Human-inspired order-based block feature in the HSI color space for image retrieval

Abstract

Color is one of the most important descriptor in image processing. Color histogram is the most commonly used color feature and has proved to be stable representation of an image, but it might be similar in different kinds of images because it describe the global intensity distribution of images. Inspired by human image classification behavior, in this paper, a new color feature representation method called Order-based Block Color Feature (OBCF) and its application in image retrieval is proposed. Firstly, RGB values of an image were transferred to HSI values. Secondly, each color channel (H, S and I) of an image is divided into M X N blocks and then statistical values are computed to characterize the block's color features. Thirdly, block features of the same statistical value in the same row are sorted in ascending order to form a row's features. Finally, all row features are concatenated to form an image's color feature. The experimental results show that the OBCF method provides high retrieval accuracy compared with color histogram method.