基于模糊度量视觉特征的非局部均值去噪

针对非局部均值(non local mean,NLM)相似性度量不够准确的问题,提出一种基于模糊度量的视觉特征相似度的改进非局部均值图像去噪算法.利用模糊度量理论构建视觉特征度量相似性函数作为衡量图像像素点相似性;将平滑核函数代替高斯加权核函数,提高运算速度和避免滤波参数的设置;利用构建视觉特征相似性度量生成的平滑核函数,对图像进行去噪.由于改进方法考虑图像视觉结构特征,更加完善了非局部均值结构相似的特点.在高斯噪声和椒盐噪声下,用峰值信噪比(peak signal to noise ratio,PSNR)和结构相似性(structural similarity index,SSIM)评价指标分别对比分析提出方法与几种优秀的改进NLM方法的降噪性能.实验结果表明,改进的新方法在去噪性能方面得到较高的提升,同时降低了相似度计算的复杂度和减少了参数设置问题.

Improved non-local means image denoising algorithm using visual features based on fuzzy metric

Given the problem that the measurement of non local mean is not accurate enough, an improved NLM image de-noising algorithm is proposed, which uses visual features based on fuzzy metric as measure similarity. Measure similarity in-dex is constructed by using the theory of fuzzy to measure the similarity between the image pixels. Then, instead of Gauss weighted kernel function, smooth kernel function is constructed by similarity index of visual feature, which can improve the operation speed and avoid the filter parameters. Finally, NLM with the smoothing kernel function and the similarity measure of visual feature is used to denoise the image. Due to the fact that the improved method considers the characteristics of image visual structure, the structure similarity of NLM is further improved. The improved method in this paper is compared with sev-eral excellent improved NLM methods by peak signal to noise ratio(PSNR) and structural similarity index(SSIM) evaluation indexes under the Gaussian noise and salt&pepper noise. Experimental results show that the improved method has better per-formance in denoising, and reduces the complexity of the similarity computation and the problem of parameters setting.