融合全局与局部特征的缩放图像质量评价

基于内容感知的图像非等比例缩放方法旨在具有不同分辨率的显示设备上保持图像的重要内容，并改 善主观质量.图像缩放质量评价是衡量图像非等比例缩放效果的客观依据，现有客观评价方法不能有效地给出 合乎主观感受的评价结果，为此提出一种融合全局与局部特征的缩放图像质量评价方法.首先提出基于B样条 函数的图像全局结构形变度，计算图像缩放前后全局图像之间的形变量；其次结合主观视觉特点改进纵横比 相似度(AspectRatioSimilarity，ARS)使其更适合缩放图像的形变检测；最后线性融合这些度量特征得到客观的 缩放评价标准.在两种公开数据集上与多种代表性算法进行对比，实验结果表明，本文提出的图像缩放客观评 价方法与主观评价具有更好的相关性.
Content-awarebasedimageretargetingmethodsaimatkeepingtheimageondisplaydeviceswithdifferent resolutionsunchangedandimprovingthesubjectivequality.Imageretargetingqualityassessmentisusedtomeasurethe effectofimageretargetingresultsfromdifferentretargetingmethods.However,theexistingobjectiveassessmentmethods cannotprovideefficientassessmentwithsubjectivefeeling.Thus,aretargetingimageassessmentmethodbasedoncom? biningglobalandlocalfeaturesisproposed.Firstly,aglobalstructuraldeformationmethodbasedonB-splinefunction isproposed,itisobtainedbycalculatingtheshapevariablesbetweentheregionsbeforeandafterimageretargeted;Sec? ondly,itismadetobemoresuitablefordeformationdetectionbycombingthesubjectivevisualfeaturestoimprovethe AspectRatioSimilarity(ARS).Finally,thesefeaturesarefusedtoobtainobjectiveretargetingassessmentcriteria.Exper? imentalresultsshowthattheproposedobjectiveassessmentmethodofimageretargetinghasabettercorrelationwiththe subjectiveassessment.