Image quality assessment via multiple features

Multimedia devices are indispensable in the information society. And, image quality highly impacts user experience of multimedia equipment. Therefore, measuring image quality accurately has great application value. The existing image quality assessment (IQA) methods have demonstrated the natural sense statistics and image structural information can measure the degradation of image. However, the generalization ability of individual IQA method is limited. In this paper, we propose a novel no-reference IQA method which is based on multiple features. For each image, we first extract natural sense statistic feature, global structural feature and local structural feature, respectively. Second, we train the quality prediction model via different features, and obtain different quality prediction scores by the models. Third, the prediction scores are collected and transformed to feature vectors. Subsequently, the IQA model is trained by support vector regression, and the input variables are the obtained feature vectors and subjective scores. The experimental results on the public databases demonstrate the proposed method can accurately predict the quality of both natural image and screen content image, and the performance is competitive with prevalent methods.