基于自适应滤波的BM3D降噪算法  

Image BM3D denoising method based on applying adaptive filtering

作　　者：崔程程 周先春[1,2,3] 昝明远 陈璟 汪志飞 殷豪 Cui Chengcheng;Zhou Xianchun;Zan Mingyuan;Chen Jing;Wang Zhifei;Yin Hao(School of Electronic and Information Engineering,Nanjing University of Information Science and Technology,Nanjing 210044,China;Jiangsu Collaborative Innovation Center of Atmospheric Environment and Equipment Technology,Nanjing 210044,China;School of Artificial Intelligence,Nanjing University of Information Science and Technology,Nanjing 210044,China;Changwang School of Honors,Nanjing University of Information Science and Technology,Nanjing 210044,China)

机构地区：[1]南京信息工程大学电子与信息工程学院,南京210044 [2]江苏省大气环境与装备技术协同创新中心,南京210044 [3]南京信息工程大学人工智能学院,南京210044 [4]南京信息工程大学长望学院,南京210044

出　　处：《电子测量技术》2021年第12期97-101,共5页Electronic Measurement Technology

基　　金：国家自然科学基金项目(11202106,61302188);江苏省“信息与通信工程”优势学科建设项目;江苏高校品牌专业建设工程项目;江苏省大学生创新创业训练计划项目(202010300128P)资助。

摘　　要：基于图像的降噪算法是构建智能视频监控系统的基础,对准确高速捕获事物具有重要意义。针对经典BM3D降噪算法中的硬阈值不能随噪声强度自适应且缺乏对图像边缘纹理信息保护的缺点,提出了一种基于自适应滤波的改进三维块匹配降噪算法。该方法首先在基础估计阶段使用自适应滤波代替硬阈值滤波进行相似块的匹配,即将软阈值应用于高噪声区域而将全变分应用于低噪声区域;然后在最终估计阶段使用K-means聚类方法寻找匹配块以得到最终去噪图像。实验结果表明,新算法将图像去噪质量平均提高了0.89 dB,图像结构相似度平均提高了1.05倍,同时也避免了传统算法造成的边缘振铃效应,有利于实际应用。Image based denoising algorithm is the basis of building intelligent video surveillance system, which is of great significance to capture things accurately. Considering that the hard threshold of BM3 D denoising algorithm cannot adapt to the noise intensity and lacks the protection of image edge texture information, an improved BM3 D image denoising algorithm based on adaptive filtering is proposed. Firstly, the adaptive filter is used to replace the hard threshold filter to match the similar blocks in the basic estimation stage. More accurately, soft threshold is applied to the high-noise region and total variation is applied to the low-noise region. Then in the final estimation stage, the K-means clustering method is used to find the matching blocks to obtain the final denoising image. The experimental results show that the new algorithm improves the PSNR of images by an average of 0.89 dB, and the SSIM of images by an average of 1.05 times. At the same time, it avoids the edge ringing effect caused by the traditional algorithm, which is beneficial to practical application.

关 键 词：三维块匹配 图像去噪 自适应阈值 K-MEANS聚类 全变分模型 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]