Method and device for face in-vivo detection

• US 9,824,280 B2
• Filed: 01/25/2016
• Issued: 11/21/2017
• Est. Priority Date: 01/26/2015
• Status: Active Grant

• Current Assignee: Advanced New Technologies Company Limited (Ant Group Co., Ltd.)
• Original Assignee: Alibaba Group Holding Ltd.
• Inventors: Li, Peng
• Primary Examiner(s): Tran, Phuoc
• Application Number:
  US15/005,427
• Publication Number:
  US 20160217338A1
• Time in Patent Office: 666 Days
• CPC Class Codes: G06V 20/40 in video content extracting overlay text G06V20/62; video retrieval G06F16/70; processing of video elementary streams in video servers H04N21/234; processing of video elementary streams in video clients H04N21/44
• CPC Class Codes: G06V 20/49 Segmenting video sequences, i.e. computational techniques such as parsing or cutting the sequence, low-level clustering or determining units such as shots or scenes
• CPC Class Codes: G06V 40/167 using comparisons between temporally consecutive images
• CPC Class Codes: G06V 40/40 Spoof detection, e.g. liveness detection

Abstract

A face detection method includes acquiring a video image sequence, performing a video shot boundary detection process on the video image sequence to determine whether a shot change exists in the video image sequence to obtain a first judgment result, and determining, when the first judgment result indicates that a shot change exists in the video image sequence, that face detection has failed. The present disclosure also provides a face detection device, which includes an acquisition unit configured to acquire a video image sequence, a first detection unit configured to perform a video shot boundary detection process on the video image sequence to determine whether a shot change exists in the video image sequence to obtain a first judgment result, and a determination unit configured to determine, when the first judgment result indicates that a shot change exists in the video image sequence, that face detection has failed.

Claims

1. A face detection method, comprising:

acquiring a video image sequence;

performing a video shot boundary detection process on the video image sequence to determine whether a shot change exists in the video image sequence to obtain a first judgment result, the video shot boundary detection process comprising;

performing an image feature extraction on the video image sequence in an image feature extraction region to obtain image feature vectors of two image frames; and

determining whether the two image frames are discontinuous in content, based on the image feature vectors; and

determining, when the first judgment result indicates that a shot change exists in the video image sequence, that face detection has failed.

2. The method according to claim 1, wherein the video shot boundary detection process further comprises:

calculating, with the image feature vectors of the two image frames, to obtain a similarity value between the two image frames to determine content continuity of the two image frames; and

determining, when it is determined that the two image frames are discontinuous in content according to the similarity value, that a shot change exists in the video image sequence.

3. The method according to claim 2, wherein the image feature extraction region comprises any one of:

a single pixel feature extraction region, a pixel block feature extraction region, or a complete frame image feature extraction region.

4. The method according to claim 2, wherein the performing an image feature extraction on the video image sequence to obtain image feature vectors comprises at least one of:

obtaining, in a selected color region, pixel gray level statistics or pixel color statistics of image frames of the video image sequence, and constituting the image feature vectors with the obtained pixel gray level statistics or pixel color statistics;

obtaining a gray level histogram or color histogram of image frames in the video image sequence;

obtaining image edge features of image frames in the video image sequence; and

transforming original images into other feature regions by means of a compressed domain to obtain corresponding image feature vectors.

5. The method according to claim 2, wherein the two image frames are the i^th image frame and the j^th image frame, where j=i+n(n≧

1), i≧

1.

6. The method according to claim 2, wherein the video shot boundary detection process comprises a video shot cut detection process.

7. The method according to claim 6, wherein, the determining, when it is determined that the two image frames are discontinuous in content according to the similarity value between the two image frames, that a shot change exists in the video image sequence comprises:

determining, when the similarity value between the two image frames is smaller than a first threshold value, that a cut change exists between the two image frames and a shot change exists in the video image sequence.

8. The method according to claim 6, wherein, the determining, when it is determined that the two image frames are discontinuous in content according to the similarity value between the two image frames, that a shot change exists in the video image sequence comprises:

judging the similarity value between the two image frames based on a neighbor algorithm k-Nearest Neighbor (KNN) classifier, a decision tree, or a support vector machine (SVM), and determining, when the judgment result indicates that a cut change exists between the two frames of images, that a shot change exists in the video image sequence.

