Rushes exploitation 2006 by CAS MCG
Sheng Tang,Yong-Dong Zhang,Jin-Tao Li,Xue-Feng Pan,Tian Xia,Ming Li,Anan Liu,Lei Bao,Shu-Chang Liu,Quan-Feng Yan,Li Tan
2006-01-01
Abstract:In our rushes exploitation task of TRECVID 2006, we propose a novel and interactive rushes video selection and editing method based on hierarchical browsing of key frames, where high level features of each key frame such as face, interview, person, crowd, building, outdoor, waterbody, and other information about redundancy and repetition are displayed at same time for helping editors to select what they really want. During high level feature extraction, we propose a multi-modal interview detection method based on audio classification and face detection, and a new repetition detection method based on spatio-temporal slice. We also detect some concepts such as crowd, building, outdoor, waterbody based on SVM classifiers. Additionally, we characterize rushes by categorization camera motion for inferring intention. Due to the difficulty of high level feature extraction and the diversity of editor’s requirements, our hierarchical browsing method along with extracted information may be a good choice for rushes exploitation.