Support Research And Realization Of Semantic Image Retrieval System

The rapid development of computer, multimedia and Internet techniques has produced too large amount of images. Therefore, it becomes an urgent problem that how to find needed image efficiently in large-scale image database. Two effective ways has been proposed to solve the problem: one is content-based image retrieval technique which search target images by low-level content feature. The other is semantic image retrieval technique which in search of ways to get images semantic information.Content-based image retrieval system remedies the deficiency existed in text-based image retrieval system by objectively describing intrinsic characteristics of image at the cost of discarding semantic cue of image and reducing retrieval precision of system.This paper presents of a project that extracts the low level features called as the representative colors from the images, at the same time we import the technique of relevance feedback and presents a new approach called word net to establish the link from the low-level feature vectors to the semantics. By cooperation between human and computer, we can make up the computer's limited ability of understanding and enhancing the effect of retrieval. The key technique of relevance feedback and semantic query is the important mark that the technique of CBIR develops from low-grade to high-grade. So the research of this paper has important meaning in learning value and practice application.The main contributions of this paper are:We first do research on the low-layer physical of image. Due to the stability and having a thick skin to the size and direction, we extract the color feature to describe the images. In order to overcome the disadvantage of the widely used methods in CBIR system: such as Color Histogram, Color Moments which can't express the spatial relationship from images. We introduce region-based representative color feature extraction, synthetically take the pixels statistic feature and the spatial information into account, at the same time we can save the storage space and the calculate time.