资源描述：
《基于全局优化的图像块填充修复方法》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。
1、计算机研究与发展ISSN1000-1239PCN11-1777PTPJournalofComputerResearchandDevelopment46(1):144-150,2009基于全局优化的图像块填充修复方法陈中贵刘利刚王国瑾(浙江大学数学系杭州310027)(浙江大学CAD&CG国家重点实验室杭州310027)(ligangliu@zju.edu.cn)Exemplar-BasedImageCompletionUsingGlobalOptimizationChenZhonggui,LiuLigang,andWangGuojin(DepartmentofMa
2、thematics,ZhejiangUniversity,Hangzhou310027)(StateKeyLaboratoryofCAD&CG,ZhejiangUniversity,Hangzhou310027)AbstractImagecompletion,whichaimstoremoveobjectsorrecoverthedamagedportionsinagivenimage,isanimportanttaskinphotoediting.Recently,exemplar-basedmethodsareconsideredtocompleteimage
3、swithlargeportionsremoved.However,structureinconsistencyofthereconstructedtextureoftenappearwhenusingthosemethods.Inthispaper,anewexemplar-basedalgorithmisproposedtoobtainglobaltextureconsistencybyusingglobaloptimization.First,anenergyfunctionisdefinedformeasuringthequalityoftherecons
4、tructedregion.Then,theimagecompletionproblemisformulatedasminimizationoftheenergyfunctionwhichisdoneinaniterativeform.Finally,theslightcolordifferencesbetweentheknownregionandthefilledregionarerevisedbythePoissonimageeditingmethod.Comparedwiththeexistingexemplar-basedmethodswhichdogre
5、edyregion-growing,theproposedmethodnotonlyreconstructsthelocalcolortextureofmissingregion,butalsopreservestheglobalstructuraltextureoftheimage.Anadaptivesamplingmethod,whichisbasedonthesaliencymapoftheimage,isalsoadoptedtoconstructthesearchingspace.Itdramaticallyreducesthesearchingspa
6、ceandacceleratesthenearestneighborsearching.Theeffectivenessoftheproposedmethodisdemonstratedonseveralexamplesandcomparisons.Keywordsimagecompletion;objectremoval;texturesynthesis;globaloptimization;adaptivesampling摘要图像修复问题公式化为一个能量函数的最优化问题,设计了合理的能量函数度量修复结果的质量,并通过迭代方法得到最优解.全局优化方法不仅保持像素
7、局部颜色的连续性,而且也很好地保持了图像整体纹理结构在修复区域内的连续性.在构建像素邻域的搜索空间时,采用了自适应的采样方法,减小了搜索空间的大小,大大提高了搜索最佳匹配邻域的速度.实验结果表明,算法对大面积的图像缺失的修补和复杂背景图像上多余物体的去除,都达到了很好的效果.关键词图像修复;对象移除;纹理合成;全局优化;自适应采样中图法分类号TP391.41收稿日期:2007-04-22;修回日期:2008-07-10基金项目:国家自然科学基金委员会与微软亚洲研究院联合资助基金项目(60776799)陈中贵等:基于全局优化的图像块填充修复方法145抠除图像上的物体
8、,再将抠除
