基于剪切波变换的光场图像质量评价方法

光场成像作为新一代成像技术可采集更丰富的视觉信息,但在其各个处理过程中都会引入失真导致光场图像主观质量的下降,因此需要设计与人类视觉系统(Human Visual System,HVS)保持一致的客观评价方法以自动评价光场图像的质量。本文提出了一种基于剪切波变换的光场图像全参考质量评价方法。首先,为了同时考虑光场的空间域和角度域失真,分别在光场视图和极平面图(Epipolar Plane Image,EPI)上提取特征。其次,在光场视图上进行剪切波变换后,分别提取局部相似性特征和全局统计特征。然后,在EPI上分别提取局部相似性特征和全局方向统计特征。最后,采用随机森林对以上特征进行池化得到最终质量预测值。在密集光场图像数据库中进行测试的实验结果表明,所提出方法优于现有的图像质量评价指标,与主观视觉评价结果的一致性较高。

Light field image quality assessment with shearlet transform

As a new generation imaging technology,light field imaging can capture more abundant visual information,however,distortion will be introduced in each process of its processing,leading to the decline of subjective quality of light field image.Therefore,it is necessary to design an objective evaluation metric consistent with human visual system(HVS)to automatically evaluate the quality of light field image.In this paper,a full reference metric for evaluating the quality of light field image based on shearlet transform is proposed.Firstly,considering the spatial and angular distortions of the light field,the features are extracted from the light field view as well as the epipolar plane image(EPI).Secondly,after shearlet transformation on the light field view,local similarity features and global statistical features are extracted respectively.Then,local similarity features and global direction statistical features are extracted on the EPI.Finally,random forest is used to pool the above features.The experimental results on a dense light field image dataset show that the proposed metric is superior to the state-of-the-art quality evaluation metrics and has a high consistency with subjective visual evaluation results.