Research On Content-Based Image Retrieval

Recently, CBIR is becoming the hotspot in the research community of multi-media technologies. Within CBIR, image retrieval in the compressed domain is the latest development. CBIR makes use of the image processing from lower level to higher level, and extracts the content features of images, and carrys out retrieval according to the content features. CBIR is able to inquire a set of images from the image data base, which are similar to the sample image given by a user.The main contents of this dissertation are as follows: 1) Some analysises are made about CBIR and the researchs about it, and an introduction is made about the retrieval based on DFT, DCT, DWT. 2) A study is made about the technology of JPEG compresses images, and the methods of CBIR in compressed domain is discussed. 3) An experiment system of CBIR is constructed for JPEG compressed image, and some experiments are conducted on compressed image retrieval based on a sample image. 4) Based on presentation of color descriptors of MPEG-7 and analysis of coding characteristic of JPEG, a new algorithm is proposed to extract color layout descriptor of compressed image on incomplete decoding, and CBIR experiment is implemented using the algorithm on the experiment system of CBIR. Experimental results show that our algorithm is fast and feasible.