Media fingerprinting and identification system
The overall architecture and details of a scalable video fingerprinting and identification system that is robust with respect to many classes of video distortions is described. In this system, a fingerprint for a piece of multimedia content is composed of a number of compact signatures, along with traversal hash signatures and associated metadata. Numerical descriptors are generated for features found in a multimedia clip, signatures are generated from these descriptors, and a reference signature database is constructed from these signatures. Query signatures are also generated for a query multimedia clip. These query signatures are searched against the reference database using a fast similarity search procedure, to produce a candidate list of matching signatures. This candidate list is further analyzed to find the most likely reference matches. Signature correlation is performed between the likely reference matches and the query clip to improve detection accuracy.
1. A computer-implemented method for video fingerprinting that is robust to video distortion, the method comprising:
determining, for respective sub-regions of a set of sub-regions that each include pixels of a frame of a query video, using a video fingerprinting system, respective intensity values corresponding to pixels included within the respective sub-regions, wherein boundaries of the set of sub-regions are defined such that some of the sub-regions include different numbers of pixels, and wherein the query video is a distorted version of a reference video corresponding to the query video;
determining, for the respective sub-regions of the set of sub-regions, using the video fingerprinting system, respective sub-region values based on the respective intensity values;
generating, using the video fingerprinting system, fingerprint data for the frame based on the respective sub-region values; and
identifying the reference video corresponding to the query video using the fingerprint data and a reference database.
2. The computer-implemented method of claim 1 , wherein the query video is a cropped version of the reference video, is encoded at a different quality than a quality at which the reference video is encoded, or includes an image overlay.
3. The computer-implemented method of claim 1 , wherein determining, for respective sub-regions of the set of sub-regions, respective sub-region values based on the respective intensity values comprises computing, for respective sub-regions of the set of sub-regions, L respective sub-region values based on the respective intensity values, wherein L is an integer that is greater than one.
4. The computer-implemented method of claim 3 , wherein generating the fingerprint data for the frame based on the respective sub-region values comprises deriving a descriptor from the L sub-region values for the respective sub-regions.
5. The computer-implemented method of claim 4 , wherein generating the fingerprint data for the frame based on the respective sub-region values further comprises determining a signature using the descriptor.
6. The computer-implemented method of claim 4 , wherein deriving the descriptor from the L sub-region values for the set of sub-regions comprises concatenating the L sub-region values for the set of sub-regions.
7. The computer-implemented method of claim 1 , wherein, for a given sub-region of the set of sub-regions, determining the respective intensity values corresponding to pixels included within the given sub-region comprises:
subdividing the sub-region into an M×M grid of pixels, wherein M is an integer that is greater than one; and
for each sub-grid area of the M×M grid of pixels, combining intensity values of pixels of the sub-grid area to produce a re-sampled intensity value,
wherein the sub-region value for the given sub-region is determined based on the re-sampled intensity value.
8. A video fingerprinting system comprising:
a processing unit; and
a non-transitory computer-readable medium having stored therein instructions that are executable by one or more computers to cause the video fingerprinting system to perform functions comprising:
determining, for respective sub-regions of a set of sub-regions that each include pixels of a frame of a query video, respective intensity values corresponding to pixels included within the respective sub-regions, wherein boundaries of the set of sub-regions are defined such that some of the sub-regions include different numbers of pixels, and wherein the query video is a distorted version of a reference video corresponding to the query video,
determining, for the respective sub-regions of the set of sub-regions, respective sub-region values based on the respective intensity values,
generating fingerprint data for the frame based on the respective sub-region values, and
identifying the reference video corresponding to the query video using the fingerprint data and a reference database.
9. The video fingerprinting system of claim 8 , wherein the query video is a cropped version of the reference video, is encoded at a different quality than a quality at which the reference video is encoded, or includes an image overlay.
10. The video fingerprinting system of claim 8 , wherein determining, for respective sub-regions of the set of sub-regions, respective sub-region values based on the respective intensity values comprises computing, for respective sub-regions of the set of sub-regions, L respective sub-region values based on the respective intensity values, wherein L is an integer that is greater than one.
11. The video fingerprinting system of claim 10 , wherein generating the fingerprint data for the frame based on the respective sub-region values comprises deriving a descriptor from the L sub-region values for the respective sub-regions.
12. The video fingerprinting system of claim 11 , wherein generating the fingerprint data for the frame based on the respective sub-region values further comprises determining a signature using the descriptor.
13. The video fingerprinting system of claim 11 , wherein deriving the descriptor from the L sub-region values for the set of sub-regions comprises concatenating the L sub-region values for the set of sub-regions.
14. The video fingerprinting system of claim 8 , wherein, for a given sub-region of the set of sub-regions, determining the respective intensity values corresponding to pixels included within the given sub-region comprises:
subdividing the sub-region into an M×M grid of pixels, wherein M is an integer that is greater than one; and
for each sub-grid area of the M×M grid of pixels, combining intensity values of pixels of the sub-grid area to produce a re-sampled intensity value,
wherein the sub-region value for the given sub-region is determined based on the re-sampled intensity value.
15. A non-transitory computer-readable medium having stored therein instructions that are executable by one or more computers to cause a video fingerprinting system to perform functions comprising:
determining, for respective sub-regions of a set of sub-regions that each include pixels of a frame of a query video, respective intensity values corresponding to pixels included within the respective sub-regions of the set of sub-regions, wherein boundaries of the set of sub-regions are defined such that some of the sub-regions include different numbers of pixels, and wherein the query video is a distorted version of a reference video corresponding to the query video,
determining, for the respective sub-regions of the set of sub-regions, respective sub-region values based on the respective intensity values,
generating fingerprint data for the frame based on the respective sub-region values, and
identifying the reference video corresponding to the query video using the fingerprint data and a reference database.
16. The non-transitory computer-readable medium of claim 15 , wherein the query video is a cropped version of the reference video, is encoded at a different quality than a quality at which the reference video is encoded, or includes an image overlay.
17. The non-transitory computer-readable medium of claim 15 , wherein determining, for respective sub-regions of the set of sub-regions, respective sub-region values based on the respective intensity values comprises computing, for respective sub-regions of the set of sub-regions, L respective sub-region values based on the respective intensity values, wherein L is an integer that is greater than one.
18. The non-transitory computer-readable medium of claim 17 , wherein generating the fingerprint data for the frame based on the respective sub-region values comprises deriving a descriptor from the L sub-region values for the respective sub-regions.
19. The non-transitory computer-readable medium of claim 18 , wherein generating the fingerprint data for the frame based on the respective sub-region values further comprises determining a signature using the descriptor.
20. The non-transitory computer-readable medium of claim 18 , wherein deriving the descriptor from the L sub-region values for the set of sub-regions comprises concatenating the L sub-region values for the set of sub-regions.