论文部分内容阅读
压缩感知理论提供了一种新的数据获取和压缩思路,能有效地把计算负担从编码端转移到解码端。高光谱数据具备数据稀疏性、空间相关性和谱间相关性,结合这3类先验知识,提出了一种基于复合正则化的高光谱图像压缩感知投影与重构方法。该方法的编码端只需要一个简单的投影操作;在重构算法实现中,基于变量分裂的思想,把具备多个正则项的优化问题转化成多个简单的优化问题,并用迭代方式求解。实验结果表明,本文算法在高光谱图像重构上能获得更高的峰值信噪比和更好的重构效果。该方法具备极低的编码复杂度,适用于资源受限的机载和星载高光谱成像平台。
Compressed sensing theory provides a new idea of data acquisition and compression, which can effectively transfer computational burden from encoder to decoder. Hyperspectral data possesses data sparsity, spatial correlation and spectral correlation. Combined with these three types of prior knowledge, a hyperspectral image compression perceived projection and reconstruction method based on complex regularization is proposed. In the implementation of reconstruction algorithm, based on the idea of variable splitting, the optimization problem with multiple regular items is transformed into several simple optimization problems and solved by iterative method. The experimental results show that the proposed algorithm can achieve higher peak signal-noise ratio and better reconstruction effect in hyperspectral image reconstruction. The method has very low coding complexity and is suitable for airborne and on-board hyperspectral imaging platforms with limited resources.