Blur Detection For Video Streams In The Compressed Domain

In ubiquitous multimedia area, the number of digital videos increases dramatically with various qualities in video frames. Artifacts such as blur may commonly exist in video treams which will disturb compression and retrieval applications. This paper proposes an algorithm that detects video frames with global blur and partial blur. The proposed algorithm extracts the features which represent blurs according to analyses of DCT coefficients characters and motion vectors information firstly. And then detects blurs by suitable model efficiently. The method presented in this paper is low in complexity while high in performance. Experimental results demonstrate the high-accuracy global and partial blurs detection of the proposed scheme.