A Novel Scene Cut Detection Method in H.264/AVC Compression Domain

As an advanced video compression standard, H.264/AVC has been applied to various fields suchas video surveillance, video conference, and wireless videocommunication. This paper presents a novel scene cut detection method in H.264/AVC baseline profile compression domain, which takes advantage of the available features from H.264/AVC bitstreams, including chroma prediction modes, motion vectors, macroblock types, and soon. Moreover, in this method, four new criterions usedfor scene cut detection have been proposed, i.e. the distribution difference of chroma prediction modes, the distribution difference of macroblock types, the accumulativemotion amount, and the difference of motion vector angles.The thresholds of the criterions are mainly determined bythe minimum error Bayesian decision. Experimental results show that the proposed method can detect the scenecuts at I-frames and P-frames correctly without the information of bi-directional prediction which is not availablein H.264/AVC baseline profile.