基于最优导向法则与距离约束的图像修复算法  

Image inpainting method based on optimal guidance rule coupled with distance constraint

作　　者：蔡鹏飞 段朝伟 

机构地区：[1]河南工学院计算机科学与技术学院 [2]南京航空航天大学自动化学院

出　　处：《电子测量与仪器学报》2018年第10期119-125,共7页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(61401150,61472119);河南省高校科技创新人才支持计划(16HJSTIT040)资助项目

摘　　要：为了解决当前图像修复算法利用置信度与数据项来完成图像修复时,忽略了优先修复块中已知信息量所占的比例,导致修复图像存在不连续以及块效应的不足,提出了一种基于最优导向法则耦合距离约束因子的图像修复算法。首先,将像素点的梯度信息引入到待修复块的优先权中,联合置信度与数据项,构造了优先权判定函数,从破损区域中选取优先修复块。以优先修复块中已知信息量所占的比例为依据,构造最优导向法则,对优先修复块中已知信息所占比例进行调整,以找出最佳的匹配块。然后,计算像素点的梯度信息,建立梯度直方图,确定待修复像素点的主方向,通过待修复块内已知像素点与待修复像素点的距离构造主方向上的距离约束因子,以对样本块大小进行动态调整。最后,在像素点之间棋盘距离的约束下,通过对像素点进行均方误差和度量,搜索最优匹配块,从而完成图像修复。实验结果与分析显示,与当前图像修复算法相比,所提算法具有更高的修复视觉质量。In order to solve the defects as discontinuity and block effect of the repaired image in current image inpainting algorithm, which induced by neglecting the proportion of the known information in the priority repair block. a novel image inpainting method based on optimal guidance rule coupled with distance constraint is proposed in this paper. Firstly, the gradient information of pixels is introduced into the priority measurement of repaired blocks, and the priority decision function is constructed with confidence items and data items to select priority repair blocks. The proportion of the known information in the priority repair block is based on the optimal guidance law, and the proportion of the known information in the priority repair block is optimized. Then, by calculating the gradient information of pixels of the gradient histogram to determine the main direction of the inpainting pixel, the distance constraint factor in the main direction was constructed based on the distance between the known pixels and repaired pixels to dynamically adjust the sample size. Finally, under the constraint of the chessboard distance between pixels, the image restoration is completed by using the methods of the mean square error and measurement of the pixels to search the optimal matching block. The simulation experiment results and analysis show that compared with the current image restoration algorithm, the proposed algorithm has better visual effect.

关 键 词：图像修复 梯度信息 优先权判定函数 最优导向法则 距离约束因子 均方误差和度量 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]