基于內(nèi)容的聚類及其在圖像編輯中的應(yīng)用.pdf

1、浙江大學(xué)碩士學(xué)位論文基于內(nèi)容的聚類及其在圖像編輯中的應(yīng)用姓名：黃海斌申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：基礎(chǔ)數(shù)學(xué)指導(dǎo)教師：孫業(yè)順；劉利剛201105浙江大學(xué)碩士學(xué)位論文摘要AbstractInthispaperwepresentanewapproachforcontextawareimageeditingmethodinimageinpaintingandtexturereplacementWeextractdifferenttexturesthr

2、oughacontentawareclusteringinimagesAftergettingtheclustersofdifferenttextures，thecontextawareimageinpaintingandtexturereplacementprocessingisnaturallyfollowedComparedwiththetradition。almethods，themethodweproposeinthepape

3、rtakesuseofthehigll—levelinformationinimageswithrepeattexturesWefirstmakeade印analysisofthecontext—awareclusteringalgorithmDifferentfromtraditionalclusteringalgorithm，thecontextawareclusteringtakesbothdistancebe—tweendata

4、pointsaswellastheneighborhoodinformationintoconsiderationTheneighborhoodinformationselvesasaregularizationtermtobalancethedatafideli哆inoriginalfeaturespaceandtheinfluenceofcontextualpatternsWewillshowthatthisclusteringal

5、gorithmCansimulatecomplexdatadistributionwithlIighcrossoverWethenapplyittOgenerateanewimageeditingmethodTotheimagewithmultipleandrepeattextures，wefirstdetectitsfeaturepointswithSIflralgorithmandclustertheminthefeaturespa

6、cewiththecontent—awareclusteringThenitiseasytOimplementtheimageinpaintingandtexturereplacingThereareseveraladvantagesforournewimageeditingmethodFirstofa11wemakegooduseoftheinformationcontentedintheimagesNotlikethetraditi

7、onalmethods，ourmethodtakestheinternlogicalrelationshipbetweendifferenttexturesratherthansimplysearchingforthepatchwithlocalmatchingSecondlythetexturereplacementeditinginourpaperisbasedonclusteringresultsofrepeattexturein

8、theglobalfieldWhattheuserneedstodoisjUStreplaceonetextureandtheotherswillbereplacedatthesametimeItismoreconveniencefortheuserwhichhelpthemavoidoperatingonebyonelocallyKeywords：ImageEditingContextawaredustering，Imageinpai