye pupil region to obtain a partial region in which the final center of the eye pupil region likely appears in the eye pupil region,

wherein the estimation unit and determination unit operate on the partial region.

26. The apparatus according to claim 25, wherein the pre-processing unit further comprises a unit configured to:

for each of the two eye pupil regions,

scan the eye pupil region by means of a rectangular sliding window with a predetermined size, and

designate an area under the rectangular sliding window in the eye pupil region with a lowest grey intensity as the partial region.

27. The apparatus according to claim 26, wherein the predetermined size is related to the size of the eye pupil region.

28. The apparatus according to claim 22, wherein the estimation range is determined based on the distance between the two eye pupil regions.