A practical method for video scene segmentation

Video segmentation is a crucial pass to content-based video summarization and retrieval. In this paper, we present a practical method to efficiently group video content into semantic segments. First we detect shots with double-threshold method to find raw shots quickly, followed by redundant frames removal though spatial color distribution to get the key frames. Finally, we cluster the key frames using the inter-shot correlation via domain color histogram and motion intensity to get the final scenes.