A chroma texture-based method in color image retrieval

Texture is an important feature generally used in the content-based image retrieval (CBIR). Traditional methods of calculating texture features ignore the texture features caused by chroma differences. A new method of calculating chroma texture features is proposed in this paper. Large numbers of experiments were performed and proved that the chroma texture feature was a very important complement to the traditional luminance texture. The effectiveness of the image retrieval is improved significantly by combination of luminance texture and chroma texture with a lower-dimensional vector. The average ranking ratio is improved by 14.57%, and there is an obvious improvement of the average recall-precision curve.