A novel approach for saliency detection based on multiscale phase spectrum

Abstract

Saliency detection is one of the core modules which is incorporated in various industrial applications such as computer vision, object based surveillance, content adaptive data delivery for low data rate systems. Saliency map represents the significance of objects in the scene. As in a visual scene all objects do not get equal attention. Objects which are different from their surroundings, automatically get more visual importance. Hardware based eye tracker devices are commercially available but the cost is comparatively very high. So to match the gaze attention, various algorithms are explored by several researchers. In this paper, multiscale phase spectrum based efficient saliency detection method is proposed. The proposed method is performing better by yielding better results than other methods. Four state-of-art of salient object detection techniques are analyzed along with the ground truth and compared against the proposed method for experimental evaluation.