资源描述：
《An Information Fidelity Criterion for Image Quality一种图像质量信息保真度准则 自然场景统计评价》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、IEEETRANSACTIONSONIMAGEPROCESSING,VOL.14,NO.12,DECEMBER20052117AnInformationFidelityCriterionforImageQualityAssessmentUsingNaturalSceneStatisticsHamidRahimSheikh,Member,IEEE,AlanConradBovik,Fellow,IEEE,andGustavodeVeciana,SeniorMember,IEEEAbstract—Measurementofvisualqual
2、ityisoffundamentalim-orvideosthatcomefromasystemrunningunderagivencon-portancetonumerousimageandvideoprocessingapplications.figuration.TheobviouswayofmeasuringqualityistosolicitThegoalofqualityassessment(QA)researchistodesignalgo-theopinionofhumanobservers.However,suchsub
3、jectiveeval-rithmsthatcanautomaticallyassessthequalityofimagesorvideosuationsarenotonlycumbersomeandexpensive,buttheyalsoinaperceptuallyconsistentmanner.Traditionally,imageQAalgo-rithmsinterpretimagequalityasfidelityorsimilaritywitha“ref-cannotbeincorporatedintoautomatics
4、ystemsthatadjustthem-erence”or“perfect”imageinsomeperceptualspace.Such“full-selvesinreal-timebasedonthefeedbackofoutputquality.Thereferenc”QAmethodsattempttoachieveconsistencyinqualitygoalofqualityassessment(QA)researchis,therefore,todesignpredictionbymodelingsalientphys
5、iologicalandpsychovisualfea-algorithmsforobjectiveevaluationofqualityinawaythatisturesofthehumanvisualsystem(HVS),orbyarbitrarysignalfi-consistentwithsubjectivehumanevaluation.SuchQAmethodsdelitycriteria.Inthispaper,weapproachtheproblemofimageQAbyproposinganovelinformatio
6、nfidelitycriterionthatisbasedonwouldproveinvaluablefortesting,optimizing,bench-marking,naturalscenestatistics.QAsystemsareinvariablyinvolvedwithandmonitoringapplications.judgingthevisualqualityofimagesandvideosthataremeantforTraditionally,researchershavefocussedonmeasurin
7、gsignal“humanconsumption.”Researchershavedevelopedsophisticatedfidelityasameansofassessingvisualquality.Signalfidelitymodelstocapturethestatisticsofnaturalsignals,thatis,picturesismeasuredwithrespecttoareferencesignalthatisassumedandvideosofthevisualenvironment.Usingtheses
8、tatisticalmodelsinaninformation-theoreticsetting,wederiveanovelQAalgorithmtohave“perfect”quality.During
