In order to overcome the disadvantages of Content Based Image Retrieval(CBIR), the study on RGB color moments and Munsell color moments shows that RGB color moments is only suitable to the simple pictures, while Munsell color moments cannot characterize the specific information of pictures. Using information fusion technology to retrieve the fusion of judgement level, multiple color space fusion color moment gives a better solution. By contrasting with the traditional fusion algorithm, an improved fusion algorithm named NewcombMNZ is proposed. Experimental results show that the improved algorithm can improve the retrieval accuracy and ranking value, and a good robustness to noise. It reduces the effect of poor features on fusion, and can be used in the case of few features.