双边非局部均值滤波图像去噪算法

摘要

In order to improve the visual effect of image denoising,according to natural images often contain more repetitive structure and bilateral filter have advantages in image denoising,present a new image denoising algorithm based on bilateral filtering and non-local means( NLM ). This method uses idea of NLM to estimate current pixel gray value. In image denoising process,not only impacts of current pixel gray value on result of prediction is taken into account,but also considered position relationship between current pixel and surrounding pixel. Build position coefficients in non-local neighborhood to constraint forecast results,considering homogeneity pixels similarity within the nonlocal neighborhood,design bilateral nonlocal mean filter. Experimental results show that running time of the algorithm is 0. 114 faster than bilateral filtering algorithm,peak signal-to-noise ratio (PSNR)is improved for 0. 9,image similarity(MSSIM)is increased for 0. 181,the image fidelity(VIF)is increased for 0. 2147. The proposed method can better maintain integrity of image information;improve image brightness and clarity of image texture.%为提高图像去噪的视觉效果，本文根据自然图像通常包含较多的重复性结构这一现象，以及双边滤波器的在图像去噪中所具有的优点，提出了一种新的基于双边滤波与非局部均值（ NLM）的图像去噪算法。利用NLM思想对当前的像素灰度值进行估计。过程中，不仅考虑到了当前像素的灰度值对预测结果的影响，而且考虑到了当前像素的位置与周围像素位置之间的关系，构建了非局部邻域内的位置系数来对预测结果进行约束，最后考虑到非局部邻域内同质像素的相似性，设计了双边NLM滤波器。实验结果表明：本文算法比双边滤波算法运行时间快了0.114 s、峰值信噪比（ PSNR）提高了0.9、图像相似度（ MSSIM）提高了0.181，图像保真度（ VIF）提高了0.2147。本文提出的方法能够更好地保留图片信息的完整性，提高了图像的亮度和图像纹理的清晰度。