基于光滑0范数的图像分块压缩感知恢复算法

Image block compressive sensing reconstruction based on smooth L0 norm

结合维纳滤波器的光滑PL分块压缩感知恢复算法(BCS-SPL)可以去除图像分块压缩感知块效应,但是收敛速度慢,尤其是在低采样率条件下,为此将BCS-SPL算法与光滑0范数压缩感知算法(SL0)相结合提出一种基于光滑0范数分块压缩感知算法(BCSSSL0PL)。通过对不同图像对比BCS-SPL和BCS-SSL0PL算法的恢复效果和恢复时间、信号恢复迭代次数,可以证明BCS-SSL0PL在较低采样率条件下能以较少的迭代次数,较快的恢复时间获得与BCS-SPL相当的恢复效果,而且与BCS-SPL相比,BCS-SSL0PL对不同的采样率恢复时间变化不大,方便于不同场合的应用。对比两种算法恢复图像的细节,BCSSSL0PL算法还能改善低采样率条件下恢复图像的块效应。

The BCS-SPL algorithm, which combines the Wiener filter with PL compressive sensing signal recover algorithm, can overcome the blocking artifacts but its convergence speed is slow. To overcome the slow convergence speed of the BCS-SPL algorithm, especially on small measure rate, the BCS-SPL algorithm is combined with smooth L0 norm compressive sensing signal recovery algorithm, and a BCS-SSLOPL algorithm is proposed, which can speed up the image compressive sensing. The performance of the BCS-SSLOPL algorithm was compared with the BCS-SPL algorithm in terms of recover image quality, recover time and iteration times. Results show that the proposed BCS-SSLOPL algorithm surpasses the BCS-SPL algorithm that it can achieve the same recover quality with less time l it needs nearly the same time for different measure rate conditions. furthermore, it can improve the blocking artifacts on small measure rate conditions.