Video Modification using Visual Saliency Map for Guiding Human's Gaze

Abstract

It is commonly believed that by drawing human's attention to particular objects, human activity can be effectively facilitated. The aim of this paper is to propose a novel video modification method for directing gaze to specific objects of interest, using a saliency map model based on L^*a^*b^* color space. First, we compute the saliency map, and modify the L^*a^*b^* components of each frame based on the obtained saliency map for adjusting saliency value. And then, we repeat this process until the specified objects has become the most salient in each frame. By applying the our attention retargeting method to a video, we can obtain the video in which the specific object shows the highest saliency.