Content-Based Image Retrieval Using Three Pixels Color Co-Occurrence Matrix *
Guanghai Liu, Zuoyong Li · 2012
A new color descriptor, namely three pixels color co-occurrence matrix, is proposed in this paper. It imple- ments content-based image retrieval via representing the spatial correlation of the same three pixels. The proposed de- scriptor first converts color image from RGB color space to Lab color space. The original color image is then quantized into 90 colors and co-occurrence matrix used for image analysis. Finally, 90-dimensional vector is adopted to describe image features. Image retrieval experimental results shown that the performances of the proposed descriptor are better than that of gray level co-occurrence matrix and MPEG-7 color layout descriptor. It has good discriminative power of color structure distribution, and can combine the color and texture features.