Abstract
Visual attention has a primary role in salient object detection. This paper presents a visual saliency detection method that extracts the salient region from an image based on Human Visual Attention System. The core idea of the proposed method contemplates the laws of the Gestalt principle for object-based visual attention. According to the law of Gestalt psychology, the convexity of an object is the most important cue to attract human attention. The convexity of an object plays a vital role to segregate the figureure from its background. This proposed method aims to unify the contrast information from the various colour channels and generate an intermediate saliency map of the desired object. Then Convex Hull-based object priors have been evaluated to estimate the final saliency map. Our approach has been validated using publicly available datasets, i.e. ECSSD and MSRA. Experimental results show that the proposed method is outperforming the existing state-of-art method.
