基于变换编码的压缩技术因其高压缩率高性能易于实现等优点，是目前应用最广泛的图像压缩方法之一。然而图像因其内容丰富多样，纹理和几何形状特征差异较大，图像的各局部统计特征也有不同。对整幅图像采用同一的变换技术不能充分描述各局部特征。因此，如何充分利用图像特征改进变换技术进而提高图像压缩算法性能是目前亟需解决的问题。本项目研究图像局部特征分析，并构造经验的相关系数矩阵- - 最优Toeplizt矩阵，进而修正变换矩阵，提高变换效率；在图像特征分析基础上，研究基于多阶提升系统的完全可逆整型变换技术，实现有损至无损渐进压缩功能，及较高的压缩性能。对高光谱图像，分析其空间特征与谱间特征，构造同组同类的KLT矩阵，实现基于聚类与三维可逆整型变换的有损至无损渐进压缩。此外，研究针对非规则任意形状的形状自适应交叠变换技术与区域自动分类技术，实现基于此技术的感兴趣区域图像压缩算法。

Transform coding technique is one of the most popular image compression method, since of its advantages such as high compression ratio, simple structure and so on. However, the local statistical charactoristic is different from each other, since that the content is of large variety, and texture and geometry shape character diverses a lot. As a result, it will not completely discribe every local character of the image if adopt one transform method to the whole image. So how to fully utilize image character is one question to be resolved now. This project will conduct research on image character analysis, designment of empirical correlation coefficient matrix- - optimal Toeplizt matrix, and then we will modify the transform matrix to improve transform efficiency. Followingly, reversibly integer transform based on multi-lifting scheme will be researched to realize progressively lossy to lossless image compression with high performance. For hyperspectral image, spatial and spectral character will be analized to design KLT matrix of same group and region. Besides, shape-adptive transform and region auto-classification technique will be researched to realize region of interest (ROI) image conding method.