计算机绘制中国画的方法研究

【作者】 李宁；

【导师】 王立新；

【作者基本信息】 华东师范大学 ， 软件工程， 2008， 硕士

【摘要】 中国画是一门历史悠久的艺术,在中国已经具有了近千年的历史。对传统绘画艺术进行仿真是计算机艺术领域最具挑战性的课题之一。在吸取传统绘画艺术的营养、弘扬传统绘画艺术成就的同时,对计算机仿真绘画技术提出了新的要求。非真实感绘制(Non-Photorealistic Rendering,NPR)从20世纪90年代中期开始逐渐成为计算机图形学的一个研究热点,是计算机技术和绘画艺术相结合的一个研究领域,指的是利用计算机生成不具有照片般真实感(Photorealistic Rendering,PR),而具有手绘风格的图形的技术。其目标不在于图形的真实性,而主要在于表现图形的艺术特质、模拟艺术作品。NPR技术允许艺术家们强调某些细节而同时忽略其他细节,因此艺术家们可以借此利用NPR的风格画模拟技术,生成不同风格的图像艺术作品。目前已经有多种艺术风格画被成功模拟,如水彩画、铅笔画、钢笔画、油画等等。中国画是我国一种传统的绘画艺术,其画风朴素抽象,笔划简单,但具有丰富的表现力。对中国画的模拟实现,不仅是传承我国优秀文化的一种新方式,也是对NPR技术的一种新的拓展。从单纯地以样条曲线生成类似于毛笔的笔划,到沿着笔划路径“贴”水墨纹理,再到逐渐考虑绘画工具的物理特性而对毛笔、纸张、墨等进行建模,现在已经可以利用手写装置实时地在电脑上进行水墨画的创作,这些模拟方法越来越完善,模拟效果越来越逼真。但这些方法或系统的应用却需要用户具备一定的绘画技术,才能创作出一幅令人满意的水墨风格画作品。本文提出一种基于图像的(Image-based Rendering,IBR)中国画风格绘制方法,该方法能够不依赖用户的先验知识,自动将一幅彩色图像转化为具有中国画效果的图像。基于图像的绘制是将计算机图形学与图像处理相结合的一种新的绘制方法,不需要建立几何模型,只需从输入图像中获取所需信息,通过视点插值、图像变形等方法,重构所希望的效果。通过观察中国画可以发现,中国画的绘制以笔法为主导,充分发挥墨法的功能,墨即是色,墨的浓淡变化就是色的层次变化。而在图像中,相同颜色的区域会因为光照的原因而呈现出不同的色彩,若根据色彩对图像进行平滑、分割,灰度化后即可产生墨色效果。中国画还有一个独特的效果在于绘画所使用的宣纸具有吸水性,而墨的主要成分为碳和水,当墨点到纸上时,墨中的水会在宣纸纤维的吸收作用下流动,从而带动碳的运动,因此在笔划的周围产生水墨扩散的晕状效果或不规则边缘。假设扩散效果由笔划边缘上的水墨流动产生,本文首先对输入的彩色图像采用Mean Shift算法进行平滑和分割,并提取出图像边缘,再为这些边缘点赋予一定的水墨量,对应为图像中每个点相应的灰度值,最后将这些边缘点作为水墨扩散的起始点,利用我们提出的一种边缘水墨扩散算法进行处理,得到水墨画中水墨在宣纸上扩散的艺术效果。Mean Shift,即均值移动,指一个迭代的步骤,先算出当前点的偏移均值,移动该点到其偏移均值,然后以此为新的起始点继续移动,直到满足一定的条件结束,表现为自适应地梯度上升搜索峰值。经证明,Mean Shift算法在满足一定条件下,一定可以收敛到最近的一个概率密度函数的稳态点,因此Mean Shift算法可以用来检测概率密度函数中存在的模态,从而能很好地进行图像的平滑和分割,同时可以提取出分割区域的边缘。水墨从笔画边缘扩散出去的过程,我们模拟实现为纤维网格上的水的运动和水带动碳粒的运动,最后根据碳的浓度计算出对应的灰度值。整个图像可视为一张二维网格,每个网格对应一个像素。为每个像素分配随机数目的纤维,正比为该点的吸水率。对每个边缘点,根据吸水率计算出该点的吸水量,多余的水会向其邻点流动,然后根据每个邻点纤维数的比例计算分别流向它们的水量,同时根据边缘点中碳的浓度计算出分别带入八个邻点的碳量,从而得到每个邻点的水墨量。再将邻点作为新一轮扩散的起点,继续扩散,直到水墨被完全吸收,则扩散停。实验表明,本文提出的基于图像的中国画水墨风格绘制方法能将真实感的图像转化为令人满意的具有水墨风格的图像。更多还原

【Abstract】 The Chinese ink wash drawing has a history of about one thousand years. One of the most challenging work in the computer art research field is to simulate thetraditional paint media and tools. The representation of normal Chinese painting is different from that of Chinese landscape painting, which is of complicated brushwork. The research on the graphical simulation of Chinese landscape painting is beneficial to the development of Chinese painting and computer art.Non-Photorealistic Rendering (NPR) has been a research hotspot of Computer Graphicssince the medium term of 1990s. It injects aesthetic and stylized elements into computer graphics,and the goal of NPR is not to render objects realistically but creatively, such as in simulatingvarious painting styles. By NPR, the artists can convey not only abstract information but also acertain artistic spirit through simulating various artistic styles.At present, many kinds of artistic style paintings, such as watercolor painting, pencilsketching, pen-and-ink illustration, canvass, etal., have been simulated successfully. Chinese inkpainting is a kind of traditional art forms in China which has a history more than three thousand years. It adopts four main tools: the hairy brush, ink stick, paper and ink stone. In Chinese inkpainting, the artist doesn’t care about producing an exact or literal reproduction, but expressingonly the essence of the considered objects. In other words, paintings typically consist of just a fewsimple strokes intended to convey the artist’s deep feelings regarding the painted object. Simplicityis therefore a key ingredient. And feelings are expressed by the speed, placement, pressure, andmovement of the brush, as well as the resultant degree of shading produced by the brush strokes. To simulate Chinese ink painting is not only to inherit our excellent traditional culture, but also todevelop a new NPR technology.The simulation of Chinese ink painting has gone through a course from simplicity tocomplexity, and now we can use the pen input devices to create Chinese ink paintings on thecomputer in real-time. However, all the systems ask for the users who wish to draw a satisfiedpicture with some certain painting skills. Without background experience about painting, thispaper presents a system for user, who can convert a color image to a picture with Chinese inkpainting effect automatically. As we known, ink is an important pigment to paint a Chinese ink painting, and the variousgray effects on the painting are produced by ink mixed with water. In addition, the ink diffusionalong the stoke edges is a very important aspect of Chinese ink painting. In order to realize theeffects described above, we present a painting method based on an input color image. At first, we use mean shift to filter and segment the image, and after we change the image to gray scale, the similar gray effect of an ink painting is shown Then, we can detect the segmented regions’ edges,let the edge points to be the start points for diffusion. We can set each point an ink value whichcorresponding to its gray value. The third step is to diffuse the ink from the edges. Since the fibersof the absorbed paper is the main cause for ink diffusion, so if we look the image as a 2D grids,and each pixel is a grid, and we allocate each grid some fibers randomly, we can achieve theabsorbed rate of each grid. Then we simulate the ink diffusion on the paper as follows:The experimental results illustrate that our methods are promising.更多还原