基于方向信息测度的非局部正则化图像去噪方法

针对非局部正则化在图像去噪过程中计算复杂度高、复原速度慢的问题,基于方向信息测度提出了改进的非局部正则化方法.在图像的边缘轮廓区域使用保边性能较好的非局部正则化方法,而在图像的平坦区域使用各向异性全变差模型,且该全变差模型由基于Bregman迭代正则化方法的快速迭代算法进行求解.实验结果表明:基于方向信息测度的非局部正则化方法在快速消除图像噪声的同时,能有效地保留图像的边缘和纹理等结构信息.

Modified non local regularization method for image denoising based on orientation information measure

Non local regularization method can achieve a state-of-art denoising result in comparison with conventional approaches.But the non local method has the disadvantages of high computing complexity and slow denoising rate.In this paper,we propose a modified non local regularization method based on orientation information measure.Furthermore,the target image is decomposed into two distinguished regions: edge regions and flat regions.Then we use the non local regularization method to denoise in the edge regions and use total variation regularization model with fast solving algorithm in the flat regions,while the total variation model is solved by a fast iterative algorithm which is constructed based on the Bregman iteration regularization.Numerical examples illustrate that the proposed method can remove noise with fast denoising rate and preserve the image edges,fine details and et al.