高光谱遥感影像包含丰富的空间、辐射以及光谱信息, 同时海量的数据也引发了高光谱成像技术在传输和存储方面的诸多问题。针对这一问题, 根据高光谱遥感影像谱间相关性强的特性, 提出了一种结合谱间多向预测的基于压缩感知的高光谱遥感影像重构方法。首先, 根据高光谱遥感影像的谱间相关性对高光谱遥感影像的波段进行分组, 每组确定一个参考波段, 使用平滑l0范数算法重构每组的参考波段。其次, 根据重构恢复的相邻组内的参考波段, 建立了一个非参考波段预测模型, 用来计算非参考波段的预测测量值; 然后, 计算实际测量值与预测测量值的差值, 使用SL0算法重构该差值得到差值向量; 最后, 利用得到的差值向量迭代更新预测测量值, 直到恢复该波段原始图像。仿真实验结果表明, 该方法提高了高光谱遥感影像的重构效果。

Hyperspectral remote sensing image contains rich space, radiation and spectral information. Meanwhile, massive data also lead to many problems in the transmission and storage of hyperspectral imaging technology. To solve this problem, a new method of hyperspectral remote sensing image reconstruction based on compressive sensing is proposed, which combines the spectral correlation and multi direction prediction. Firstly, according to the spectral correlation of hyperspectral remote sensing image, the bands of hyperspectral remote sensing image are grouped into each group, each group is defined as a reference band, and the I0 norm algorithm is used to reconstruct the reference band. Secondly, according to the reference band adjacent reconstructed within the group, a non reference band prediction model is set up, which is used to calculate the predicted measurement non reference band value. Then, we calculate the difference between the actual measured value and the predicted measured values, and reconstruct the difference vector difference using SL0 algorithm. Finally, by using difference vector iteration, we update the prediction measurements until the band recoreries the original image. The simulation results show that the proposed method improves the reconstruction effect of hyperspectral remote sensing image.