In this paper we present a new detector of points of focus. Based on psychological works stating that the symmetry attracts the eye, we have chosen objects' centers as points having a high visual interest [1]. Various works based on gray scale gradients used symmetry to detect points of focus. In this work, we propose the integration of color information in the process of detection which will provide more distinctive symmetry maps. We have test the robustness of our approach under image transformations like illumination changes, noise and occlusion. The results obtained prove that our detector offers best results comparing to the gray scale approach with more robustness face to image transformations.
展开▼
