矢量维数分割量化的超光谱图像压缩方法

Compression algorithm of hyperspectral image based on vector dimension segmentation quantization

提出了一种矢量维数分割量化的超光谱图像压缩算法,通过维数分割将矢量分为几个部分,然后利用哈达玛变换的性质,在哈达玛域内单独设计每个部分的码书。设计过程中采用最优矢量量化器设计原则,并结合分步判断排除不等式算法与LBG(Linde Bazo Gray)聚类算法快速生成矢量量化过程的最终码书,使各个部分的码书性能达到最优,改善整体码书的性能。实验表明,这种算法在码书尺寸相同的情况下,图像的恢复质量以及复杂度都优于其他几种算法。

A hyperspectral image compression algorithm with dimension segmentation quantization on each vector is introduced. The algorithm adopts dimension segmentation to divide vector into several parts and de- signs each part of the codebook based on the nature of Hadamard transformation. Optimal vector quantizer de- sign principle is used in the designing process. Combined with a step-by-step exclude inequality algorithm and LBG （Linda Bazo Gray） clustering algorithm, the final codebook can be quickly generated. The whole codebook performance can be improved by designing each part of the codehook to achieve optimal performance in Ha- . damard domain. Experimental results show that the algorithm is superior to other algorithms in image recovery quality and complexity at the same codebook size.