图像平滑的自适应保真全变分模型

提出一种自适应保真的全变分平滑模型，证明其解的稳定性。结合各向同性和全变分的优点，利用扩散系数构造保真项，从而增强图像边缘，该模型根据图像的梯度信息确定门限值进而选择合适的图像平滑方法。在去除噪声和保持边缘的同时避免了“阶梯”效应，实验结果表明，该模型能有效地去除噪声，提高图像的峰值信噪比。

Adaptive Fidelity Total Variation Model for Image Smoothing

An adaptive fidelity total variation image smoothing model is proposed. The proof of the stability of solution for this model is given. The method combines the merits of isotropic and total variation. A fidelity term constructed by the diffusion coefficient is used to enhance the edge of the image. The thresholds are introduced based on the gradient information of image, which helps the model to choose the appropriate smoothing scheme. This model removes noises highly and retains sharp edges without exhibiting staircasing. Experimental results show this model is succeessful in overcoming staircasing and efficiently removing noise, and the PSNR is high.