| Abstrakt: |
Achieving high effectiveness in Shot Boundary Detection (SBD) paves the way for high-level analysis of the video (keyframe extraction, story segmentation, etc.), which makes this step very important. Thus, the SBD problem has been extensively addressed, and many approaches have been proposed. As these approaches have their own different strengths and weaknesses, merging the outcomes of different detectors in order to obtain a better detector comes naturally. In this paper we propose an approach to SBD which takes into account the outcomes of two shot boundary detectors, using a rank fusion technique. This new detector is tested with videos from the TRECVid initiative, finding that it outperforms the two original methods. Moreover, the computation of this merging method is very fast, so it is very attractive for an operational environment. [ABSTRACT FROM AUTHOR] |
