In order to help disabled people in their current life, our laboratory develops a system with a mobile robot supporting a camera and an arm manipulator, and guided remotely by the handicapped person. The robot moves in an indoor environment so, in the camera images, much of the informations are related to textures and colors. A previous work permits to localize and guide the robot in 3D space, by use of segments from gray levels stereoscopic vision. Our aim is to use color, so we have to choose the color space, which is the best to detect and match segments. We have built a color images database with some different scenes and we have defined two ways for obtaining a notation for each color space, or color axis. One is an absolute evaluation of the space called “weighted notation” and the other is a comparative evaluation based on the segment histograms from color and gray level images. So, after a test stage, we hope to be able to determine the best three space components, which will be used by the mobile robot.