Image Retrieval Based on Texton Autocorrelograms

A new color descriptor is proposed in this paper.The descriptor can represent the spatial correlation of color based on textons.Firstly,quantify the original images into 64 colors and analyze textons context,and then calculate the statistical information of textons to describe image features.Image retrieval experimental results show that the performances of our proposed method are better than that of color histogram and color autocorrelograms,and the computational complexity and feature dimension are smaller.