Shot boundary detection using second order statistics of gray level co- occurrence matrix

The readily and easily available nature of capturing devices made enormous amounts of video available in day-to-day life. Processing of such a lengthy video is a time consuming process, therefore researchers have introduced key frames. Key frame in short can be visualized as a frame that represents the information present in entire video shot. Detecting shot boundaries plays a vital role in extracting key fames. The results of shot boundary detection shows effect on performance of further stages of processing, therefore a reliable shot boundary detection task forms corner stone in several applications such as video analysis and summarization, video abstraction and higher contextual segmentation etc. In this article a novel image gray level co-occurrence matrix based technique for shot boundary detection by calculating statistics of the current frame such as homogeneity, energy, correlation, contrast and comparing the same with the next frame. The proposed algorithm successfully detects the shot boundaries by considering the statistics captured by gray level co-occurrence matrix. The method is experimented on animation videos. Performance of the method is evaluated with evaluation parameters boundary recall, accuracy, detection percentage, missing factor. The investigational results demonstrate that the proposed algorithm performs better than state-of-art methods. The results are tabulated, plotted and discussed briefly.