二阶TGV结合小波变换模的图像去噪算法

Image Denoising by Second- order Total Generalized Variation Combined with Wavelet Transform Modulus

研究了总变分(total variation,TV)模型在图像去噪中的应用,针对TV正则化模型在图像去噪过程中容易导致阶梯效应的缺陷,提出利用二阶总广义变分(total generalized variation,TGV)正则项代替TV正则项的图像去噪模型,并利用小波变换模极大值在检测图像边缘应用中的优点,在TGV正则化模型中引入以小波变换模为参数值的边缘检测函数,利用边缘检测函数引导扩散。该模型具备良好去噪和保持图像边缘的优点,还能缓解阶梯效应的产生。

The application of Total Variation model in image denoising was studied. This model has disadvantage of causing staircasing phenomenon. As for this defect,a second- order Total Generalized Variation regularization was proposed instead of Total Variation regularization in denoising model. The advantage of the application of the wavelet transform modulus maxima in edge detection was discussed. The edge detection function whose parameter is the wavelet transform modulus was introduced to the new image denoising model. The proposed model can effectively denoise as well as protect the edges and details of image. It can also ease the produce of staircase effect.