Using Fuzzy Information Granulation for Spatio-temporal Salient Unit Detection in Video Sequences
Bing Li · Dianzi xuebao · 2007
In this paper,we propose an approach to automatic extraction of spatio-temporal salient unit from video sequences.Firstly,a class-related feature granulation algorithm is proposed,which can map original feature space to concept space based on fuzzy information granular.To detect spatial salient regions,segmented homogenous regions are classified according their prominent importance.Then,salient regions are tracked and the results of tracking are sequences of temporal coherent regions,called spatio-temporal salient unit.Experimental results verify efficiency of proposed approach in spatial and temporal aspects.