【作者】 刘洋；

【导师】 孙正兴；

【作者基本信息】 南京大学 ， 计算机技术（专业学位）， 2015， 硕士

【摘要】 随着数字化技术的快速发展,艺术和科技的结合日渐紧密,如何利用新技术将传统艺术的创作和生产过程数字化,进而推动传统艺术的发展成为当前计算机领域的研究热点。乱针绣作为传统刺绣艺术中的代表,有着独特的绣制技巧,善于刻画作品的纹理和色彩变化,尤其对油画、摄影作品具有突出的表现力,但其从业门槛高、生产效率低、价格昂贵的特点严重制约了自身的发展,研究如何利用计算机技术来辅助乱针绣的创作和生产具有重大的文化和经济意义。乱针绣的辅助创作本质上是完成由原始图像到绣稿再到绣品的转换过程。其中绣稿是对原始图像结构和内容的抽象,用于指导绣制过程的建模,绣品是对模拟绣制结果的仿真显示。如何有效地存储和管理这些图像信息是增强乱针绣辅助创作系统可用性的关键。本文根据视觉信息,对原始图像库进行层次化组织,并设计数据管理系统存储和管理图像数据,方便用户快速查找到目标图像信息。主要工作可以归纳以下三点:(1)图像库的特征表示。对于常常作为乱针绣创作素材的艺术画图像库,分别提取图像的全局特征和局部特征,并根据艺术画空间布局的特点设计空间匹配模型以度量图像在不同特征下的相似性,根据每个特征的聚类性能为图像库的层次化组织选择合适的特征。最后验证了人脸检测描述子对于艺术画图像的有效性。(2)图像库的层次化组织和浏览。提出了一种基于内容的图像层次化组织和展示方法,首先设计合适的MLAP聚类机制将图像库层次聚类构建浏览树结构,并利用t-SNE,MDS等技术对浏览树进行可视化展示供用户层次化浏览图像库,同时为用户提供浏览树的局部调整操作以弥补语义鸿沟和个体差异。(3)图像数据的存储和管理。根据数据间的相互关系设计数据库系统存储和管理图像数据,并设计信息收集机制,完善数据信息,以辅助用户按照语义标签检索数据库。更多还原

【Abstract】 Along with the rapid development of digital technology and a growing rapprochement between art and technology,it has become a research focus to study how to digitize the creation and production of traditional arts,which can promote the development of traditional arts.As a representative of traditional embroidery art,Random-Needle Embroidery has a unique embroidering shill,and it’s good at characterizing the variation of color and texture,especially for oil paintings and photography works.But the Random-Needle Embroidery is hard to learn,hard to manufacture,and expensive for customers,which seriously restrict its development.How to assist the creation and production using computer technology has major cultural and economic significance.The computer aided creation for random-needle embroidery is essentially to complete the process of turning the original image into embroidery draft and then turning the embroidery draft into embroidery.The embroidery draft is the abstract of the structure and content of the original image,which is used to guide the simulation of the embroidery.Embroidery is the digital display on the simulation result.How to effectively store and manage the image information is the key to enhance the usability of the computer aided random-needle embroidery system.In this paper,we Hierarchical organize the Original images according to visual information,and design database management system to store and manage the image data,in order to help users to quickly find the target image information.The main achievements of this paper are threefold as follows.(1)The feature representation of image dataset.We extract global features and local features for artistic image dataset which often used as the creation material for random-needle embroidery respectively.And then,based on the spatial layout characteristics of the artistic image,we design a pyramid spatial match model to measure the similarity with different features,so that we can choose an appropriate feature representation for artistic images.At last we verify the face detection descriptor on artistic images.(2)The hierarchical organizing and browsing of image dataset.We propose a content-based hierarchical organization and display method for image dataset.Firstly,a suitable MLAP clustering mechanism is designed to construct the hierarchical clustering of image dataset to build the browsing tree structure.Then we display the images of every level by the means of t-SNE,MDS.At last we provide the local adjustment operation for the users to make up the semantic gap and individual differences.(3)Storage and management of image data.According to the relationship between data,we design database system to store and manage image data,and design information gathering mechanism to perfect image data’s information,in order to assistant users to search the database according to sematic tag.更多还原