基于表现手法的国画分类方法研究

发布时间：2018-05-26 21:16

本文选题：国画 + 工笔　；参考：《计算机学报》2017年12期

【摘要】：传统国画的数字化是信息时代对国画有效保存和有效分享的重要手段.海量的数字化国画已成为数字博物馆乃至现代图书馆重要的组成部分.如何实现对数字化国画的高效检索和管理已成为一个亟需解决的问题.针对该问题,文中探索研究基于表现手法的国画分类方法.根据表现手法的不同,国画可以分为工笔画和写意画两种基本形式.画作的对象内容不同及不同画家的画法、风格不同等因素对工笔画和写意画的计算机分类带来了重大挑战.通过对国画表现手法的研究及对大量国画的分析,文中首先提出一种国画关键区域检测方法,通过融合Scale Invariant Feature Transform(SIFT)特征检测子和边缘检测得到国画关键区域.然后,提出国画关键区域描述方法,通过对关键区域视觉特征及内部邻域差异性的描述,有效捕获工笔画和写意画在绘画技法上的不同.最后,提出一种级联分类策略,融合不同维度特征,实现对国画的分类.在文中所构建的包含1718幅古代、现代、当代国画大家作品的数据集上的实验结果表明,文中所提方法在工笔画和写意画分类上取得了很好的效果,分类正确率比现有最好方法提高了3.4%以上.
[Abstract]:The digitalization of traditional Chinese painting is an important means to effectively preserve and share traditional Chinese painting in the information age. Mass digital Chinese painting has become an important part of digital museum and even modern library. How to achieve the efficient retrieval and management of digital Chinese painting has become an urgent problem. Aiming at this problem, this paper explores and studies the classification method of traditional Chinese painting based on expression technique. According to the different performance techniques, traditional Chinese painting can be divided into two basic forms of fine brushwork and freehand brushwork. The difference of object content and painting method of different painters and different styles bring great challenge to the computer classification of fine brushwork and freehand brushwork. Based on the study of the expression techniques of traditional Chinese painting and the analysis of a large number of traditional Chinese paintings, this paper first puts forward a method to detect the key areas of traditional Chinese painting, and obtains the key areas of traditional Chinese painting through the fusion of Scale Invariant Feature transform and sift feature detection and edge detection. Then, the paper puts forward the description method of the key area of traditional Chinese painting. Through the description of the visual characteristics of the key area and the difference of the internal neighborhood, it can effectively capture the difference between the fine brushwork and freehand brushwork in the painting technique. Finally, a concatenated classification strategy is proposed to realize the classification of traditional Chinese painting. The experimental results on the data set of 1718 ancient, modern and contemporary Chinese paintings show that the method proposed in this paper has achieved good results in the classification of fine brushwork and freehand brushwork. The classification accuracy is more than 3.4% higher than the best method.
【作者单位】：北京大学信息科学技术学院;中国海洋大学信息科学与工程学院;
【分类号】：J212;TP391.41

【相似文献】