线条画的提取与风格转换方法研究

发布时间：2018-03-11 16:49

本文选题：非真实感绘制　切入点：形态学细化　出处：《曲阜师范大学》2008年硕士论文　论文类型：学位论文

【摘要】： 随着计算机图形学的飞速发展,以非真实感为目标的图形学越来越受到人们的重视。非真实感绘制是计算机图形学中一个崭新而富有活力的分支。线条画作为一种有效表示形状抽象信息的工具,属于非真实感绘制研究的领域。因其独特的表现力和抽象性广泛应用于美术创作、动漫制作及平面广告设计等领域。线条是最简单、最有效的交流方式,也是表达画者主观情感的重要手段,线条能够准确地表现自然物体的特征和轮廓。线条画的这些特点使我们能够快速识别和鉴别出图像特定内容而很少受到无关信息的干扰。另外,基于线条的物体表示还在处理时间和存储空间方面有很好的优势。为了得到非真实感图片,人们研究了各种方法。在这个领域中,如何从图像中提取能代表该图像内容的线条画和风格转换是一个研究热点。本文主要是探索基于图像的线条画提取,矢量化控制和风格转换方法。对于艺术形式的多样性和复杂性,如果找到一个合理的表示模型把某种艺术形式的风格数量化,风格重用与调整的问题就好解决了。本文给出了基于二维图像自动提取线条画的处理框架。框架有三部分组成:线条提取,线条绘制和线条风格转换。在线条提取部分中,给出两种提取图像内容的方法:一是形态学细化,并给出了一种加速运算的方法。二是依靠经典边缘检测算子获得图像中物体的边缘信息。在线条绘制部分中,首先把离散像素点用深度优先遍历的方法生成笔划路径集合,然后用三次B样条逼近绘制出每一条笔划,根据抽象程度的不同来控制线条,达到不同的绘制效果。风格转换部分是在线条矢量化的基础上,应用平面形状演化理论,通过对笔划的几何属性进行控制和调整,提供了一种灵活的调整线条画风格的手段,生成了具有夸张风格的线条画。实验结果显示用本文方法能从图像中抽象出生动的线条画,是原图像内容的抽象。本文主要是探索如何从二维图像中得到矢量化的线条问题。基于本文的工作,可以对提取出的线条画采用样本学习的方法对线条进行风格转换和定制方面的研究,还可研究其他复杂艺术形式的非真实感绘制的模型和方法,从而更好的实现由真实感照片到各种非真实照片艺术化转化。
[Abstract]:With the rapid development of computer graphics, People pay more and more attention to non-realistic graphics. Non-realistic rendering is a new and dynamic branch of computer graphics. Line painting is an effective tool for representing abstract information of shapes. It belongs to the field of non-realistic rendering. Because of its unique expressiveness and abstractness, it is widely used in the fields of art creation, animation production and graphic advertising design. Line is the simplest and most effective way of communication. It is also an important means of expressing the subjective feelings of the painters. Lines can accurately represent the features and contours of natural objects. These features of line paintings enable us to quickly identify and identify specific contents of an image with little interference from irrelevant information. Line-based object representation also has a good advantage in processing time and storage space. In order to get non-realistic pictures, people have studied various methods. In this field, How to extract line painting and style transformation which can represent the content of the image is a research hotspot. For the diversity and complexity of art forms, if a reasonable representation model is found to quantify the style of a certain art form, the problem of style reuse and adjustment can be solved. This paper presents a processing framework for automatic line drawing based on two-dimensional images. The framework consists of three parts: line extraction, line drawing and line style transformation. This paper presents two methods for extracting image content: one is morphological thinning, and the other is a method of accelerating operation, and the other is to obtain the edge information of objects in the image by classical edge detection operator. Firstly, the discrete pixels are generated by depth-first traversing method, then each stroke is drawn by cubic B-spline approximation, and the lines are controlled according to the different degree of abstraction. On the basis of line vectorization, applying the theory of plane shape evolution, through controlling and adjusting the geometric attribute of stroke, a flexible method of adjusting the style of line drawing is provided. The experimental results show that the method of this paper can abstract the vivid line painting from the image, which is the abstraction of the original image content. This paper is mainly to explore how to get vectorized lines from two-dimensional images. Based on the work of this paper, we can use the method of sample learning to study the style conversion and customization of lines. It is also possible to study the models and methods of non-realistic rendering of other complex art forms, so as to better realize the artistic transformation from realistic photos to various kinds of non-real photos.
【学位授予单位】：曲阜师范大学
【学位级别】：硕士
【学位授予年份】：2008
【分类号】：TP391.41

【引证文献】