Back to Publications

Author(s) Li, X., Zhan, Y., Ke, J., Zheng, H.
Title Shot retrieval based on fuzzy evolutionary aiNet and hybrid features
Abstract There are noises and interferences in video sequence, which make shot retrieval efficiency disappointing in some situations. In this paper, a novel method for shot retrieval is proposed which is based on fuzzy evolutionary aiNet and hybrid features. First of all, our proposed fuzzy evolutionary aiNet algorithm is utilized to extract key-frames in a video-sequence. To represent a key-frame, hybrid features of color feature, texture feature and spatial structure feature are extracted. Then, the features of extracted key-frames in the same shot are taken as an ensemble and mapped to high dimension space by nonlinear mapping, obeying Gaussian distribution. Furthermore, shot similarity is obtained by computing the probabilistic distance between distributions of the key-frame feature ensembles for two shots. Finally, the similar shots are retrieved effectively by using the similarity measure method based on hybrid features. Experimental results show the validity of the proposed method.
Download ZhLiKe11.pdf