基于多种正则化的改进超分辨率重建算法

为了解决超分辨率图像重建过程中无法同时降低平滑区域噪声和保持图像细节的问题，结合改进的非局部变分（NLTV）和全变分（TV）正则项方法提出一种新的超分辨率重建算法。首先，根据图像重尾分布特性，结合高斯分布、拉普拉斯分布及柯西分布改进了传统NLTV正则项系数，提出了改进的ANLTV正则项。然后利用ANLTV正则项基于分裂Bregman算法重建了初始的高分辨率图像。最后结合TV正则项对重建的高分辨率图像进行去模糊操作，进而得到最终的超分辨率图像重建结果。为验证所提算法的性能，分别利用该算法与传统的TV和NLTV算法进行超分辨率图像重建并对比。实验结果表明，所提出的方法相比于传统的TV和NLTV重建算法，其峰值信噪比、信噪比和结构相似度均有所提高，能够同时满足超分辨率图像重建过程中抑制噪声和保持边缘细节的需求。

Improved image super resolution reconstruction method based on multiple regularizations

For better balancing the problems of noise-reducing and details-keeping in the process of the Super-Resolution（SR） image reconstruction, this paper proposes a novel multiple regularizations-based SR image reconstruction method, which combines an Amended Non-Local Total Variation（ANLTV） and Total Variation（TV） regularizations in the SR image reconstruction framework. Firstly, according to heavy-tailed distribution properties of the natural images, the ANLTV regularization is given by reformulating the weighting coefficients of the traditional Non-Local Total Variation（NLTV） with the combination of the Gaussian, Laplacian and Cauchy distributions. Accordingly, the initial SR image is then reconstructed using the split Bregman algorithm, from which, the final SR image is obtained by deblurring the initial reconstructed image with a TV regularization. In order to verify the performance of the proposed algorithm, the quantitative comparison between the proposed method and the traditional TV and NLTV based methods are carried out. The experimental results illustrate that the peak signal to noise ratio, signal to noise ratio and structural similarity obtained by the proposed method is obviously higher than the traditional ones, which is able to satisfy the requirement of reducing noise and preserving edge details in SR reconstruction simultaneously.