Clothing image retrieval using color moment

Color Histogram based on HSV and Color Moments are widely used in image retrieval. In this paper, we focus on the research about the image retrieval and propose a new color feature, called Cascade Color Moments. By dividing the image into blocks, we add spatial information of the image into the color features. Cascade Color Moments feature is formed by cascading the color Moments of each block. The experimental results show that Cascade Color Moments is better than HSV Color Histogram and Color Moments when taking the retrieval precision into consideration. Meanwhile, a preprocessing procedure of the clothing image is proposed which mainly includes histogram equalization, foreground extraction and finally normalizing the size of the clothing area of the image into 64*128. And the preprocessing can greatly remove effect of the background, illumination and size of the image.