基于小波域马尔可夫随机场模型的压缩传感图像重构

目前在压缩传感重构算法中利用图像的可稀疏性表示先验知识,从比奈奎斯特采样少得多的观测值中恢复原始图像。除了稀疏性之外,邻域系数的相关性也可以作为先验知识加速重构算法收敛。为了克服目前算法中没有利用邻域系数相关性的缺点,提出了基于小波域马尔可夫随机场模型的压缩传感图像重构算法,根据显著性度量对变换系数进行分类得到具有马尔可夫性的初始掩模,利用ICM算法完成掩模优化,实现系数更新,并将算法与未考虑邻域相关性的算法进行了比较。实验结果证明了算法的有效性。

Image compressed sensing based on Markov random field model in wavelet domain

The current image compressed sensing algorithms can reconstruct the original image using the sparse prior of image from far fewer measurements than the Nyquist samples.However,the dependency of neighborhood coefficients is also a prior to accelerate the convergence of reconstruction algorithm besides the sparsity.To overcome the disadvantage that the current algorithms do not exploit the dependency of neighborhood coefficients,the reconstruction algorithm based on the Markov random field model in wavelet domain is proposed,the coefficients are classified according to the significant measurements to obtain the initial mask,and then the coefficients estimation is performed after optimizing the mask by ICM,and the proposed algorithm with the algorithm of no taking account of the neighborhood is compared.The results show the effectiveness of the proposed algorithm.