计算机工程与应用 ›› 2018, Vol. 54 ›› Issue (16): 169-175.DOI: 10.3778/j.issn.1002-8331.1704-0030

• 图形图像处理 • 上一篇    下一篇

基于兴趣区域的无参考图像质量评价方法

文继李1,丁立新1,万润泽1,邹桢苹2   

  1. 1.武汉大学 计算机学院,武汉 430072
    2.武汉大学 经济与管理学院,武汉 430072
  • 出版日期:2018-08-15 发布日期:2018-08-09

Non-reference image quality assessment method based on region of interest

WEN Jili1, DING Lixin1, WAN Runze1, ZOU Zhenping2   

  1. 1.School of Computer Science, Wuhan University, Wuhan 430072, China
    2.School of Economics and Management, Wuhan University, Wuhan 430072, China
  • Online:2018-08-15 Published:2018-08-09

摘要: 为了评价不同失真类型图像的质量,提出了一种基于兴趣区域和自然图像统计特性的无参考图像质量评价方法。该方法对Itti模型进行改进,并利用改进的Itti模型提取失真图像的感兴趣区域和非感兴趣区域,在非下采样Contourlet域提取图像的统计特性,通过计算失真图像的不同区域与自然图像统计特性的差异来获得图像的质量分数。在LIVE数据集上与已有方法进行对比,实验结果表明,提出的方法和主观感知具有较好的一致性。

关键词: 兴趣区域, 非下采样Contourlet变换, 自然场景统计特性(NSS)

Abstract: In order to evaluate the quality of images with different distortion types, this paper proposes a non-reference image quality assessment method based on Regions of Interests(ROI) and Natural Scene Statistics(NSS). Firstly, the existing Itti model is improved, and this improved Itti model is used to extract regions of both interests and non-interests from the distorted image. Secondly, the NSS of the image is extracted from the nonsubsampled contourlet domain. At last, the discrepancy between the NSS of the distorted image and that of the natural image is calculated as the quality score of the image. Comparing the proposed methods with existing methods on the LIVE dataset, experimental results show that the proposed method correlates well with human subjective perception.

Key words: Region of Interest(ROI), nonsubsampled Contourlet transform, Natural Scene Statistics(NSS)