公开(公告)号：US20160321788A1

公开(公告)日：2016-11-03

申请号：US15022872

申请日：2015-02-10

Applicant: HUAZHONG UNIVERSITY OF SCIENCE AND TECHNOLOGY

Inventor： Tianxu ZHANG ,  Gang ZHOU ,  Ao ZHONG ,  Liangliang WANG ,  Ming LI ,  Cen LU ,  Wen ZHANG ,  Zhiyong ZUO

IPC: G06T5/00

CPC classification number: G06T5/003 ,  G06T5/00 ,  G06T2207/20004 ,  G06T2207/20012 ,  G06T2207/20056

Abstract: The invention discloses a direction-adaptive image deblurring method, comprising steps of: (1) defining a minimum cost function for deblurring an image by direction-adaptive total variation regularization; (2) converting the unconstrained minimization problem in step (1) to a constrained problem by auxiliary variables d1=Hu, d2=∇xu and d3=∇yu; (3) obtaining a new minimum cost function from the constrained problem in step (2) by introducing penalty terms; and (4) converting the minimization problem in step (3) to an alternating minimization problem about u, d1, d2 and d3, where a minimum of a variable is calculated as other variables are determined, and obtaining a deblurred image by solving the alternating minimization problem by an alternative and iterative minimization process. Compared with the prior art, the present invention obtains a new direction-adaptive cost function by introducing local direction information into a maximum a posteriori algorithm, solves a problem of edges of an image restored by traditional TV regularization terms being blurred, and can restore images of complex blurring types or images with abundant textures.

Abstract translation: 本发明公开了一种方向自适应图像去模糊方法，包括以下步骤：（1）通过方向自适应总变异规则化定义去除图像的最小成本函数; （2）通过辅助变量d1 = Hu，d2 =∇xu和d3 =∇yu将步骤（1）中的无约束最小化问题转换为约束问题; （3）通过引入惩罚条件，在步骤（2）从约束问题中获得新的最小成本函数; 和（4）将步骤（3）中的最小化问题转换为关于u，d1，d2和d3的交替最小化问题，其中确定变量的最小值作为其他变量，并且通过求解交替 通过替代和迭代最小化过程的最小化问题。 与现有技术相比，本发明通过将局部方向信息引入到最大后验算法中来获得新的方向自适应成本函数，解决了由传统TV正则化项被模糊而恢复的图像的边缘的问题，并且可以恢复图像 复杂的模糊类型或具有丰富纹理的图像。