对于带噪声的周期性纹理图像,提出一种基于二维秩约束的混合正则化去噪方法。该方法结合了全变分去噪理论和方法，并且利用该类图像低块秩的特性，对图像进行了低块秩约束。通过和全变分去噪方法比较可知，对于周期性纹理图像，混合正则化方法能有效地分离出噪声，并且能让图像很好地保持边缘。即使非严格的周期性纹理，该方法依然有很好的去噪效果。-For periodic texture images with noise, a new method based on two dimensional rank constraint is proposed. This method combines the theory and method of total variation denoising, and uses the characteristic of the low rank of the image to carry on the low rank constraint. By comparing with the total variation denoising method, we can know that the mixed regularization method can effectively separate the noise　the periodic texture image, and the image can be well maintained. Even if the non strict periodic texture, this method still has the very good denoising effect.