Application Research Of Formal Concept Lattices In Semantic Image Retrieval

Along with the development of computer vision, Multimedia and database technology, image information which is thought as a main kind of multimedia information, is widely employed in various fields. For management and retrieval of the information, the CBIR (Content-Based Image Retrieval) has emerged to be one of the important research topics in image domain. The low-level image features are firstly extracted automatically, and then the similarity between images in feature space can be calculated. At last, the desire retrieval results are obtained. This method is devoted to dealing with the weaknesses of texture-based image retrieval of much effort on artificial annotation and subject of human annotation. So the CBIR is developed so fast that it becomes the main technology in image retrieval filed and many retrieval systems are designed and implemented.Because there is existed serious difference between low-level image features and human understanding to image, the former cannot describe image content exactly. That is to say, there is a "semantic gap" between low-level features and image semantics. This leads to the semantic image retrieval and classification. The approach combines the semantic information of images with visual features, among at retrieving or classifying images. The Semantic image retrieval has recently attracted many researchers and become an interesting issue in multimedia information retrieval. It is difficult to extract semantics, represent and apply it for the complexity of image semantics. Thus there is a challenging problem.This thesis discusses the key techniques on semantic image retrieval, and proposes a new method of semantic image retrieval based on the concept lattices. The potential concept structures and mutual relations of the concepts in the image are analyzed by using of formal concepts. The linguistic variables are employed to describe semantic feature of image and concept lattice is constructed in terms of these semantic fuzzy value. Then semantic image retrieval based on the concept lattices will work. The experiment shows that the result obtained by this method is better consistency with human visual perception.