Color segmentation for visual attention of mobile robots

Abstract

Segmentation is a very iniportant step in most qf the nmachine vision applications. A method for color segmentation is proposed as a basic requirement qf visuial attention system of mobile robots. After experiments with different color spaces, the HSI color space is selected for segmentation with nminimtm influence of shades and shadows. The proposed method first classifies the color ofseed pixel and then selects a method for region growing according to nature of the seed pixel Results obtained are very encouraging and accor-ding to requtiremcents qf the application area. This paper addresses the problem of obtaining regions from colored images such that the regions may have minimum influence of shades and shadows in order to produce good input for higher level processing in visual attention systems of mobile robots. HSI (HueSaturation-Intensity) space, sometimes also called HSV (Hue-Saturation-Value) or HSL (Hue-SaturationLuminance), has a good representation of the colors of human perception [3] and has good capability of dealing with highlights, shades, and shadows [2]. As these features suite the requirements of our application area hence our approach performs segmentation in HSI color space and proposes an improved algorithm to obtain better results.