欢迎来到天天文库
浏览记录
ID:51416648
大小:4.58 MB
页数:9页
时间:2020-03-23
《基于再模糊理论的无参考图像质量评价.pdf》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、第37卷第7期2016年7月仪器仪表学报ChineseJournalofScientificInstrumentV01.37No.7Jul.2016基于再模糊理论的无参考图像质量评价王红玉,冯筠,牛维,卜起荣,贺小伟(西北大学信息科学与技术学院西安710127)摘要:针对目前无参考图像质量评价算法依赖训练样本集、算法复杂性较高,难以应用于图像视频采集设备等问题,提出了一种基于有效再模糊的无参考图像质量评价算法(NRIQAVR)。首先研究了便携式设备采集图像的特点,然后在再模糊理论基础上通过计算最
2、小和最大再模糊标准差,构造了与人类视觉感知相近的NRIQAVR算法,通过理论证明该算法可以反映自然场景图像模糊度。最后在多个公开数据集和头盔摄像机采集的图像上测试,对比其他流行算法,该算法性能稳定且无需训练,评价结果与主观感知之间的相关系数平均值高于0.92,离出率低于0.05,平均执行时间0.088S。该算法符合人类视觉感知、不依赖图像内容、简单易行、可以达到ll帧/S的处理速度,具有重要实际应用价值。关键词:便携式设备;图像质量评价;视觉感知;再模糊理论;中图分类号:TP391THl64文献
3、标识码:A国家标准学科分类代码:520.6040No··referenceimagequalityassessmentbasedonre--blurtheoryWangHongyu,FengJun,NiuWei,BuQirong,HeXiaowei(CollegeofInformationScienceandTechnology,NorthwestUniversity,Xi’all710127,China)Abstract:Aimingattheproblemthatcurrentno—refe
4、renceimagequalityassessmentalgorithmsrelyontrainingsampleset,havehighcomputingcomplexityandaredifficulttobeappliedinportableimageacquisitionequipment.ANo—referenceimagequalityassessmentalgorithmbasedonvalidre-blur(NRIQAVR)isproposedinthispaper.Firsdy
5、,theimageacquisitioncharacteristicofportableequipmentisstudied.Then,onthebasisofre-blurtheory,throughcalculatingtheminimumandmaximunire—blurstandarddeviationsofvalidre-blur,theNRIQAVRalgorithmisconstructed,whichhasthefeatureofsimilartohumanvisualperc
6、eption.Itisprovedtheoreticallythatthealgorithmcallreflecttheblurextentinnaturalsceneimagesandhasgreatsimilaritywithvisualperception.Finally,experimentsonbothpublicdata-setandrealdata-setoftheimagesacquiredwithportablecamerawereconducted,andthepropose
7、dalgorithmWastested.Theresultsshowthatcomparedwithotherpopularalgorithms,theNRIQAVRalgorithmhasasteadyperformanceandrequiresnotraining.TheaveragecorrelationcoefficientbetweentheevaluationresultoftheNRIQAVRalgorithmandsubjectiveperceptionishigherthan0
8、.92.Theoutlierratioislessthan0.05andtheaverageexecutiontimeis0.088S.Theproposedalgorithmaccordswithhumanvisualperception,doesnotrelyontheimagecontent,issimpletobeimplementedandcallachievetheprocessingspeedof11矗amespersecond.TheproposedalgorithmCanbei
此文档下载收益归作者所有