Hierarchical indexing of ocean survey video by mean shift clustering and MDL principle

Our contribution is proposing a novel framework for building a hierarchical index for videos. First, a set of key frames for the finest level of the hierarchy is computed by mean-shift clustering. Second, clusters are successively merged based on the minimum description length (MDL) principle, to form the upper levels of the hierarchy in a fine-to-coarse sequence. The advantage is providing the user with the flexibility of browsing the video at arbitrary levels of detail. For performance evaluation, we show the indexing results on two test videos, which are very close to the ground truth. In addition, we demonstrate a hierarchical browsing environment on a ocean survey video, which makes it more efficient for users to browse a long video in order to find interesting information based on the hierarchical index.