基于邻域信息配准的计算机图像修复技术

打开文本图片集

中图分类号：G633.67 文献标志码：A 文章编号：2095-2945（2025）21-0167-04

Abstract：ImagesareimportantcariersofinformationandplayacrucialroleintheinformationsocietyInordertoachieve effectiveestorationofdamagedareasinimages，thispaperproposesanewneighborhoodbasedrestorationmethod.Thismethod firstidentifiesthedamagedareaintheimageandcomparesthebrightnessdiferencebetweenthedamagedareaandthe surroundingarea.Furthermore，suitableneighboringimagesubblocksareselectedandcomparedwithneighboring imagesub blocksaround’thedamagedareatodeterminetheoptimalimagesubblock.Afterthat，thepixelsinthedamagedareaare repairedstepbysteptorestorethecompleteimagearea.Toverifytheefectivenessofthismethod，repairexperimentswere conductedonnetworkvideoimages，imagesofwomenwearingheadscarves，imagesofbabonfaces，andimagesofboardwalks. Theresultsshowedthattheproposedrepair methodcanefectivelyrepairdamagedareasinimagesandhasrobustness inrepair performance for different types of images.

Keywords： image inpainting; single frame inpainting; dynamic inpainting; robustnes;theoretical model

人类进入信息时代以来，各种类型的信息呈现出爆炸式增长的态势。（剩余4389字）