Hyperspectral remote sensing imaging is capahle of providing rich spalial and spectal information of target scene. However, because of huge amount of the acquired data. there are heavy difficulties in storing and transmitting the data of hyperspectral remote sensing image. In this paper, three-dimensional integer wavelet transform and wavelet support vector regression are applied to the hyperspectral remote sensing image compression. Firstly, the image is decomposed into subbands with different scales by means of three-dimensional integer wavelet transform. Then the low frequency subband is directly coded hy using DPCM. For the high frequency subbands, the wavelet support vector regression can learn from the dependency between the coefficients of high frequency subbands. The wavelet coefficients are represented sparsely by small training samples or support vectors. As a result, the coefficients of high frequency subbands are compressed. Finally, effective entropy coding technique is used to encode support vectora and corresponding weights. The experimental result given in the paper shows that the proposed method has higher peak signal to noise ratio than the hyperspectral remote sensing image compression methods based on 3D SPIHT and JPEG2000 at the same bit rate.%高光谱遥感成像技术能够获取探测对象丰富的空间信息和光谱信息,但所得图像海量数据的存储与传输制约了其应用.为此,提出了一种基于三维整数小波变换和小波支持向量回归的高光谱遥感图像压缩方法.首先采用三维整数小波变换把高光谱遥感图像分解成不同尺度的多个子带.然后对低频子带直接进行DPCM编码,而对高频子带则利用小波支持向量回归学习其小波系数之间的相关性,并采用小部分训练样本即支持向量来稀疏表示小波系数,以此达到压缩高频小波系数的目的.最后对支持向量及其相应的权重进行熵编码.文中给出了实验结果,并与基于3D SPIHT和JPEG2000的高光谱遥感图像压缩方法进行了比较,结果表明:所提出的方法在相同比特率下能够获得更高的峰值信噪比.
展开▼
