图像美学质量评价技术发展趋势

 图像质量评价是利用计算机模拟人类视觉系统自动评价图像的失真程度，而图像美学质量评价是利用计算机模拟人类对美的感知与认知，自动评价图像的"美感"，分析图像在构图、颜色、光影、景深、虚实等美学因素影响下所形成的美感刺激，这是计算美学与计算机视觉、心理学、虚拟现实等领域的学科交叉新方向，其在摄影摄像、电影电视、电子商务、服装设计、平面设计、文稿设计、工业设计、汽车行业、建筑行业、美容美妆等多个行业具有良好的应用前景。本文回顾了图像美学质量评价的发展历史，从基于人工设计美学特征的方法、基于美学特征深度学习的方法、图像美学质量评价的新任务、图像美学质量评价数据集构建等方面综述了该领域关键技术的发展情况，并展望了图像美学质量评价及其关键技术的发展趋势。

Image aesthetic quality assessment: A survey

Image quality assessment (IQA) is to measure the perceived image degradation while image aesthetic quality assessment (IAQA) is to automatically give a review of aesthetics of an image. The composition, color, illumination, depth of field, blurness, sharpness, etc. are analyzed by means of computer to analog the aesthetic perception and cognition of human beings. IAQA is a novel interdisciplinary field of computer aesthetics and computer vision, psychology, and virtual reality. IAQA can be used for photography, movie and television, ecommerce, fashion design, graphic design, document typesetting, industry design, automobile industry, architectural design, beauty makeup, etc. In this article, we give a brief history of IAQA. Then we review the methods based on hand-crafted aesthetic feature design, the methods based on deep aesthetic feature learning, new topics in IAQA, and dataset collection for IAQA. Finally, we give the future directions of IAQA.