Visual saliency detection based on region descriptors and prior knowledge

Abstract

Visual saliency detection not only plays a significant role, but it is also a challenging task in computer vision. In this paper we propose a new method for saliency detection. It incorporates visual features and spatial information with a guidance of prior saliency knowledge. To provide more accurate visual cues, region descriptors are introduced for image segments by computing two saliency measures, namely feature distinctiveness and spatial distribution. In contrast to previous models which linearly combine basic features for visual cues, we provide nonlinear integration of features. In addition, by taking the advantage of the prior saliency distribution obtained from a convex hull of salient points, we heighten the contrast of fore- and background. Thereby we enhance the final saliency map that uniformly covers the salient objects, while tone down the nonsalient background. Experimental results on a benchmark dataset show that our saliency detection model performs favorably against the state-of-the-art approaches. A detailed experimental evaluation demonstrates that our algorithm excels at saliency detection in cluttered images.