Shot Boundary Detection using Pixel-to-Neighbor Image Differences in Video

This paper introduces a framework for shot boundary detection used by the Language and Media Processing Laboratory (LAMP) at University of Maryland. It handles detection of cuts, dissolves and fades. Generally, it computes pixel-to-neightbor image differences in the videos. The cut detection applies the so-called secondmax ratio criterion in a sequential image buffer. The dissolve detection is based on a skipping image difference and linearity error in a sequential image buffer. We also filter out the effects of camera flash. We submitted five runs to TRECVID 2004. Their run-ids are D0, D1, . . . , D4. They use the same set of algorithms with different parameter settings.