Image Retrieval Based on Regional Color-Texture Features Description and DPF Matching

A new method for color image retrieval is introduced in this paper. At first, the image is parted into some connected regions. Rational Legendre chromaticity distribution moments and texture co-occurrence matrices are computed to represent the color and texture features (Regional Color Texture: RCT). Then, Dynamic Partial distance Function (DPF) and dynamic regions matching are used to weight RCT features, for short RCT_DPF. Experimental results indicate that this method RCT-DPF has good performance in image retrieval. To compare with Texture Co-occurrence Matrices(TCM), Multi-component Co-occurrence Matrices(MCM) and RCT, RCT-DPF is better precision.