摘要
作为图像质量的监测和评价工具,图像质量评价(image quality assessment,IQA)在各种图像处理系统中发挥着重要的作用,理想的IQA方法应该与人类视觉系统(human visual system,HVS)相一致.目前HVS对图像的理解主要是依据图像的底层特征,本文提出了一种新的全参考(full reference,FR)彩色图像IQA方法.首先,提取了结构对比度指标(structural contrast index,SCI)、梯度、局部二值模式(local binary pattern,LBP)和色度四类底层特征图,用于刻画图像的不同特征属性;其次,利用不同的特征池化策略对每类特征分别处理,将其组成一组相似特征向量作为图像质量的检测器并采用极限学习机(extreme learning machine,ELM)建立回归模型,得到客观的质量分数;最后,与目前流行的8种FR IQA方法在5个标准IQA数据库上进行数值实验.结果表明,该方法整体性能优于其他方法,能够有效地提高大多数失真类型的预测精度.
As an image quality monitoring and evaluation tool,image quality assessment(IQA)plays an important role in various image processing systems.The ideal IQA method should be consistent with human visual system(HVS).Suppose HVS understanding an image mainly according to its low-level features,a novel Full Reference(FR)color IQA method.Firstly,four different types of low-level feature maps are extracted,namely structural contrast index(SCI),gradient,local binary pattern(LBP),and chroma,which are used to characterize different feature attributes of the image.Secondly,different feature pooling strategies are employed to process each type of features respectively,and a set of similar feature vectors are formed as the detector of image quality.Then,extreme learning machine(ELM)is used to establish regression model and map the feature vectors into an objective quality score.Finally,extensive experiments performed on five benchmark IQA databases and compared with eight state-of-the-art FR IQA metrics.The results demonstrate that the overall performance of proposed method is better than other methods,and can effectively improve the accuracy of IQA index on most of distortions.
作者
马月梅
付浩
刘国军
杨玲
魏立力
Ma Yuemei;Fu Hao;Liu Guojun;Yang Ling;Wei Lili(School of Mathematics and Statistics,Ningxia University,Yinchuan 750021,China)
出处
《南京师大学报（自然科学版）》
CAS
CSCD
北大核心
2022年第4期91-101,共11页
Journal of Nanjing Normal University(Natural Science Edition)
基金
国家自然科学基金项目(62061040)
宁夏区重点研发计划项目(2019BEG03056)
宁夏自然科学基金项目(2021AAC03039).
关键词
彩色图像质量评价
底层特征
局部二值模式
梯度
结构对比度指标
极限学习机
color image quality assessment
low-level features
local binary pattern
gradient
structural contrast index
extreme learning machine
