摘要
以全变分去噪模型为例,从梯度下降法解相应的欧拉方程着手,提出一种新的定参方法对图像进行去噪,根据新方法选取的参数,同时保证均值和方差估计式。且求解偏微分方程选取的初值不是噪声图像而是对噪声图像进行小波分解后,保留低频系数,只对高频系数设置阈值,再重构后的图像。根据新选取的初值对相应的偏微分方程进行差分迭代求解,数值仿真结果表明,该方法选取的初值具有更好的去噪效果。
This paper takes the example of total variation denoising model to propose a newly- selected parameter method when solving the corresponding Euler equation by the gradient method in denoising image,and realizes the mean and the variance estimate formula simultaneously by adopting this method. While ones solve PDE,ones do not select the noise image but the reconstructed image.as the initial value, under the wavelet decomposition of noise image,keeping the low frequency coefficients and only setting up a threshold for high frequency coefficients. With the newly- selected initial value,PDE is solved by the finite difference iteration and the equation is simulated numerically. The experiment results show that this method has a better denoising effect.
出处
《现代电子技术》
2008年第19期180-183,共4页
Modern Electronics Technique
关键词
变分
去噪
初值
阚值
小波分解
total variation
denoising
initial value
threshold
wavelet decomposition