摘要

In this paper, image retrieval method based on statistics and projection is presented. At first, images are transferred from the RGB color model to the HSI color model, and then images are segmented into sub-images (or blocks) and extract the mainhue and mainintensity in these sub-images. Images's mainhue and mainintensity are projected onto the horizontal and vertical directions, respectively, and four projective histograms were obtained. Finally, the first three central moments are taken as feature as to compute the similarity between images.