A Novel RS-based Key Frame Representation for Video Mining in Compressed-Domain

It is a challenging issue to analyze video content for video mining tasks due to lacking of effective representation of video. In this paper, we propose a novel key frame representation algorithm based on rough sets (RS) in discrete cosine transform (DCT) compressed-domain. Firstly, we extract DCT coefficients in compressed-domain, select and preprocess the DC coefficients that derived from DCT coefficients. Secondly, we construct information system with DC coefficients. Finally, we reduce information system using attributes reduced theory of RS, and obtained the representation of the video frames by reduced DC coefficients. Experimental results show that the proposed algorithm is fast and effective. Compared to conventional algorithm, our algorithm enjoys the following advantages: (1) the numbers of the key frame extracted using our algorithm become more scientific; (2) the algorithm can avoid the expensive computations in decompression processes.