10th EAI International Conference on Mobile Multimedia Communications

Research Article

A No-reference IQA Metric Based on BEMD and Riesz Tansform

Download12 downloads
  • @INPROCEEDINGS{10.4108/eai.13-7-2017.2269980,
        author={Guangyi Yang and Yue Liao and Qingyi Zhang and Ye Zhi and Jiahua Cao and Deshi Li},
        title={A No-reference IQA Metric Based on BEMD and Riesz Tansform},
        proceedings={10th EAI International Conference on Mobile Multimedia Communications},
        publisher={EAI},
        proceedings_a={MOBIMEDIA},
        year={2017},
        month={12},
        keywords={no-reference image quality assessment bidimensional empirical mode decomposition riesz transform human visual system},
        doi={10.4108/eai.13-7-2017.2269980}
    }
    
  • Guangyi Yang
    Yue Liao
    Qingyi Zhang
    Ye Zhi
    Jiahua Cao
    Deshi Li
    Year: 2017
    A No-reference IQA Metric Based on BEMD and Riesz Tansform
    MOBIMEDIA
    EAI
    DOI: 10.4108/eai.13-7-2017.2269980
Guangyi Yang1, Yue Liao1, Qingyi Zhang1, Ye Zhi1, Jiahua Cao1, Deshi Li1,*
  • 1: School of Electroinc Information, Wuhan University, Wuhan, Hubei, China, 430072
*Contact email: dsli@whu.edu.cn

Abstract

In recent years, the research of the no-reference (NR) image quality assessment (IQA) has been very active. But the NR-IQA research for noise-distorted images hasn't been proposed yet. In this paper, we propose a NR-IQA metric for noise-distorted images based on bidimensional empirical mode decomposition (BEMD) and Riesz transform, namely BRIQ index. Firstly, the denoised image is obtained by using the BEMD, Riesz transform and the visual contrast sensitivity function (CSF) to denoise the distorted image. Then the similarity calculation is performed using five Riesz transform feature maps of the distorted image and the denoised image. Finally, we combine these five similarity indices to obtain the final quality score for the distorted image. Experimental results on pubic database show that the BRIQ index evaluates the noise-distorted images with good performance, as consistent as subjective quality ratings.