Abstract: There are the problems that the extracted color feature is high dimension based on the
traditional color histogram, and the direction of texture is neglected based on the traditional
co-occurrence matrix. A new image retrieval algorithm combining the improved color histogram and
gray level co-occurrence matrix algorithm is proposed. The K-means clustering is used to cluster the
detected images in order to reduce the number of colors. The image codes are computed to form the
color histogram based on vector codes in the HSV space. So the color features are extracted. The gray
level co-occurrence matrix is used to extract the four eigenvalues of the detected image, and the four
eigenvalues are combined with the weighting factor determined by the direction measure. The texture
eigenvectors are obtained from normalizing the fused components. Finally, weighted average is used to
fuse the feature distance of color and texture. Compared with the other two algorithms, experimental
results show that our algorithm has higher recall and precision in general images and textured images.

Key words: image retrieval, color histogram, color clustering, vector coding, gray level co-occurrence
matrix, direction measurement