Using Similarity Sequence for Shot Organization in Video Retrieval
相似顺序图用于视频镜头的组织

Shot organization is very important in content-based video retrieval, but in most cases automatic shot organization methods fail to give satisfactory results, thus human-involved interactive organization is required. In this paper, a new shot organization method called similarity sequence is proposed, which reorders the r-frames of shot and makes similar r-frame appear near each other in the generated sequence, thus provides an effective interactive organizing interface. To compare the performance of different sequences, a distance measure called similarity distance is proposed, upon which a gene algorithm calculating similarity sequence is developed and applied to r-frame sequence reordering in shot organization. Basic ideas and algorithm modules are given in detail, also with experimental results and analysis.