9. The method according to claim 2, wherein the video shot boundary detection process comprises a video shot gradual change detection process, and wherein, the determining, when it is determined that the two image frames are discontinuous in content according to the similarity value between the two frames of images, that a shot change exists in the video image sequence comprises:

obtaining, when it is determined that the similarity value between the two frames of images is greater than a third threshold value and is smaller than a fourth threshold value, subsequent image frames of the two image frames;

calculating a similarity value between pairwise image frames of the subsequent image frames until the similarity value between the pairwise image frames is greater than the fourth threshold value; and

calculating a sum of the similarity values of the obtained image frames, and determining, when the sum of the similarity values is smaller than a fifth threshold value, that a gradual change exists in the video image sequence and a shot change exists in the video image sequence.

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

carrying out the face detection based on a face feature point detection process to obtain a second judgment result; and

determining, when the second judgment result indicates that face detection conditions are satisfied and the first judgment result shows that no shot change exists in the video image sequence, that the face detection has passed.

11. The method according to claim 1, wherein the video shot boundary detection process further comprises:

calculating, with the image feature vectors of the two image frames, to obtain a frame difference value between the two image frames to determine content continuity of the two image frames; and

determining, when it is determined that the two image frames are discontinuous in content according to the calculated frame difference value between the two image frames, that a shot change exists in the video image sequence.

12. The method according to claim 11, wherein the image feature extraction region comprises any one of:

a single pixel feature extraction region, a pixel block feature extraction region, or a complete frame image feature extraction region.

13. The method according to claim 11, wherein the performing an image feature extraction on the video image sequence to obtain image feature vectors comprises at least one of:

obtaining, in a selected color region, pixel gray level statistics or pixel color statistics of image frames in the video image sequence, and constituting the image feature vectors with the obtained pixel gray level statistics or pixel color statistics;

obtaining a gray level histogram or color histogram of image frames in the video image sequence;

obtaining image edge features of image frames in the video image sequence; and

transforming original images into other feature regions by means of a compressed domain to obtain corresponding image feature vectors.

14. The method according to claim 11, wherein the video shot boundary detection process comprises a video shot cut detection process.

15. The method according to claim 14, wherein, the determining, when it is determined that the two image frames are discontinuous in content according to the frame difference value between the two frames of images, that a shot change exists in the video image sequence comprises:

determining, when the frame difference value between the two image frames is greater than a second threshold value, that a cut change exists between the two image frames and a shot change exists in the video image sequence.

16. The method according to claim 14, wherein, the determining, when it is determined that the two image frames are discontinuous in content according to the similarity value between the two image frames, that a shot change exists in the video image sequence comprises:

judging the frame difference value between the two image frames based on a neighbor algorithm k-Nearest Neighbor (KNN) classifier, a decision tree or a support vector machine (SVM), and determining, when the judgment result indicates that a cut change exists between the two frames of images, that a shot change exists in the video image sequence.

17. The method according to claim 11, wherein the video shot boundary detection process comprises a video shot gradual change detection process, and wherein, the determining, when it is determined that the two image frames are discontinuous in content according to the frame difference value between the two image frames, that a shot change exists in the video image sequence comprises:

obtaining, when it is determined that the frame difference value between the two image frames is greater than a six threshold value and is smaller than a seven threshold value, subsequent image frames of the two image frames;

calculating the frame difference value between pairwise image frames of the subsequent image frames until the frame difference value between the pairwise image frames is smaller than the six threshold value; and

calculating a sum of the frame difference values, and determining, when the sum is greater than the seventh threshold value, that a gradual change exists in the video image sequence and a shot change exists in the video image sequence.

18. A face detection device, comprising:

an acquisition unit configured to acquire a video image sequence;

a first detection unit configured to perform a video shot boundary detection process on the video image sequence to determine whether a shot change exists in the video image sequence to obtain a first judgment result, the video shot boundary detection process comprising;

performing an image feature extraction on the video image sequence in an image feature extraction region to obtain image feature vectors of two image frames; and

determining whether the two image frames are discontinuous in content, based on the image feature vectors; and

a determination unit configured to determine, when the first judgment result indicates that a shot change exists in the video image sequence, that face detection has failed.

19. The device according to claim 18, wherein the first detection unit is further configured to:

calculate a similarity or frame difference value based on the image feature vectors of the two image frames to determine content continuity of the two image frames; and

determine, when it is determined that the two image frames are discontinuous in content according to the similarity or frame difference value, that a shot change exists in the video image sequence.

20. The device according to claim 19, wherein the image feature extraction region comprises any one of:

a single pixel feature extraction region, a pixel block feature extraction region, or a complete frame image feature extraction region.