A novel image retrieval model

Image retrieval is the hot point of researchers in many domains. Traditional text-based query methods use caption and keywords to annotate and retrieval image databases, which often consumes a mass of human labor. Feature vector based retrieval methods only can provide the query by example, and can't provide retrieval on semantic level. In this paper, we propose a novel image retrieval model that combines good qualities of those two methods above-mentioned. It utilizes the image low-level features and the. user relevance feedback mechanism to classify images and acquire high-level semantic information. Furthermore, the image classification and the semantic information are not fixed, which can be changed by the user according to his preference. Experiments show that our scheme can achieve high efficiency.