Saliency Based Color Segmentation for Visual Attention in Mobile Robots

Image Segmentation is an essential and critical step in computer vision applications. This paper proposes a method for color segmentation as a basic requirement of visual attention systems of mobile robots. Segmentation is incorporated with computation of color contrast in order to gain computational speedup for visual attention. After experiments with different color spaces, the HIS color space is selected for segmentation allowing minimum influence of shades and shadows. The proposed method picks salient seed pixels according their color contrast values to start region construction then the color of the seed pixel is classified. The method to grow a particular region is decided according to nature of the seed pixel. Results obtained are very encouraging and are according to requirements of the application area.