Color image segmentation based on visual perception

Color image segmentation is a critical pre-process in image processing. Also it´s important in the field of computer vision and pattern recognition. In this paper, we first state some evidence in the human vision research. Not all the intensity from 0 to 255 in RGB spaces can be distinguished by human vision. So we reduce the level of the intensity in RGB space to 26,28,26 respectively, while maintaining the image´s visual features and reducing the categories of image´s information. So that we can achieve the most correct peak points, from which we can form all the possible cluster centers. As all these cluster centres consider the image´s self-information, so the cluster centres is accurate. It can reduce the times of the interation for segmentation, so to some extent reduce the time consumption. Our experimental results show that our method can be highly regulated, and the segmentation results are as good as traditional method.