融合全局与局部特征的缩放图像质量评价

Image retargeting quality assessment combining global and local features

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(Aspect Ratio Similarity,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-aware based image retargeting methods aim at keeping the image on display devices with different resolutions unchanged and improving the subjective quality.Image retargeting quality assessment is used to measure the effect of image retargeting results from different retargeting methods.However,the existing objective assessment methods cannot provide efficient assessment with subjective feeling.Thus,a retargeting image assessment method based on combining global and local features is proposed.Firstly,a global structural deformation method based on B-spline function is proposed,it is obtained by calculating the shape variables between the regions before and after image retargeted;Secondly,it is made to be more suitable for deformation detection by combing the subjective visual features to improve the Aspect Ratio Similarity(ARS).Finally,these features are fused to obtain objective retargeting assessment criteria.Experimental results show that the proposed objective assessment method of image retargeting has a better correlation with the subjective assessment.