Researches On Remote-sensing Image Coding

Remote-sensing images have found wide applications in human life and military affairs, so the image coding technologies suitable for various requirements are of great practical significance. Different applications need the suitable coding algorithms to satisfy the special requirements. Main researches in recent years on image coding are proposed systematically in this thesis, with emphasis on the theory analysis and the algorithm study for wavelet-based embedded image coding.EBCOT(Embedded Block Coding with Optimized Truncation) is studied seriously in the thesis, then a fast algorithm for JPEG2000 hardware implementation is proposed with lower complexity. An algorithm based on rate-distortion slope lifting for ROI(Region of Interest) coding is also proposed such that the lifting and reverse lifting of the coefficients in wavelet domain are not required. The whole reconstructed image has better quality when the distortion is given. So it is suitable for the systems in which the ROI distortion is given or for the clients/server coding systems.The image compression based on absolute error measure is of great importance on measurement engineering and the digital information storage. The algorithm is proposed in which wavelet domain coding and error image coding are mixed, then their code-streams are optimized so that better compression efficiency is obtained and the error propagation decreased.The hyper-spectral image technology is developing rapidly, and has been widely used in human life and military affairs. Because of the specificity of the image, the hyper-spectral image compression must be able to prevent the error propagation. The variable run algorithm with lower complexity is proposed suitable for easier hardware implementation and with good ability for deducing the error propagation (the error propagation is limited within a few rows).There are offsets between the multi-spectral images. The 3-D (Three Dimensional) wavelet transform based algorithms are too complicated for hardware implementation. In this thesis the offset between two adjacent images is worked out by matching in wavelet domain. Then the difference between the images and masks is coded, which is based on wavelet transform. Therefore, some shortcomings based on 3-DWT algorithm such as large amount of storage and long coding delay can be avoided. This algorithm has the same complexity as the one of single image, but the reconstructed image has better quality.