Image super-resolution representation via image patches based on extreme learning machine

《Image super-resolution representation via image patches based on extreme learning machine》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、InternationalConferenceonSoftwareEngineeringandComputerScience(ICSECS2013)Imagesuper-resolutionrepresentationviaimagepatchesbasedonextremelearningmachineQiuxiZhu,XiaodongLi,WeijieMaoDepartmentofControlScienceandEngineeringZhejiangUniversityHangzhou,Chinae-

2、mail:zhuqiuxi0743@126.comAbstract—Inthispaper,aimedattheextensivelyexistingproblemandmakeitpossibletoincreasetheresolutionofLRimagesbyofslownessinmainstreamimagesuper-resolutions,anefficientmorethan3or4times[2].However,mostoftheappliedwell-approachispropos

3、edforsuper-resolutionbasedontheextremedevelopedlearningalgorithmssuchasbackpropagation(BP)learningmachine(ELM)forsingle-hiddenlayerfeedforwardandsupportvectormachine(SVM),arestilltrappedbytheneuralnetworks(SLFNs).Featuresandissues(e.g.parameterbottleneckof

4、slownesscausedbyiterativesolutionsandthusselections)intheapplicationofELMarediscussed,onthebasisfailtobeacceptedinfieldsthatlaystressonthespeedofofwhichageneralframeworkforavarietyofsuper-resolutionimaging[2].problemsisproposed,andcorrespondingexperimentsa

5、reInthispaper,extremelearningmachine(ELM)[10]forconducted.Itisshownintheresultsthattheproposedapproachsinglehidden-layerfeedforwardneuralnetworks(SLFNs)iscanachieverelativelygoodqualityandmuchfasterspeedappliedtoimagesuper-resolutioninordertoprovidelearnin

6、g-comparedtotraditionalreconstruction-basedsuper-resolutions,basedsuper-resolutionswithafastandefficientlearningthereforetheeffectivenessofthismethodisdemonstrated.algorithm.TheproposedmethodmainlyfocusesonprovidingKeywords-ELM;neuralnetwork;imageprocessin

7、g;super-thesuper-resolutionproblemwithageneralframeworkresolution(includingtheconfigurationsoftheneuralnetwork)thatcanbeappliedtodifferenttypesofsuper-resolutionproblemswithfewmodificationsneededintransplantations.BytrainingtheI.INTRODUCTIONSLFNwithrelativ

8、etrainingimages,super-resolutionimagesImagesuper-resolutionisasignificantbranchinimagecanberebuiltbytheintegrationofnetworkoutputs.fusion,whichrebuildshigh-resolution(HR)imagesbyutilizingTheremainingpartsofth