图像去噪的自适应非局部均值滤波方法

Adaptive Non-local Means Filtering Method for Image Denoising

提出一种自适应的非局部均值滤波算法.针对传统非局部均值滤波算法不能自适应地调节滤波参数的不足,本文统计和分析了不同图像的最优滤波参数与其小波系数能量的关系,并运用最小二乘拟合法建立了最佳滤波器参数值的预测函数,该函数可为待滤波图像选取合适的非局部均值滤波参数,进而实现了自适应的非局部均值滤波.与传统非局部均值滤波时需手动调节参数相比,本文的算法更加灵活.实验结果表明,对于具有不同内容或结构特性的图像,本文算法在峰值信噪比和主观去噪效果方面均优于传统的非局部均值滤波算法.

This study proposes an adaptive non-local means （ NLM ） filtering algorithm for image denoising. Conventional NLM method is not able to adaptively choose filter parameter. We thus analyze the relation between optimal filter parameters of different images and wavelet coefficients＇ energy. Subsequently, we build a prediction function employing least square fitting method, which is used to choose a suitable NLM parameter for the image to be filtered, so as to realize adaptive NLM filtering. Compared with conventional NLM method, our proposed method is more flexible. Experimental results illustrate that the proposed method is superior to conventional NLM in terms of both peak signal-to-noise ratio and subjective denoising quality for images with various contents or structural characteristics.