A Saliency Detection Model Based on Multi-feature Fusion

In this paper, we develop a saliency detection model which combines low and high level features. This model can be useful to the problems of content missing in image with large scale foreground and false detection in image with complex background when detecting salient regions with the existing models. To seek a solution to avoid content missing, our approach firstly adopted the blur region inhibition to reduce false detection to some extent, and then merged the saliency information both in space and in frequency domain. Experimental results show that the proposed model can solve the problem of content missing and false detection. Comparing to the other five typical models, our method can increase the detection accuracy by at least 16% while can outperform other systems in ROC analysis.