摘要: |
基于图像高频细节的重构问题,建立了全变分(Total Variation,TV)约束重建模型,选取交替方向乘子法(Alternating Direction Method of Multipliers,ADMM)求解分析。TVADMM重建算法能够保持图像边缘信息,纹理细节的刻画却不够理想,图像平滑部分的重建出现阶梯效应和过平滑现象。为此,提出使用分数阶约束的模型算法FOTVADMM求解。该算法对图像纹理高频细节重建效果较好,能锐化图像边缘区域,同时为降低经验调节参数对图像重建的影响,减少调节参数的时间,引入L曲线调节参数,找出了正则化参数最优解。实验结果表明,基于L曲线调参的FOTVADMM算法能够更好地保留图像的纹理和平滑部分的细节特征,在峰值信噪比和结构相似度评价指标上,FOTV对高频细节的重建改善效果更佳。 |
关键词: 超分辨率图像重建;分数阶全变分;交替方向乘子法;L曲线 自适应正则化调参 |
DOI: |
|
基金项目: |
|
Super-resolution image reconstruction based on L-curve parameters adjustment FOTV-ADMM |
XU Wen,YANG Xiaomei,XU Qiuyi,TIAN Qiaoyu,LIU Kai |
(School of Electrical and Electronic Information Engineering,Sichuan University Jinjiang College,Meishan 620860,China;College of Electrical Engineering,Sichuan University,Chengdu 610207,China;School of Information Engineering,Minzu University of China,Beijing 100081,China) |
Abstract: |
To solve the problem of reconstructing high-frequency details of images,a model with total variation(TV) regularization is established and solved by alternating direction method of multipliers(ADMM) algorithm.TV-ADMM reconstruction algorithm can keep image edge information,but the texture details of image is not ideally reconstructed and ladder effect and over smooth effect appear.A fractional order regularization model algorithm FOTV-ADMM is proposed to solve this problem.The algorithm has a good effect on high-frequency details of image texture and can sharpen the edge at the same time.In order to reduce the influence of empirical adjustment parameters on image reconstruction and the time of adjustment parameters,the L-curve parameters adjustment method is introduced to find the optimal solution of the regularization parameters.Experiment results show that FOTV-ADMM can better retain the texture and smooth details of the image,and adaptive FOTV-ADMM regularization has a better effect on high-frequency details in terms of peak signal-to-noise ratio and structural similarity. |
Key words: super-resolution image reconstruction fractional order total variation ADMM L-curve adaptive fractional regularization |