A novel color image segmentation scheme was proposed for retinopathic fundus diagnosis purpose, combing with the hue-preserving color image equalizaion and segmentation by means of mean shift filtering of feature space clustering algorithms. With the image equalization algorithm, incorrectly exposed fundus image was enhanced and pieces of information for diagnosis were improve effectively and accurately. After image segmentation with the equalized fundus, experiments showed that the remote and new hemorrhagic spots and areas and the main vessels were well-segmenting compared with the original underexposed retina fundus and its segmentation.