2019 | vol. 67 | nr. 4 | art. 10

Automatic Re-Ranking in Clip Matching of Stereoscopic Video based on Clustering and Density

Fengfeng Duan, Jiaojiao Lu
Abstract
Re-ranking of video clips is one of the keys in content-based video retrieval and presentation. In order to better realize the presentation and location of stereoscopic video clips, we propose a re-ranking algorithm in clip matching of stereoscopic video based on clustering and density to solve the problems of low accuracy of retrieval and poor performance of ranking. In the algorithm, the similarity of stereoscopic video clips is calculated according to the visual, order and semantic similarity factors. With the clip similarity, the clustering is carried out firstly and the correlation among classes is calculated, by which the clustering re-ranking is achieved. Then the density is calculated in each class to measure the similarity between the elements in the class and query clip, so as to realize the re-ranking based on density. In clip matching and re-ranking experiment, compared with the typical algorithms, the similarity matching accuracy and re-ranking performance both are greatly improved. Experimental results show that the re-ranking of stereoscopic video clips in matching and retrieval can be better achieved with the proposed algorithm in this paper.
Keywords: Re-ranking; Clip Matching; Stereoscopic Video; Clustering; Density
To cite this article: Duan F.F., Lu J.J., “Automatic Re-ranking in Clip Matching of Stereoscopic Video Based on Clustering and Density”, in Electrotehnica, Electronica, Automatica (EEA), 2019, vol. 67, no. 4, pp. 75-84, ISSN 1582-5175.

 

Download