Image salient region detection based on spatial and frequency domains
Ji Cha · Journal of Jilin University · 2014
According to the human visual attention mechanism,an image salient region detection method is proposed.The salient features of image are extracted in spatial domain using local clutter density contrast and global color distribution estimation,and extracted in frequency domain using the efficient band divided method.Then,referring to the principle that cells respond to stimuli,such features are combined using the theory of feature combination.Local clutter density contrast imitates the law of universal gravitation,computing the visual attention on sparse matrix model.The efficient band divided method is used in combining information of amplitude for salient feature detection in frequency domain,later utilizes all features with weight.Experiment results show that the algorithm can effectively extract salient regions.The visual attention model is effectively applied to virtual and reality interaction to detect the effective region in real scene.