基于压缩感知的遥感成像稀疏重构性能分析

Analysis on Sparse Reconstruction Performance of Remote Sensing Imaging Based on Compressive Sensing

摘要:

压缩感知是一种新型的信息论，打破了传统的Shannon-Nyquist采样定理，能够以少量数据完成信号采样。稀疏重构是压缩感知由理论到实际的关键环节，为了将压缩感知有效地应用于遥感成像领域，研究了稀疏重构对遥感成像过程的影响。针对稀疏重构理论模型，分析了重构误差的成因；同时，针对典型的凸优化类算法和贪婪类算法，利用峰值信噪比指标对遥感图像重构误差进行评价。在仿真实验中，定量考察遥感图像在不同压缩采样率、不同重构算法下的稀疏重构性能。结果表明，稀疏重构算法能够成功重构遥感图像，各算法在不同压缩采样率下均表现出了较好的重构质量，整体上能够满足遥感成像应用，验证了压缩感知稀疏重构方法在遥感成像中应用的可行性。

Abstract:

Compressive sensing is a new information theory which breaks the traditional Shannon-Nyquist sampling theorem and can perform signal sampling with a small amount of data. Sparse reconstruction is the key factor of compressive sensing from theory to practice. In order to apply compressive sensing effectively to remote sensing imaging, the effect of sparse reconstruction on remote sensing imaging is studied. Based on the sparse reconstruction model, the causes of reconstruction error are analyzed. Meanwhile, according to the typical convex optimization algorithms and greedy algorithms, the reconstruction errors of remote sensing image are evaluated by Peak Signal-to-Noise Ratio (PSNR). In the simulation, the sparse reconstruction performance of remote sensing image is quantitatively investigated with regard to different compression sampling rates and reconstruction algorithms. The result shows that sparse reconstruction algorithm can successfully reconstruct remote sensing image. The algorithms give good reconstruction quality with different compression sampling rates, which can meet the requirements of remote sensing imaging. The conclusion proves the feasibility of applying compressive sensing sparse reconstruction method in remote sensing imaging.