基于Curvelet稀疏和共轭梯度法的压缩传感图像重构

压缩传感以低采样率、抗干扰性强等特点备受关注。在图像能够稀疏表示的先验条件下,可以通过较少的随机投影,就能对原始图像进行精确重构。本文将Curvelet阈值收缩和共轭梯度相结合进行压缩传感重构,克服了正交小波方向选择性差,传统重构算法需要内存大、收敛速度慢、重构图像的细节与平滑不能兼备的缺点。实验结果表明,该算法提高了重构图像的峰值信噪比,加快了收敛速度,平衡了图像的细节与平滑成分。

Image compressed sensing reconstruction based on Curvelet-shrinkage and conjugate-gradient solution

Compressed sensing is famous for its low-sampling rate and stronger noise-resistance.In the prior condition that image have sparse representation,it can reconstruct the original image accurately from fewer measurements of random projection.Orthogonal wavelets have bad directional selectivity.Traditional reconstruction algorithms requires big memory,has slow convergence speed,and can＇t balance image details and smoothing components.Aiming at this problem,a reconstruction algorithm is represented that bases on sparse representation of the image in curvelet-shrinkage transform domain and conjugate-gradient.Experiment results show that the algorithm improves the peak signal-to-noise ratio,fasters convergence speed and balances the image details and smoothing component.