基于非局部总广义变分的图像去噪

针对全变分(TV)模型在去除图像噪声时容易产生阶梯效应的缺点,将二阶总广义变分(TGV)作为正则项应用于全变分模型中可以有效地去除阶梯效应,并且还能够更好地保持图像边缘纹理结构;利用非局部均值滤波算法的思想来构造非局部微分算子,将非局部微分算子应用于总广义变分模型中,综合提出了一种基于非局部总广义变分的图像去噪新模型。新模型充分利用了图像的全局信息进行去噪。实验结果显示了该模型的有效性和优越性。

An image denoising method based on nonlocal total generalized variation

The total variation model can remove noise effectively, however, it also brings in staircase effect. To overcome this shortcoming, we use the second order total generalized variation (TGV) as the regularization term in the new denoising model. The TGV model can not only eliminate the staircase effect, but also preserve structures such as edges and textures better. The nonlocal differential operators which are constructed based on the idea of the nonlocal means filtering algorithm are applied to the TGV model, and the new method makes good use of the global information of the image to remove noise. Experimental results demonstrate the validity and superiority of the proposed method.