Roi Coding And Its Application In Multispectral Image Compression

As the advancing of remote sensing technology, remote sensors with large amount of data keep on emerging and the multi-spectral image data compression is receiving more and more attention. Because of various objectives of the multispectral images, users may only be interested in certain areas of the image. During the compression, regions-of-interest (ROI) are compressed with high fidelity or lossless methods based on different purposes, while other non-ROI are processed with low fidelity or loss compression to achieve higher compression ratio without missing important information.First, this thesis discusses the significance of image compression and summarizes the evolution and current status of image compression technology, as well as introduces multi-spectral data compressing technology. Next, the structure of JPEG2000 system and its working principles are introduced. The wavelet transformation algorithm lifting scheme of wavelet and EBCOT coding technic, which are recommended in JPEG2000 standard, are elaborated. Then the characteristics of ROI of JPEG2000 standard are studied. Two ROI algorithms defined by JPEG2000, i.e., Maxshift and Scaling-based Method, are also elaborated and numerical simulations are also performed. Results from different methods are compared and analyzed. To solve the problem that Maxshift method can't control the relative quality between ROI and background, two improved algorithms, such as partial significant bitplane shift (PSBShift) and bitplane-by-bitplane shift (BbBShift), are implemented. Simulation results show that both algorithms have the merits of Maxshift withou its disadvantages, so that the ROI and background compression qualities can be adjusted conveniently. Finally, a multi-spectral image ROI compression algorithm that combines JPEG2000 and predictive coding method is implemented. Two modified ROI compression methods, i.e., PSBShift, BbBShift, are adopted into the multi-spectral image ROI compression algorithm. Numerical simulation results show that PSBShift and BbBShift coding methods can produce desired high quality ROI images, improve background image quality, and significantly enhance relative quality of ROI and BG images in Maxshift method.