Research And System Implementation On Content-based Image Retrieval

Content-Based Image Retrieval (CBIR) is an integrated technique, which retrieves image from image database on the basis of the content feature of the image. CBIR combines several technologies, such as image processing, computer vision, artificial intelligence and database, and has become an urgent research area.This dissertation makes extensive and deep research on CBIR technology. Several key technologies of CBIR are deeply analyzed.Three improved methods are proposed and capabilities of the algorithms are tested and compared with some other classical algorithms. At last, a CBIR system for testing retrieval algorithms is developed. The main content of the dissertation is summarized as follows:1. Several key techniques and algorithms of CBIR are deeply analyzed.2. A method for image retrieval based on wavelet generalized histogram is proposed. This new approach integrates the advantages of generalized histogram and wavelet transform, and it not only consideres the spatial correlation info but also takes full advantage of excellent character of wavelet transform.3. An improved approach for image retrieval based on generalized co-occurrence matrix is proposed. The new approach constructes four co-occurrence matrices based on generalized image in four directions, and then it pickes up the texture parameters of the four co-occurrence matrices for image retrieval.4. A remote sensing image retrieval method based on target object is proposed. In this method, the image is divided into target and background, and the pixels of the target and background are endowed with different powers during the retrieving.5. A CBIR system for testing retrieval algorithms is developed, which is used as the tested platform for the new approaches and other classical methods.