压缩感知的高分辨率天文图像去噪

High resolution astronomical image denoising based on compressed sensing

摘要:

为提高高分辨率天文图像的重构质量, 在传统压缩感知 (compressed sensing, CS) 迭代小波阈值算法的基础上, 提出了一种基于小波维纳滤波的压缩感知去噪重构算法.该算法的设计方法为：在每次迭代过程中, 使用设计的小波维纳滤波算子替代传统的小波阈值算子对获得的天文图像小波系数进行筛选, 从而对小波阈值去噪方法重建图像过程中出现的伪吉布斯现象进行有效地抑制；然后使用全变差方法对去噪重建后的天文图像进行调整, 以进一步提高重构图像的质量.仿真实验结果表明, 与传统的迭代小波阈值算法相比, 本算法可以获得较优的去噪重建性能, 并且能有效地保护高分辨率天文图像的细节特征信息.此外, 在压缩比较高的情况下, 该算法仍然可以获得相对较高的视觉质量和峰值信噪比.

Abstract:

To improve the quality of the reconstruction for high resolution astronomical image, a compressed sensing denoising and reconstruction algorithm, which combines wavelet with wiener filtering, is proposed based on the traditional compressed sensing (CS) iterative wavelet thresholding algorithm. The design method for this algorithm is that: a predesigned wavelet wiener filtering operator is used to replace the traditional wavelet threshold operator to select the wavelet coefficient of astronomical image in each iteration, thus the pseudo-gibbs phenomenon caused by the threshold denoising method in the reconstructed image can be suppressed effectively, and then the total variation method is used to adjust the reconstructed image for improving its quality. The experimental results show that the proposed algorithm can achieve better denoising and reconstruction performance, and can effectively protect the detailed feature information of high resolution astronomical image, compared with the traditional iterative wavelet thresholding algorithm. In addition, when the compression ratio is higher, the proposed algorithm can also help to the relatively higher visual quality and peak signal to noise ratio.