Detecting visual saliency using image signature

Saliency detection is one of the prominent areas of research since it is used in many computer vision applications. In this paper a visual saliency model is proposed which is bottom-up approach that is used to detect salient regions of an image. The proposed system combines two properties to find the salient regions. First, the proposed system uses a special image descriptor called image signature which approximates the foreground of the image and the foreground separated image is filtered to highlight visually salient regions. Secondly, it utilizes the property of human visual system called human visual acuity to enhance the saliency map. This saliency approach predicts human fixations in a better way than other existing methods on Bruce and Tsotsos[1] benchmark dataset.