论文部分内容阅读
In this paper,we present a novel video salient object detection method based on group sparsity.Specially,both temporal and spatial saliency detection are considered.First,the temporal and spatial dictionaries are firstly constructed separately based on the appearance model and the motion feature contrast of each frame.Then the temporal saliency detection of each frame can be reduced to the sparse reconstruction problem based on the temporal dictionary to generate the temporal saliency map.A similar sparse reconstruction process is adopted to generate the spatial saliency map.Finally,the temporal saliency map and spatial saliency map are combined adaptively to generate the final spatiotemporal saliency map.Several experiments show that the proposed algorithm can detect salient objects accurately under scenes with large salient objects or dynamic backgrounds compared with the existing methods.