视频指纹文献专辑.pdf

Temporally Tolerant Video Matching Nicholas Diakopoulos Stephan Volmer nicholas.diakopoulos@igd.fraunhofer.de stephan.volmer@igd.fraunhofer.de Fraunhofer Institute for Computer Graphics Fraunhoferstra?e 5 64283 Darmstadt Germany ABSTRACT advertisement of a particular competitor in a given market A temporally tolerant notion of visual similarity for videos is segment, or if a particular company is running several re- de?ned in the domain of television advertisements. This no- lated advertisements in a given advertising campaign. tion encompasses robustness to channel and encoding noise, encoding rate, and reediting of the advertisements. Using The technical demands on such an automatic advertisement a color-based visual feature as a basis for describing the monitoring system are considerable. The detection of ad- frames of the video, a sub-shot segmentation of the video vertisements must ?rst of all be robust to the speci?c char- that is consistent across encoding rates and extensible to acteristics of the broadcast channel that is being monitored. streaming media is produced. This segmentation is subse- This is due to the fact that di?erent broadcasters may ap- quently utilized in a similarity matrix based matching al- ply di?erent signal ?lters to the broadcast media (e.g. color gorithm that e?ectively matches temporally reencoded and boosting). Additionally it must be robust against encoding reedited videos. Experimental matching results for the situ- rates (e.g. 15fps, 25fps, 30fps) and against encoding varia- ation of discrete video ?les are given. The method is devel-