Method and apparatus for analyzing video based on spatiotemporal patterns

1. A video analysis apparatus comprising:

a generating means to generate at least one spatiotemporal pattern by performing pixel sampling on a plurality of frames in an input video;

a clustering means to generate a segmentation map using a first spatiotemporal pattern of the at least one spatiotemporal pattern;

a region of interest extracting means to extract the at least one region of interest having a sinusoidal pattern from the segmentation map; and

an analysis means to perform a frequency analysis of the at least one region of interest to determine whether the input video includes a predetermined type of content.

2. The video analysis apparatus of claim 1 , wherein the generating means generates the at least one spatiotemporal pattern by sampling the plurality of frames in at least one predetermined direction along a time axis.

3. The video analysis apparatus of claim 1 , wherein the generating means generates the at least one spatiotemporal pattern by executing scene-cut through at least one of color histogram analysis and color correlation analysis of the plurality of frames.

4. The video analysis apparatus of claim 1 , wherein the generating means generates the at least one spatiotemporal pattern by executing scene-cut based on continuity of the at least one spatiotemporal pattern.

5. The video analysis apparatus of claim 1 further comprising:

a segmenting means to generate a plurality of partial spatiotemporal patterns by segmenting each of the at least one spatiotemporal pattern based on a first time length,

wherein the clustering means generates the segmentation map by a region segmentation from a first partial spatiotemporal pattern among the plurality of partial spatiotemporal patterns.

6. The video analysis apparatus of claim 1 , wherein the region of interest extracting means extracts a plurality of bounding boxes including each of boundaries of the segmentation map, and selects at least one bounding box satisfying a predetermined condition as the at least one region of interest.

7. The video analysis apparatus of claim 1 , wherein the analysis means comprises:

a first transforming means to transform the at least one region of interest into a one dimensional signal to generate a first signal;

a second transforming means to transform the first signal to a frequency domain to generate a second signal; and

a determining means to determine whether the input video includes the predetermined type of content by analyzing a frequency distribution of the second signal.

8. The video analysis apparatus of claim 7 , wherein the first transforming means generates the first signal by applying a correlation function to the at least one region of interest.

9. The video analysis apparatus of claim 7 , wherein the analysis means determines that the input video includes the predetermined type of content when a dominant frequency in the frequency distribution of the second signal is within a predetermined frequency range.

10. The video analysis apparatus of claim 1 , wherein the determining means determines whether the input video includes the predetermined type of content by analyzing color information of the at least one region of interest and a region other than the at least one region of interest.

11. A video analysis method comprising:

generating at least one spatiotemporal pattern by performing pixel sampling on a plurality of frames in an input video;

generating a segmentation map using a first spatiotemporal pattern of the at least one spatiotemporal pattern;

extracting at least one region of interest having a sinusoidal pattern from the segmentation map; and

performing a frequency analysis of the at least one region of interest to determine whether the input video includes a predetermined type of content.

12. The video analysis method of claim 11 , wherein generating at least one spatiotemporal pattern comprises generating the at least one spatiotemporal pattern by sampling the plurality of frames in at least one predetermined direction along a time axis.

13. The video analysis method of claim 11 , wherein generating at least one spatiotemporal pattern comprises generating the at least one spatiotemporal pattern by executing scene-cut through at least one of color histogram analysis and color correlation analysis of the plurality of frames; and

generating the at least one spatiotemporal pattern by executing scene-cut based on continuity of the at least one spatiotemporal pattern.

14. The video analysis method of claim 11 , wherein the extracting generating a segmentation map comprises:

generating a plurality of partial spatiotemporal patterns by segmenting each of the at least one spatiotemporal pattern based on a first time length; and

generating the segmentation map by a region segmentation from a first partial spatiotemporal pattern among the plurality of partial spatiotemporal patterns.

15. The video analysis method of claim 11 , wherein the extracting of the region of interest comprises:

extracting a plurality of bounding boxes including each of boundaries of the segmentation map; and

selecting at least one bounding box satisfying a predetermined condition as the at least one region of interest.

16. The video analysis method of claim 11 , wherein the performing a frequency analysis comprises:

transforming the at least one region of interest into a one-dimensional signal to generate a first signal;

transforming the first signal to a frequency domain to generate a second signal; and

determining whether the input video includes the predetermined type of content by analyzing a frequency distribution of the second signal.

17. A non-transitory computer-readable medium comprising a program for instructing a computer to perform a video analysis method, the video analysis method comprising:

generating at least one spatiotemporal pattern by performing pixel sampling on a plurality of frames in an input video;

generating a segmentation map using a first partial spatiotemporal pattern of the at least one spatiotemporal pattern;

extracting at least one region of interest having a sinusoidal pattern from the segmentation map; and

performing a frequency analysis of the at least one region of interest to determine whether the input video includes a predetermined type of content.