基于局部相关性和偏微分方程的图像去噪

本文分析了各向异性偏微分方程在图像去噪中的特性，针对P—M扩散方程的缺点。定义了一种像素的局部区域相关性。根据局部区域相关性进行孤立噪声点检测。并用中值滤波的方法进行预处理，对噪声图像进行可变阈值参数的各向异性扩散。实验结果显示去噪后的图像从视觉效果和峰值信噪比上都要优于P—M扩散的结果。

Local correlation and PDE-based image denoising

The characteristic of anisotropic partial differential equation(PDE) is analyzed in image denoising.In view of the disadvantages of P-M diffusion equation,a Local-Correlation of pixel is defined.According to the Local-Correlation,the isolated noise is detected and median filtering method is used for preprocessing.Then an anisotropic diffusion with variable threshold parameter is implemented on noisy image.The experimental results show that the vision effect and peak signal-to-noise-ration(PSNR) of de-noised image is better than that of P-M's.