Content-based Video Retrieval

Due to the dramatically increase of the video data and the wide use of the video in human's daily life, people's demand for the video search becomes much higher than before. The traditional search method based on text could no longer satisfy the new requirement to the video search. Only through searching by the content of the video can the video search meet the need of the video user.This paper focuses on the research of three aspects, which include shot boundary detection, key frame extraction and segment retrieval based on the content.In the research of the shot boundary detection, the paper is inspired by a method called detection of Lens mutations based on spatio-temporal slice(STS). Through spatio-temporal slice vertically and horizontally, the method can detect cut by close matching method. This paper improves it and extends it to detect the gradual shot boundary. The experiments proved that the improved method could detect not only the abrupt shots but also the gradual shots.In the research of key-frame extraction, considering the robustness to small movement of HSV-histogram algorithm and the sensitivity of Sub-block algorithm, this paper integrates these two algorithms and comes up with a new algorithm of key-frame extraction, The key-frame extracted with this method can summary the shot much better.In the end, an algorithm of clip retrieval is modified. First, a method to calculate the number of similar shots is proposed. Then, by calculate the similarity granularity, mapping the similarity relation to array and analysis the array, this paper has obtained the similarity of the clips. Base on some factors, the similarity model is established. This model would reduce the complexity of the similarity calculation. Experiment has proved that a good retrieval result could be obtained with the help of this model.Base on the research discussed above, this paper also developes a software which realizes the clip retrieval based on content, and this has proved the better performance of the improved method.