By means of self-organizing clustering

a new color-based image retrieval method is proposed in the paper.According to the colors' distributing information in the image

every pixel is assigned a weighing value and thus the initial number of clustering can be confirmed.Therefore

those weighed pixels are clustered and the dominant colors' statistical features are acquired.Based on the dominant colors spread in the image

the colors' moment features are extracted to present their special features simultaneously.Therefore

the whole image's content can be expressed from general statistic to partial distributing by these two kinds of features.The experiments verify the method mentioned above is more efficiently than those ways based on color histogram.