An accurate and robust method for detecting video shot boundaries
Yi Gong · 2003
Many shot boundary detection methods have been proposed in recent years. Comparison studies have revealed that, among the existing methods, the local histogram-based method gives the highest precision rate if the highest shot boundary detection rate is not the first priority. In this paper, we present a new shot boundary detection method that surpasses the local histogram-based method in both the boundary detection rate and the precision rate. We propose a novel method of extracting color distributions from video frames to effectively suppress both the intra-frame and the inter-frame color variations caused by noise, minor illumination changes, as well as camera and object motions. We also provide a systematic approach to combine the color and the spatial features of video frames for more accurate shot boundary detection. An experimental evaluation confirms the superior performance of our proposed method.