多方向几何非线性扩散图像去噪方法
2016-10-08刘国才王帅卿
刘国才 王帅卿



摘 要:现有几何非线性扩散滤波器同时在水平和垂直方向进行扩散,没有考虑图像边缘的方向性,有损其去噪效果.提出根据像素8邻域方向的梯度值检测边缘点及其边缘方向,对边缘点仅沿该边缘方向进行几何非线性扩散,而对非边缘点进行4个方向上的几何非线性扩散.5个等级噪声强度和5个典型加噪图像去噪实验结果表明,本文方法能够更有效地去除图像中的噪声,同时更好地保护图像边缘.
关键词:图像去噪;各向异性扩散滤波;非线性扩散滤波;高斯噪声;图像处理
中图分类号:TP391.4 文献标识码:A
Abstract:As the direction of edges in images is not taken into account, the original geometric nonlinear diffusion filter (GNLDF) is not in the best state for images denoising, which diffuses only in the horizontal and vertical directions. In this study, we detected the edges and their directions in images based on the magnitude of the gradient in the direction of the eight neighborhoods of each pixel, then conducted GNLDF at the edge pixel just along the directions of the edges while at any no edge pixel along all four directions. Experiment results of five typical images with five levels of noise intensity show that the proposed method can not only effectively remove the noise but also better protect the edges in images.
Key words:image denoising; anisotropic diffusion filter; nonlinear diffusion filtering; Gaussian noise;image processing
数字图像处理技术在人们日常中的应用越来越重要,而图像去噪是图像处理工作中一直都存在的基础性问题[1-2].随着技术的发展,人们对图像去噪技术有了更高的要求,传统的方法通常是采用线性滤波,如中值滤波、均值滤波等[3].但是这些方法通常都会导致图像细节的丢失,致使图像模糊,图像边缘[4]、纹理等高频信号被滤除.许多学者通过对图像内部像素及像素块之间相关性[5-6]的深入研究,逐渐认识到图像噪声之间的无关性和图像纹理之间的相关性.出于对图像边缘纹理[7]信息的需要,具有保留边缘[8]特性的去噪方法越来越得到重视.
自Perona等人[9]于1990年提出各向异性扩散滤波器(Anisotropic Diffusion Filter, ADF)以来,非线性滤波器得到了长足发展,提出了许多非线性滤波器,如全变差滤波器(Total Variation Filter, TV)[10]、双边滤波器(Bilateral Filter, BF)[11]、非局部均值滤波器(NonLocal Mean Filter, NLM)[12]等.由于各向异性扩散……
