基于深度学习的照片图像美感质量评估综述

Overview of Aesthetic Quality Evaluation of Photo Images based on Deep Learning

图像美感质量评估作为计算美学中重要的研究领域,是指利用计算机模拟人类的主观审美思维,并在此基础上对图像的美感进行定性或定量评估。作为图像美感质量评估中的一类主要研究对象,照片图像美感质量评估在检索与排序、照片图像美化等方面有着广泛的应用。本文主要对基于深度学习技术的照片图像美感质量评估研究进行归纳和总结,并从基于深度学习的美感评估这一基本思路出发,依次对照片图像美感质量评估类别、照片图像美感评估数据集及其建立方法进行综述,并对相关研究内容提出建议。

As an important research field in computational aesthetics, image aesthetic quality evaluation refers to the use of computer to simulate human subjective aesthetic thinking, and on this basis, to carry out a qualitative or quantitative evaluation of the image aesthetics. Photo image aesthetic quality evaluation, as one of the main research objects in image aesthetic quality evaluation, has been widely applied in retrieval and sorting, photo image beautification and so on. This paper mainly proposes to summarize the research on the aesthetic quality evaluation of photo images based on deep learning technology. Starting from the basic idea of the aesthetic evaluation based on deep learning, categories of photo image aesthetic quality evaluation, photo image aesthetic evaluation datasets and establishment methods are reviewed in turn, and suggestions for relevant research are put forward.