基于残差偏置和查找表的高光谱图像无损压缩

为了提高高光谱遥感图像的压缩比，提出一种基于残差偏置和查找表的高光谱图像无损压缩方法。在高光谱图像的第一谱段图像采用了无损压缩标准中值预测器方法进行谱段内预测，其它谱段图像采用谱间预测方法。首先，在多级查找表（LAIS-LUT）预测方法的基础上搜索当前预测值，用当前预测值周围特定的5个像素点和当前像素值周围相同位置的5个像素点进行比较，通过比较结果，得出一个偏置值；然后在预测残差上加上偏置值；最后，将最终预测残差进行算术编码，并进行了理论分析和实验验证。结果表明，针对美国航空航天局的高光谱图像，所提出的方法比LAIS-LUT压缩比平均提高0.05；针对国内高光谱图像，该方法比LAIS-LUT压缩比平均提高0.07。这一结果对提高高光谱图像压缩效率是有帮助的。

Lossless compression of hyperspectral images based on lookup table and residual offset

In order to improve the compression ratio of the hyperspectral remote sensing images, a new lookup table(LUT) prediction method was proposed based on residual offset. In the first spectral band of the hyperspectral images, the prediction was conducted within the spectral band by the median prediction method of lossless compression standard. In other spectral bands, the prediction was conducted between the spectral bands. Firstly, the current prediction value was found through locally averaged interband scaling lookup table (LAIS-LUT) prediction method. Then, the specific five pixels around the current prediction value were compared with the corresponding five pixels around the current value. After the comparison, the offset was obtained. The offset was added to the prediction residual error. Finally, the prediction residual error will be coded with algorithm coding. Theoretical analysis and experimental verification show that the lossless compression ratio of the proposed method is increased by about 0.05 in National Aeronautics and Space Administration data and by about 0.07 in Chinese data. This result is helpful to improve the compression efficiency of hyperspectral images.