基于提升方案的多光谱遥感图像有损压缩算法

Lossy compression algorithm of remotely sensed multispectral images based on lifting scheme

在分析多光谱遥感图像谱间和空间数据特点的基础上,提出了一种DPCM线性预测与基于提升方案的整数小波变换相结合的多光谱遥感图像有损压缩算法。在谱间采用DPCM预测去除谱间相关性;在谱内采用整数小波变换去除空间相关性,根据不同子带对目标识别的重要程度,选择不同的量化阈值和量化步长进行量化,并分别对各个子带量化后的数据和重要图表采用固定比特平面编码和游程编码,实现高效的多光谱遥感图像压缩。实验结果表明,该算法在一定的压缩比下,重构图像具有较高的峰值信噪比,并且算法硬件实现简单,对内存的需求低。

On the basis of analyzing characteristic of spectral bands and spatial dimensions data of remotely sensed multispectral images,a lossy compression algorithm of remotely sensed multispectral hybrid DPCM and integer wavelet transform is proposed.Between spectral bands,DPCM prediction approach is applied to remove the correlation between spectral bands.In spectral band,integer wavelet transform is applied to remove the correlation of spatial.According to the important degree of the different subbands for target recognition,different quantification threshold values and quantification steps are chosen in the quantification,and fixed bit-plane coding and RLE are individually used to quantify data of every subband and important graph,which realizes the high efficiency compression of remotely sensed multispectral images.The result of experiment shows that reconstructed image by the algorithm has higher PSNR in certain compression ratio.In addition,the algorithm requires small storage and is easy to be realized in hardware.