基于光滑0范数压缩感知的多光谱图像去马赛克算法

提出了一种基于压缩感知(CS)的多光谱滤波阵列(MSFA)的多光谱图像去马赛克算法(DMA)。 首先,通过将MSFA采样得 到马赛克图像的过程等效为CS理论中的感知矩阵采样的过程,并充分利用多光谱图 像的空间和谱间 相关性,通过在三维空间傅里叶基上对多光谱图像进行稀疏表示；然后由随机MSFA模式和CS 理论构造的测量矩阵对多光谱图像进行观测投影,最后采用CS重构算法求解0范 数下的最优化问 题,从而得到多光谱图像的稀疏表示系数。给出对算法性能的评估数据和Matlab仿真 图片。实验结果证明,本文算法的峰值信噪比(PSNR)值高于克罗内克CS(KCS)和组稀疏(GS)两种算法,且有效地减少了上述两种算法中出现的模糊现 象,改善了图像的视觉效果。

Multi-spectral demosaicking method based on compressive sensing of smooth norm 0

A multi-spectral demosaicking algorithm (DMA) based on compressive sensing (CS) is proposed.First,the process of obtaining mosaic image by multi-spectral filter array (MSFA) is equivalent to the process of sensing matrix sampling in the CS theory,and making use of sparse representa tion of multi-spectral images is obtained in 3dimensional spatialand spectral domain in spatial and spectral correlation multi-spectral image.Finally,the CS signal reconstruction alg orithm is used to solve the optimization problem of the 0norm,and the sparse representation coefficients o f the multi-spectral images are obtained.The evaluation data for the algorithm performance and the simulation images of MATLAB are given.Experimental results s how that the peak signal to noise ratio (PSNR) of algorithm in this paper is higher than those of the Kronecker CS and group-spar se (GS) algorithms, and the proposed algorithm effectively reduces the in zipper effect above two al gorithms and improves the image visual effect.