基于边缘和梯度分布先验模型的运动消模糊

Motion deblurring based on prior model of edge and gradient distribution

基于模糊核估计的运动模糊图像盲复原一直是图像处理中较难的问题,尤其是模糊核的精确估计和复原效果控制。本文采用模糊图像中的强边缘信息作为模糊核估计的依据,首先应用双边滤波和冲击滤波保持增强模糊图中的边缘信息,再利用梯度算子提取有效清晰的显著边缘并加以控制,最后对利用正则约束的方法求解到的模糊核进行修正和优化。在图像复原阶段,针对模糊核估计的偏差和复原过程产生的降质,提出一种基于引导滤波和自然梯度先验分布的图像约束复原算法。通过复原仿真及对比实验证明,该算法能复原出高质量的清晰图像,并能有效抑制振铃效应和图像噪声。

Fuzzy image restoration based on blur kernel estimation has been a difficult problem in the field of image processing, especially the accurate estimation of the blur kernel. This paper propose to consider the strong fuzzy image edge information as the basis for estimation of a blur kernel. Firstly, to hold and enhance the edge information in a blurring image, the proposed method utilizes the Bilateral filter and Shock filter. And the effective clear significant edge which should be control to become more effective, can be extracted by using gradient operator easily. Then to constraint and optimize the blur kernel which is obtained by solving the constraint regularization objective function is the most important prerequisite for recovery process. For the bad effect from the not very accurate blur kernel and processing of restoration,this paper propose a constrained restoration algorithm that use natural gradient prior distribution model and guide filter to relieve bad effect. Experimental results show that the algorithm can obtain a sharp image with high quality, and can effectively suppress the ringing effect and image noise.