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 method comprising:
obtaining, by the one or more computing devices, a plurality of query signatures of a query media clip;
determining, by one or more computing devices, that a reference media clip is a possible match to the query media clip;
accessing, by the one or more computing devices, a set of reference signatures of the reference media clip from a reference database;
for each query signature of the plurality of signatures, performing a similarity search between the query signature and the set of reference signatures, wherein performing the similarity search comprises:
determining distances between the query signature and respective reference signatures of the set of reference signatures, and
determining an individual score based on the distances between the query signature and the respective reference signatures of the set of reference signatures;
determining, by the one or more computing devices, an overall score for the reference media clip based on the individual scores for the query signatures of the plurality of signatures;
determining, by the one or more computing devices, that the overall score for the reference media clip satisfies a threshold; and
based on the determining that the overall score satisfies the threshold, providing, by the one or more computing devices, an indication that the reference media clip is a match to the query media clip.
2. The method of claim 1 :
wherein determining that the reference media clip is a possible match to the query media clip comprises determining that a segment occurring at a time range within the reference media clip is a possible match to the query media clip, and
wherein accessing the set of reference signatures of the reference media clip comprises accessing the set of reference signatures based on the reference signatures of the set of reference signatures corresponding to the time range.
3. The method of claim 1 , wherein the individual score for each query signature of the plurality of signatures is a minimum distance of the distances between the query signature and the respective reference signatures.
4. The method of claim 3 , wherein the overall score is an average of the individual scores for the query signatures of the plurality of signatures.
5. The method of claim 1 :
wherein each query signature of the plurality of signatures is an N-dimensional query signature,
wherein each reference signature of the set of reference signatures is an N-dimensional reference signature, and
wherein for each query signature of the plurality of signatures, determining the distances between the query signature and the respective reference signatures of the set of reference signatures comprises determining bit distances between the query signature and the respective reference signatures of the set of reference signatures.
6. The method of claim 1 , further comprising determining, using the plurality of query signatures, a plurality of reference signatures matching at least one of the query signatures, wherein the plurality of reference signatures correspond to plurality of reference media clips, and wherein the reference media clip is one of the plurality of reference media clips.
7. The method of claim 6 , wherein determining that the reference media clip is a possible match to the query media clip comprises determining that, from among the plurality of reference media clips, the reference media clip has the most corresponding reference signatures of the plurality of reference signatures.
8. A non-transitory computer-readable medium having stored therein instructions that are executable to cause a computing system to perform functions comprising:
obtaining a plurality of query signatures of a query media clip;
determining that a reference media clip is a possible match to the query media clip;
accessing a set of reference signatures of the reference media clip from a reference database;
for each query signature of the plurality of signatures, performing a similarity search between the query signature and the set of reference signatures, wherein performing the similarity search comprises:
determining distances between the query signature and respective reference signatures of the set of reference signatures, and
determining an individual score based on the distances between the query signature and the respective reference signatures of the set of reference signatures;
determining an overall score for the reference media clip based on the individual scores for the query signatures of the plurality of signatures;
determining that the overall score for the reference media clip satisfies a threshold, and
based on the determining that the overall score satisfies the threshold, providing an indication that the reference media clip is a match to the query media clip.
9. The non-transitory computer-readable medium of claim 8 :
wherein determining that the reference media clip is a possible match to the query media clip comprises determining that a segment occurring at a time range within the reference media clip is a possible match to the query media clip, and
wherein accessing the set of reference signatures of the reference media clip comprises accessing the set of reference signatures based on the reference signatures of the set of reference signatures corresponding to the time range.
10. The non-transitory computer-readable medium of claim 8 , wherein the individual score for each query signature of the plurality of signatures is a minimum distance of the distances between the query signature and the respective reference signatures.
11. The non-transitory computer-readable medium of claim 10 , wherein the overall score is an average of the individual scores for the query signatures of the plurality of signatures.
12. The non-transitory computer-readable medium of claim 8 :
wherein each query signature of the plurality of signatures is an N-dimensional query signature,
wherein each reference signature of the set of reference signatures is an N-dimensional reference signature, and
wherein for each query signature of the plurality of signatures, determining the distances between the query signature and the respective reference signatures of the set of reference signatures comprises determining bit distances between the query signature and the respective reference signatures of the set of reference signatures.
13. The non-transitory computer-readable medium of claim 8 , wherein the functions further comprise determining, using the plurality of query signatures, a plurality of reference signatures matching at least one of the query signatures, wherein the plurality of reference signatures correspond to plurality of reference media clips, and wherein the reference media clip is one of the plurality of reference media clips.
14. The non-transitory computer-readable medium of claim 8 , wherein determining that the reference media clip is a possible match to the query media clip comprises determining that, from among the plurality of reference media clips, the reference media clip has the most corresponding reference signatures of the plurality of reference signatures.
15. A computing system comprising:
a processing unit; and
a computer-readable medium having stored therein instructions that are executable to cause the computing system to perform functions comprising:
obtaining a plurality of query signatures of a query media clip,
determining that a reference media clip is a possible match to the query media clip,
accessing a set of reference signatures of the reference media clip from a reference database,
for each query signature of the plurality of signatures, performing a similarity search between the query signature and the set of reference signatures, wherein performing the similarity search comprises:
determining distances between the query signature and respective reference signatures of the set of reference signatures; and
determining an individual score based on the distances between the query signature and the respective reference signatures of the set of reference signatures;
determining an overall score for the reference media clip based on the individual scores for the query signatures of the plurality of signatures,
determining that the overall score for the reference media clip satisfies a threshold, and
based on the determining that the overall score satisfies the threshold, providing an indication that the reference media clip is a match to the query media clip.
16. The computing system of claim 15 :
wherein determining that the reference media clip is a possible match to the query media clip comprises determining that a segment occurring at a time range within the reference media clip is a possible match to the query media clip, and
wherein accessing the set of reference signatures of the reference media clip comprises accessing the set of reference signatures based on the reference signatures of the set of reference signatures corresponding to the time range.
17. The computing system of claim 15 , wherein the individual score for each query signature of the plurality of signatures is a minimum distance of the distances between the query signature and the respective reference signatures.
18. The computing system of claim 15 :
wherein each query signature of the plurality of signatures is an N-dimensional query signature,
wherein each reference signature of the set of reference signatures is an N-dimensional reference signature, and
wherein for each query signature of the plurality of signatures, determining the distances between the query signature and the respective reference signatures of the set of reference signatures comprises determining bit distances between the query signature and the respective reference signatures of the set of reference signatures.
19. The computing system of claim 15 , wherein the functions further comprise determining, using the plurality of query signatures, a plurality of reference signatures matching at least one of the query signatures, wherein the plurality of reference signatures correspond to plurality of reference media clips, and wherein the reference media clip is one of the plurality of reference media clips.
20. The computing system of claim 15 , wherein determining that the reference media clip is a possible match to the query media clip comprises determining that, from among the plurality of reference media clips, the reference media clip has the most corresponding reference signatures of the plurality of reference signatures.