A Image Retrieval Scheme Based On JPEG2000 Compressed Stream

As intuitionistic, visual, pellucid and full of information, image information is the richest source of visual information. There is much image information on the Internet. How to retrieve these image resources effectively becomes a subject of study. Image retrieval technology experienced development from TBIR(Text-based Image Retrieval) to CBIR(Content-based Image Retrieval). Nowadays, compressed-domain image retrieval techniques become the research focus in CBIR. Because it extracts feature in compressed-domain without decompressing compressed image fully, it improves processing speed and reduces the demand for computer resources.With the development of computer networks and the increasing demands of image information, JPEG2000 standard, due to its performance advantages, will become the new darlings. Image retrieval techniques based on JPEG2000 standard get more and more attention by researchers.This thesis analyzes the development of image retrieval technology, and researches compressed-domain image retrieval focused on DCT domain and DWT domain. Then the thesis analyzes the coding process of JPEG2000 standard, and researches some core technologies such as wavelet transform, bit-plane arithmetic coding, EBCOT and so on.This thesis proposed a new concept of basic quality level. Experimental analysis, on which the characteristics of the code block contributed to the basic quality level, draws a conclusion that the more a code block contributed to the basic quality level, the greater energy the code block including. An image retrieval scheme is proposed based on this conclusion in this thesis. Preliminary experimental results show that the proposed scheme is efficient and effective.