图像质量评价是图像处理领域的研究热点。该文综合论述了图像质量的主观和客观评价方法,重点阐述了单视点图像质量的客观评价方法。对目前比较常用的峰值信噪比和均方误差全参考评价算法进行了分析并指出其存在的问题。然后,对基于误差敏感度和基于结构相似度的评价算法进行了论述和分析,并对质降和无参考评价方法进行了综述。根据视点的个数,图像质量评价可分为对传统单视点图像和立体图像的评价。该文还对立体图像质量评价算法进行了分析讨论。最后,就图像质量评价算法的进一步发展提出了若干技术与研究方向的展望。
Image Quality Assessment (IQA) is a hot research area in the field of image processing. In this paper, objective and subjective IQA methods are reviewed, and more attention is paid to the former. PSNR and MSE, which are commonly used to assess the quality, are analyzed in detail and their defects are given. The models based on error sensitivity and structure distortion of images are two critical methods in IQA, and the survey presents their key techniques and challenge problems. The reduced reference and no reference methods are also presented in this survey. Based on the number of view, IQA are classified into two major categories, namely, monoscopic image IQA and stereoscopic image IQA. This survey also makes an introduction of the stereoscopic image IQA. Finally, the survey lists several perspective sub-fields and topics in IQA progress.