The Research Of Color Image Segmentation Algorithm Based On Color And Texture Features

The rapid accumulation of large digital image collections has created the need for efficient and intelligent schemes for color image processing technology.One of the most important and challenging components of color image processing technology is image segmentation. Since humans are the ultimate users of image processing systems, the research of image segmentaion algorithm based on visual peceptual characters that are meaningful to humans.A new approach for color-texture image segmentation is proposed that is based on viaual perceptual models about the processing of color and texture features.First, perceptual-based low-level color and texture features are derived, respectively. Then, a hierarchical image segmentation algorithm that combines these features is developed to obtain image segmentations.Two types of features are proposed. The first describes the local color composition in terms of image dominant colors, which on one hand, reflect the fact that the human visual system cannot simultaneously perceive a large number of colors, and on the other, consider the spatial relation between pixels. An adaptive color segmetation algorithm based on the type-2 fuzzy sets is proposed to obtain the dominant colors of images.The optimal color composition distance (OCCD) metric is introduced to measure the perceptual similarity between two color compostion feature vectors. The second is based on a multi-scale frequency decomposition, which approximates early processing in the human visual system. The local median energy of the subband coefficients is used as a simple but effective characterization of spatial texture. The median filter on one hand, suppress the enegy of texture region, and on the other, hold the energy of region boundaries between texture regions. Images can be segmented into sevaral diffirent directional texture classes according to the local median enegy of the subband coeffients which reflects the direction feature of textures.According to the spatial texture feature, first of all, the image is partitioned into different directional texture regions, and then, in each texture region, the mean shift algorithm, combining with the color feature, is used to obtain the segmentation regions with the same textue and simlilar color feather. Finally, a region merging process is implemented to avoid over-segmentation and obtain the final segmentation image.The performance of the proposed algorithms is demonstrated in the domain of low resolution images of natural scenes. The experimental results show that the proposed algorithm is an effective segmentation algorithm and robust to illumination variation.