通用型无参考图像质量评价算法综述.pdfVIP

ComputerEngineeringandApplications$l算机工程与应用 2015，51(19) 13 通用型无参考图像质量评价算法综述 张淑芳，张 聪，张 涛，雷志春 ZHANG Shufang，ZHANGCong，ZHANG Tao，LEIZhichun 天津大学 电子信息工程学院，天津 300072 SchoolofElectronicInformationEngineering，TianjinUniversity,Tianjin300072，China ZHANG Shufang，ZHANG Cong，ZHANG Tao，eta1．Review onuniversalno—referenceimagequalityassessment algorithm．ComputerEngineeringandApplications，2015，51(19)：13—23． Abstract：Imagequalityassessmentcaneffectivelyevaluatedistortionordegradationcausedby imageacquisitionand transmissionprocess，whichhasabroadapplicationprospectinthefieldofdigitalmultimedia．Andbecauseofnoneed anypristineknowledgeofreferenceimages，no·referenceimagequalityassessmenthasbecomeanadvancedresearchhot— ． spotinthefieldofimagequalityassessment．Onthebasisofextensiveresearchofliteraturesathomeandabroad，inboth ofalgorithm principleandperformancecomparison，thispapersystematicallyintroducesseveralstate-of-the-artno-refer- enceIQAalgorithms，suchasBIQI，DIIVINE，BLIINDS，BLIINDS·II，BRISQUE，NIQEandGRNN．Firstly，themethods offeatureextractionandtheprincipleofqualityassessmentofeachalgorithm areintroduced．Secondly,thealgorithms abovearesimulatedandevaluatedontheLIVE imagedatabase，andtheperfomr anceandexecution speedofthealgo— rithmsareanalyzedandcompared．Atlast，thefurtherresearchtrendsofno-referenceimagequaliyt assessmentarepro— posed．AlthoughthesenO—referenceimagequalityassessmentsreviewedinthispaperhavesatisfactoryperfomr ance，their processesofevaluatingimagequalityheavilydependonopiniondataofimagequalityintheimagedatabase，andthere stillexistsomedeficienciesinevaluationperfomr anceandalgorithm complexity．Therefore，itisnecessary tomakefur- therstudyinthisfield． Keywords：imagequalityassessment；noreference；universal；featureextraction 摘 要 ：图像质量评价可有效评估图像采集和传输过程引起的失真或退化，在数字多媒体领域具有广阔的应用前景， 无参考图像质量评价算法由于不需要参考图像先验知识，近年来成为图像质量评价领域研究的热点。在对国内外文 献进行广泛调研的基础上，从评价算法原理和性能比较两个方面，系统综述了