一种基于总体变分的自适应图像去噪方法

An Adaptive Image Noise Removal Method Based on Total Variation

建立了总体变分自适应图像去噪模型,并给出了其非线性各向异性扩散方程的数值解法。该方法采用高斯滤波器对噪声图像进行预处理,并利用图像每个像素的梯度信息,使其扩散方程在沿边缘方向上具有较大的扩散系数,而在垂直边缘的方向上具有较小的扩散系数。因此,总体变分自适应图像去噪方法不但能抑制噪声,还能很好地保持图像的边缘和纹理特征。图像去噪仿真实验表明,该方法的降噪效果明显优于总体变分去噪方法和中值滤波、维纳滤波等传统方法。

An adaptive image noise removal method based on total variation is illustrated and the numerical solution for the nonlinear anisotropic diffuse equation is provided. In this method, a Gaussian filter is used for noised images preprocessing and the gradient of every pixel is used in the diffuse equation, which has a bigger diffuse coefficient along the direction of image edges and a smaller one on vertical directions against image edges. Therefore, the adaptive image denoising method based on total variation achieves noise restriction and detail preservation simultaneously. Image denoising experiments are carried out and simulated results show that the performance of this method is obviously better than the total variation denoising method and traditional smooth methods such as median filtering, Wiener filtering and so on.