Current Issue Cover
绘画特征提取方法与情感分析研究综述

贾春花,郭小英(山西大学软件学院)

摘 要
图像分类与情感分析是当前计算机视觉领域的研究热点,为人类绘画图像数字化研究提供了有效方法,在人类绘画艺术作品保护与作品创新中具有重要的应用价值。本文的工作:(1)分析了中西方绘画的不同表征方式及形成原因,为绘画图像特征提取与绘画情感分析提供了切入点,并概括了当前研究中常用绘画数据库;(2)概述了支持向量机、决策树、人工神经网络和深度学习、K最近邻分类等绘画图像分类中常用的机器学习方法,并分析了各类方法的优劣;以绘画图像的笔触特征、颜色特征、形状特征和纹理特征等为依据,详细综述了中西方绘画特征提取技术与分类方法的研究现状及发展;简要描述了绘画图像分类模型中常用的评价方法,并分析了当前研究中的常用评价指标;(3)从颜色特征出发,探讨了西方绘画情感分析的研究进展情况,为中国传统绘画情感分析提供了有效的思路;(4)提出了当前绘画分类和绘画情感研究中存在的问题和挑战,并探讨了存在问题的应对之策。在绘画图像分类的进一步研究中,特别是中国传统水墨画情感分析和绘画艺术创作方面,本文内容可以起到一定的启发和指导作用。
关键词
Reviews on methods of feature extraction and researches of affective analysis for the painting

jiachunhua,guoxiaoying(College of Engineering, Shanxi University)

Abstract
Image classification and affective analysis are the hot issues in the field of computer vision, which provide effective methods for the research of digital painting and play very important roles in the art protection and painting innovation for the human beings. Brush strokes, color, shape, texture, and white space are the very important visual features of the paintings, the classification based on these visual features can help us to better identify painting style and painter, analyze painting affection, further understand the meanings of painting creation and inherit cultural. The contributions of this paper are as follows: (1) The paper firstly shows the different representation modes of Chinese painting and western painting, and then analyzes why these two modes are different. The reasons mainly derive from the different cultural backgrounds and the different ways of thinking. These analyses offer a starting point for feature extraction and affective analysis of painting image; in addition, the paper sums up the painting database commonly used in the present study. Chinese traditional painting has its unique artistic expression technique, which makes the painting have the artistic effect of "false or true complement" by skillfully using white space. In addition, traditional Chinese painting attaches importance to the combination of calligraphy and poetry, decorated with a seal, and it emphasizes the connection between art and nature to present spirit by form. The line is the basic modeling way and the color is the auxiliary characteristic for Chinese traditional painting, which does not pay great attention to the bright and dark change of the light and shadow. Besides, the line is a form of expression for the traditional Chinese painting affective characteristics. For an appreciator, Chinese traditional painting needs more association and imagination, rather than visual effect. Chinese painting style mainly includes two major categories of traditional ink painting and mural, the formation of the ink and wash style is mainly distinguished by the representative painters, who are Qi baishi, Zheng banqiao, Xu beihong, Wu guanzhong, Wu changshuo, Huang gongwang, and so on; and mural research is represented by Chinese Dunhuang mogao grottoes murals with distinctive national style, which have rich and colorful content and the form of painting, they are expectations of people"s good wishes. Western painting is very different from traditional Chinese painting style. Traditional western painting highlights realism, emphasizing similar appearance, reproduction, space time and effect of light color. Western painting attaches great importance to the use of the change of color, light and shadow to show images, besides that, elaborately and tactfully use of color also can reflect the painting affection. Because of the shading of object, the whole painting has a better sense of texture and space. Contrary to the traditional Chinese painting, western painting gives the viewer more visual effects. Western painting style is associated with the development of literature and art movement, the formation of painting style is mainly divided into baroque style, three-dimensional style, impressionist style, romantic style, the rococo style, the Renaissance style, etc. (2) The paper outlines machine learning methods of the support vector machines, decision tree, artificial neural networks and deep learning, K nearest neighbor classification, which are commonly used in painting classification. The paper also analyzes the advantages and disadvantages of all kinds of methods. Besides, the paper reviews in detail the research status, the development of feature extraction technology and classification method for Chinese and western paintings from the view of the characteristics of the brushstrokes, color features, shape features and texture features and so on. This paper describes briefly the commonly used evaluation methods of the painting classification model, which are error rate and accuracy, precision and recall ratio, P-R curve and F1 measurement, ROC curve and AUC. Furthermore, some commonly used evaluation indexes in the current researches are analyzed. Computer vision technology has the distinct advantage in the fields of object recognition, scene classification, image classification, image affective and semantic analysis, which can make meaningful judgement for the perception target images and scenes by simulating human intelligent of vision. The essential feature of the image is the key of right judgement. As a kind of image resource, the selection of painting characteristics is very important for the painting classification and affective analysis. Machine learning is closely related with painting feature selection and classification research, which investigates how to use a computer to simulate or realize people"s learning activities, so that it can acquire new knowledge and skills. Machine learning is widely used in the field of each branch of artificial intelligence. (3) The paper probes into the affective investigation of western painting based on color features, and provides an efficient idea for the affective analysis of the Chinese traditional painting. Painting can reflect the objective social life and rich affection of the painter, using computer intelligence to analyze painting affection can help us better understand the history and culture of various periods. Chinese painting has its unique style, with the characteristics of the ink brush stroke, shade of ink, white space of the painting, painting content, and so on, which all can also convey different affections that the painter want to express. Combining the cognitive and psychological knowledge, methods of the image feature extraction can also realize affective analysis and affective extraction of the Chinese ink and wash painting. (4) Based on the views of the database of the painting classification, the classification goal and the limitation of affective analysis of the painting, the paper puts forward existing problems and challenges in the study of painting classification and painting affective, meanwhile discusses the solutions to the existing problems. In the further study of the painting classification, especially in the research of the traditional Chinese ink painting affective analysis and painting art creation, the content of this article can play a certain enlightening and guiding role.
Keywords
QQ在线


订阅号|日报