Video signature based on image hashing and shot detection

1. A method of comparing a source video shot to a reference video shot, the method comprising:

electronically receiving a first group of frames corresponding to the source video shot;

selecting a sub-set of frames from the first group of frames as a first key frame set, wherein the sub-set of frames includes two or more different frames from the source video shot;

selecting a second key frame set from a second group of frames corresponding to a reference video shot;

identifying multiple pairs of counterpart frames for comparing the source video shot to the reference video shot, wherein each pair of counterpart frames includes one source frame from the first key frame set and one reference frame from the second key frame set;

calculating a hash distance for each pair of counterpart frames to obtain a plurality of hash distances between source frames of the source video shot and reference frames of the reference video shot;

selecting a maximum hash distance in the plurality of hash distances as a Hausdorff distance for comparing the source video shot to the reference video shot, the maximum hash distance being a distance between an image hash in a source frame and a reference frame of a single pair of counterpart frames; and

determining that the source video shot is similar to the reference video shot when the Hausdorff distance is below a predetermined threshold.

2. The method of claim 1 , wherein selecting the sub-set of frames from the first group of frames comprises selecting the sub-set of frames using a regional histogram difference of the first group of frames.

3. The method of claim 2 , wherein selecting the sub-set of frames further comprises truncating results of the regional histogram difference when a number of selected frames is greater than a predetermined number.

4. The method of claim 3 , further comprising selecting the predetermined number in accordance with a video type associated with the first group of frames.

5. The method of claim 1 , further comprising calculating an image hash for each frame in the first key frame set.

6. The method of claim 5 , wherein calculating an image hash comprises:

extracting feature points of the each frame of the first key frame set;

performing a pseudorandom permutation of the feature points to produce permutated feature points;

calculating transform coefficients of the permutated feature points; and

hashing the transform coefficients.

7. The method of claim 1 , further comprising partitioning the source video shot from a source video.

8. The method of claim 1 , wherein the plurality of hash distances include at least a first hash distance and a second hash distance, the first hash distance and the second hash distance corresponding to different source frames in the source video shot.

9. A method of determining whether a source video shot is similar to a reference video shot, the method comprising:

electronically receiving a first group of frames corresponding to the source video shot;

selecting a sub-set of frames from the first group of frames as a first key frame set, wherein the sub-set of frames includes two or more different frames from the source video shot;

selecting a second key frame set from a second group of frames corresponding to the reference video shot;

identifying multiple pairs of counterpart frames for comparing the source video shot to the reference video shot, wherein each pair of counterpart frames includes one source frame from the first key frame set and one reference frame from the second key frame set;

calculating a hashing distance for each pair of counterpart frames to obtain a plurality of hash distances, wherein each hashing distance comprises a distance between an image hash in a source frame and the image hash in a reference frame of the respective pair of counterpart frames;

determining that a maximum hashing distance in the plurality of hashing distances is a Hausdorff distance for comparing the source video shot to the reference video shot; and

determining that the source video shot is similar to the reference video shot when the Hausdorff distance is below a threshold.

10. The method of claim 9 , further comprising partitioning the source video shot from a video sequence.

11. The method of claim 9 , wherein the plurality of hash distances include at least a first hash distance and a second hash distance, the first hash distance and the second hash distance corresponding to different source frames in the source video shot.

12. An apparatus comprising:

a processor; and

a computer readable storage medium storing programming for execution by the processor, the programming including instructions to:

receive a first group of frames corresponding to a source video shot;

select a first key frame set from the first group of frames, wherein the first key frame set includes two or more different frames from the source video shot;

select a second key frame set from a second group of frames corresponding to a reference video shot;

identify multiple pairs of counterpart frames for comparing the source video shot to the reference video shot, wherein each pair of counterpart frames includes one source frame from the first key frame set and one reference frame from the second key frame set;

calculate a hash distance for each pair of counterpart frames to obtain a plurality of hash distances between source frames of the source video shot and reference frames of the reference video shot;

select a maximum hash distance in the plurality of hash distances as a Hausdorff distance for comparing the source video shot to the reference video shot, the maximum hash distance being a distance between an image hash in a source frame and a reference frame of a single pair of counterpart frames; and

determine that the source video shot is similar to the reference video shot when the Hausdorff distance is below a threshold.

13. The apparatus of claim 12 , wherein the plurality of hash distances include at least a first hash distance and a second hash distance, the first hash distance and the second hash distance corresponding to different source frames in the source video shot.

14. A non-transitory computer readable medium with an executable program stored thereon, wherein the executable program includes instructions to:

receive a first group of frames corresponding to a source video shot;

select a first key frame set from the first group of frames, wherein the first key frame set includes two or more different frames from the source video shot;

select a second key frame set from a second group of frames corresponding to a reference video shot;

identify multiple pairs of counterpart frames for comparing the source video shot to the reference video shot, wherein each pair of counterpart frames includes one source frame from the first key frame set and one reference frame from the second key frame set;

calculate a hash distance for each pair of counterpart frames to obtain a plurality of hash distances between source frames of the source video shot and reference frames of the reference video shot;

select a maximum hash distance in the plurality of hash distances as a Hausdorff distance for comparing the source video shot to the reference video shot, the maximum hash distance being a distance between an image hash in a source frame and a reference frame of a single pair of counterpart frames; and

determine that the source video shot is similar to the reference video shot when the Hausdorff distance is below a threshold.

15. The non-transitory computer readable medium of claim 14 , wherein the plurality of hash distances include at least a first hash distance and a second hash distance, the first hash distance and the second hash distance corresponding to different source frames in the source video shot.