基于改进优先权计算耦合色调相似性的图像修复算法  

Image Inpainting Algorithm Based on Improved Priority Calculation on Similarity of Coupled Hues

作　　者：范海峰[1] FAN Haifeng(Jilin Police College,Changchun 130000,China)

机构地区：[1]吉林警察学院信息工程系,吉林长春130000

出　　处：《洛阳理工学院学报（自然科学版）》2020年第3期64-68,共5页Journal of Luoyang Institute of Science and Technology：Natural Science Edition

摘　　要：为了准确获取优先修复块、提高算法的修复性能,设计了基于改进优先权计算耦合色调相似性的图像修复算法。在确定待修复块的优先级时,通过加入调节权值因子,对置信度项与数据项的权重进行调节,构造优先权测量函数,以测量待修复块的优先级,获取优先修复块。获取最优像素匹配块时,为了克服误差平方和(sum of squared differences,SSD)匹配准则,通过图像的R、G、B值计算其色调相似性,利用其来辅助SSD匹配准则加强邻域对比选出最优像素匹配块,从而减少错误匹配的出现,改善修复效果。在Criminisi算法更新置信度时,以图像块间的SSD值为依据,设计置信度更新模型,以增加置信度更新的合理性,提高算法的修复性能。不同算法的仿真结果显示,本文算法所得修复图像较为清晰且质量较好,算法的修复性能较为优秀。In order to identify priority restoration blocks and improve the accuracy of computing,this paper designs an image inpainting algorithm based on the improved priority computing of coupling tone similarity.Firstly,the priority measurement function is constructed by adding adjustment weight factors.Then,the R,G,B values of the image are used to calculate the hue similarity and select the optimal pixel matching block.Finally,when the Criminnisi algorithm updates the confidence,a confidence update model is designed to increase the rationality of confidence update and improve the inpainting performance.The results show that the image repaired by this algorithm has better visual effect,and the algorithm has better performance.

关 键 词：图像修复 Criminisi算法 优先权 色调相似性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]