Use of the compressed domain features for summarizing videos

This paper presents a new technique in high efficiency video coding (HEVC) standard which is video summarization. By this standard, the features of the frames are extracted from compressed mode, the conversion block features are used and by a clustering algorithm, a method is presented for summarizing. A lot of research has been done on summarizing, but most of them are pixel-based which uses a lot of time and memory for summarization. In the proposed algorithm, first, the characteristics of the conversion blocks of the compressed video are extracted according to the HEVC standard. Then, by counting the conversion blocks and plotting the histogram, a threshold is determined. In the following, an incremental clustering algorithm accommodates similar frames in specific clusters. A representative set of each cluster is considered as a summarized frame. The results indicate that the proposed algorithm significantly acted better in detecting similar frames and has less computational load.