An integrated model for effective saliency prediction

摘要: In this paper, we proposed an integrated model of both semantic-aware and contrast-aware saliency (SCA) combining bottom-up top-down cues for effective eye fixation prediction. The contains two pathways. first pathway is a deep neural network customized saliency, which aims to capture the semantic information in images, especially presence meaningful objects object parts. second based on on-line feature learning maximization, learns adaptive representation input discovers high contrast salient patterns within image context. pathways characterize long-term short-term attention are using maxima normalization. Experimental results artificial images several benchmark dataset demonstrate superior performance better plausibility over classic approaches recent models.