Shot Boundary Detection with Sparse Presentation

Shot boundary detection is usually among the first steps to process produced videos like movies and to serve as basis for following steps like scene detection. In this paper, we propose an algorithm for shot boundary detection based on sparse presentation where each frame has a sparse presentation using basis pursuit denoising(BPDN). And then we feed this sparse presentation to SVM for the learning/detection task. Wel also consider other classifier as alternative for the second step. We compare our algorithm to baseline algorithms and the experiment results demonstrate our algorithm performs comparably well.