Abstract: Motion blur is a common phenomenon which is caused by the relative motion between the camera and scene during the capture processing.Obviously，the blurred images can reduce the reference value for people.In this paper，a modified adaptive non-convex total variation regularization model is proposed for restoration of motion blurred image based the local structure architectural features and orientation information measure.The proposed model can suppress the ringing effect effectively by using the global and local priori information of the target image.The experimental results show that the introduced model can deblur the blurry image with suppressing ringing effect effectively.And the peak signal to noise ratio，mean structural similarity and subjective visual effect of the denoised images are improved obviously.

Key words: image restoration, blur kernel, total variation regularization, mean structural similarity

摘要： 拍摄过程中的相对运动，导致获取图像存在一定程度的模糊，降低了其利用价值。在贝叶斯框架下，基于图像的局部结构特征和方向信息测度，提出了改进的自适应非凸全变分正则化图像复原模型，充分利用图像的全局和局部先验信息，有效抑制了复原图像中存在的振铃效应。实验结果表明，提出的改进模型在复原图像的同时能够保留图像的边缘轮廓等结构信息，得到的复原图像在峰值信噪比、平均结构相似度和主观视觉效果方面均有所提高。

关键词: 图像复原, 模糊核, 全变分正则化, 平均结构相似度