Abstract: The performance of content-based image retrieval (CBIR) depends to a great extent on the image feature descriptor. Among these descriptors, color difference histogram (CDH) has showed the great discriminative performance in CBIR. However, there are still some limitations in it: 1)only taking color difference of pixels in global region into account; 2)not considering the spatial structure among pixels. In this paper, to solve these problems, we propose a novel image representation, called texton correlation descriptor (TCD), which is applied to CBIR. First, we define uniform regions which contain discriminative information of images and then detect them by analyzing the relationship among low-level features (color value and local binary patterns) of pixels. Second, in order to character contrast and spatial structure information in uniform regions respectively, we propose the color difference feature which fuses color difference correlation and global color difference histogram, and texton frequency feature which fuses texton frequency correlation and texton frequency histogram. Finally, by combining these feature vectors, TCD not only characters two orthogonal properties: spatial structure and contrast, but also takes these properties in local and global uniform regions into account simultaneously so that TCD has better performance in CBIR. The experimental results show that the retrieval results of TCD is higher than that of other descriptors in image datasets, and thus demonstrate that TCD is more robust and discriminative in CBIR.