基于块匹配改进和维纳滤波优化的BM3D降噪算法  

BM3D Noise Reduction Algorithm Based on Improved Block Matching and Wiener Filter Optimization

作　　者：曹阳 周先春[3] 王文艳 杨传兵 CAO Yang;ZHOU Xianchun;WANG Wenyan;YANG Chuanbing(School of Electronic and Information Engineering,Nanjing University of Information Science and Technology,Nanjing 210044;Jiangsu Collaborative Creation Center of Atmospheric Environment and Equipment Technology,Nanjing University of Information Science and Technology,Nanjing 210044;School of Artificial Intelligence,Nanjing University of Information Science and Technology,Nanjing 210044)

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]南京信息工程大学江苏省大气环境与装备技术协同创作中心,南京210044 [3]南京信息工程大学人工智能学院,南京210044

出　　处：《计算机与数字工程》2025年第3期811-814,839,共5页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:11202106,61302188)资助。

摘　　要：针对BM3D降噪算法中的块匹配过程受噪声影响较大且缺乏对图像边缘纹理信息保护的缺点,提出了一种基于块匹配改进和维纳滤波优化的BM3D降噪算法。该方法首先在基础估计阶段使用新的块相似性度量代替欧式距离进行相似块的匹配,新的块相似性度量考虑了观测图像中包含的噪声特征,以便在存在高强度噪声的情况下保证块匹配的准确性。然后在最终估计阶段通过最大化真实图像与估计图像之间的结构相似性(SSIM)来改进维纳滤波。达到了既要有效去噪,又要保护图像的边缘、纹理等细节信息的效果。实验结果表明,在同一实验条件下,新算法的峰值信噪比与BM3D算法相比平均提高了0.84 dB,图像结构相似度平均提高了1.08倍。A BM3D denoising algorithm based on improved block matching and Wiener filter optimization is proposed to overcome the shortcomings of the BM3D denoising algorithm,which is greatly affected by noise and lacks the protection of image edge texture information.Firstly,in the basic estimation stage,a new block similarity measure is used to replace the Euclidean distance to match similar blocks.The new block similarity measure takes into account the noise characteristics contained in the observed image,so as to ensure the accuracy of block matching in the presence of high-intensity noise.Then in the final estimation stage,Wiener filtering is improved by maximizing the structural similarity(SSIM)between the real image and the estimated image.It achieves the effect of not only effectively denoising,but also protecting the image edge,texture and other details.The experimental results show that under the same experimental conditions,the peak signal-to-noise ratio of the new algorithm is increased by 0.84 dB on average compared with BM3D algorithm,and the image structure similarity is increased by 1.08 times on average.

关 键 词：块匹配 BM3D 图像降噪 维纳滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]