基于矢量量化的高光谱图像无损压缩算法研究信息与通信工程专业论文.docxVIP

哈尔滨工业大学工学硕士学位论文 哈尔滨工业大学工学硕士学位论文 II II Abstract Remote sensing technology since the sixties has enjoyed rapid development of hyperspectral remote sensing images people will probe the environment for resource monitoring, military reconnaissance and many other fields. However, with the rapid development of imaging spectrometer, hyperspectral image data was massive increase in transmission and storage areas to the tremendous pressure. As the remote sensing information is very valuable and accurate information retrieval requires a higher degree, which for lossless or near lossless compression for more attention. Hyperspectral remote sensing image as a three-dimensional image is different from ordinary two-dimensional image, According to its characteristics, this thesis of hyperspectral remote sensing images of lossless compression algorithm. Firstly, from the perspective of the correlation characteristics of hyperspectral remote sensing images are analyzed, including spatial correlation, spectral correlation, entropy analysis, focusing on analysis of hyperspectral image correlation spectroscopy. Reached its spatial correlation with both, but also has spectral correlation, the spectral correlation is greater than the spatial correlation of the conclusions. Vector quantization by the three key technology components, codebook design, vector image generation, code search, this paper three techniques and implemented a system vector quantization lossless compression program. Codebook generation is one of the key vector quantization. In this paper, the codebook design algorithms, experimental analysis of code size and the codeword length of the compression performance, discussed optimum combinations to get good compression ratio. Image vector generation algorithm is one of key technologies. In this paper, the traditional M / RVQ algorithm was modified to further improve the hyperspectral images lossless compression performance of VQ. JPEG-LS with the classical method, compar