基于压缩感知的红外遥感图像重构算法分析

摘要

With the rapid development of aerospace remote sensing technology,the performance of sampling and storage devices needs to be improved because of the longer sampling time and larger data of remote sensing image.To resolve the above problem,compressed sensing is introduced into meteorological satellite infrared remote sensing image processing.The performance of Orthogonal Matching Pursuit(OMP) algo-rithm,Gradient Projection for Sparse Reconstruction(GPSR) algorithm,Subspace Pursuit(SP) algorithm and Smoothed l0norm(SL0) algorithm are analyzed by Matlab modeling and simulation.A large number of infrared images are sampled and compressed at different sampling rates,then images are reconstructed with multiple algorithms.Comparative experiments show these algorithms can get the whole infrared image in the lower sampling frequencies,but smoothed l0norm algorithm is better than other algorithms in the accuracy of reconstruction and the runtime.It's proved that compressed sensing has great practical value.%随着航天遥感技术的飞速发展,遥感图像采集数据耗时长、图像数据量大等问题的出现对采样设备和存储设备提出了更高的性能要求.为了解决以上问题,在气象卫星的红外遥感图像的处理中采用了压缩感知理论.通过Matlab建模和仿真,分析了正交匹配追踪算法、梯度投影算法、子空间追踪算法、平滑l0范数算法的性能,并对大量红外图像以不同的采样率进行采样压缩,然后使用多种重构算法重构图像.对比试验显示,几种算法都能以较低的采样率得到完整的红外图像,但平滑l0范数算法在重构精度和运行时间方面都优于其余几种算法,证明了压缩感知在红外遥感图像的处理中具有较大的实用价值.