一种保持图像纹理的去噪方法

总变分去噪方法自提出以来得到了广泛的发展,该方法处理的数字图像收敛于一个分段光滑的函数,这导致图像纹理特征的大量丢失。将图像视为结构和纹理的组合,定义能够分辨纹理特征的局部方差,并利用定义的局部方差限制总变差保真项,进而在去除噪声的同时很好地保持了图像的纹理特征。针对各类图像进行了仿真试验,结果表明提出的方法是有效可行的。

A Denoising Method of Image Details Preserving

Total Variation noise removal method is developed extensively since presented.It modify images towards piecewise constant functions,so leading to lose a lot of details.We propose a mechanism that better preserves fine features in such denoising processes.This is accomplished by constraining fidelity item constructed by local variance.Our results show improvement in the signal-to-noise ratio over total variation method,and they are more appealing visually.