Re-ranking of Stereo Video Retrieval Results Based on Clustering and Density

Retrieval result re-ranking is one of the keys of content-based video retrieval and presentation. In order to better realize the display and location of returned stereo video clips, a re-ranking method of stereo video clip retrieval based on clustering and density is proposed to solve the problems of low accuracy in re-ranking and low efficiency in retrieval. According to the similarity of content, the similar clustering is implemented firstly in the method, and the correlation among classes is calculated, so the clustering ranking can be realized according to the correlation. Then, the density is calculated within the classes to measure the similarity between the elements in each class and the query clip, so as to achieve density-based ranking. In the experiment, compared with the situation without re-ranking, it can better meet the user's needs of browsing in similarity order. The experimental results show that the method can achieve better re-ranking in retrieval results of stereo video clips.