Surface detail in non-photorealistic and realistic images

纹理在计算机图形学中使用，以提供细节世界的错觉；虚拟建筑、街道和风景使用纹理表达粗糙和锈迹的微小细节。纹理合成创建了游戏或电影的虚拟世界所需的大量细节。 纹理也用于非照片真实感渲染(NPR)，它寻求以艺术风格创建图像，如绘画或钢笔素描。历史艺术品和插图有时非常详细。在计算机图形学中，这样的细节通常由纹理提供；但在NPR中，纹理被保留用于描述介质(例如，油漆的粗糙度)，而不是用于为图像中显示的场景添加细节。因此，非照片级真实感图像往往远不如计算机生成的真实感图像那么详细。我的意思是改变这一点，将相对成熟的纹理技术引入艺术风格化，使风格化过程不像现在那样破坏细节，而是增加细节，丰富风格化图像的视觉趣味。 即使是对于逼真的图像，纹理也可以得到改善。算法开发人员一直专注于“静止”纹理，这种纹理的特性在图像中不会发生变化。变化不定的纹理相对较少受到关注，但自然界中的大多数纹理都是非静止的。我将探索高质量非静止纹理的新纹理合成方法，重点是针对树皮和龟甲等自然表面的专门算法。我打算首先专注于描绘岩石表面；岩石是千变万化的，但计算机图形学社区一直专注于合成大规模地形，而不是组成它们的岩石。 在建议的作品中，我将改进逼真和艺术形象的纹理。我将改进计算机生成的艺术图像中的细节描述，使用简化的纹理来描述场景中对象的表面细节，而不仅仅是描绘用于绘制场景的媒体。我将提高自动化能力，以创建描绘自然表面的高度可变的纹理。最后，我将通过考虑图像内容的矢量表示来改进纹理和图像编辑功能。纹理的改进将对整个商业计算机图形产生影响。纹理无处不在，是描绘表面细节的首选工具。在商业应用中，许多纹理是由艺术家构建的；辅助纹理操作的工具可以提高艺术家的生产率，而更好的自动合成方法可以让艺术家自由地在其他地方工作。

Texture is used in computer graphics to provide the illusion of a detailed world; virtual buildings, streets, and landscapes have minute details of roughness and rust expressed using texture. Texture synthesis creates the huge quantities of detail needed for the virtual world of a game or film. Texture is used as well in non-photorealistic rendering (NPR) which seeks to create images in artistic styles, such as painting or pen-and-ink sketching. Historical artworks and illustrations are sometimes magnificently detailed. In computer graphics, such detail would normally be supplied by texture; in NPR, though, texture has been reserved for depicting the media (say, the roughness of the paint) and not used to add detail to the scene shown in the image. Thus, non-photorealistic images have tended to be far less detailed than computer-generated realistic images. I mean to change that by bringing the relatively mature technology of textures into artistic stylization proper, so that the stylization process, far from destroying detail as it does at present, rather adds detail and enriches the visual interest of the stylized image. Even for realistic images, texture can be improved. Algorithm developers have concentrated on "stationary" texture, whose properties do not vary over an image. "Nonstationary" texture, that does vary, has received comparatively little attention, yet most textures in nature are nonstationary. I will explore new texture synthesis methods for high-quality nonstationary textures, emphasizing specialized algorithms for natural surfaces such as bark and tortoiseshell. I intend first to concentrate on depicting rock surfaces; rocks are immensely varied, but the computer graphics community has concentrated on synthesizing large-scale terrains rather than the rocks of which they are made. In the proposed work, I will improve textures in realistic and artistic images. I will improve detail depiction in computer-generated artistic images, using simplified textures to depict surface detail of objects in the scene, not only portraying the media used to draw the scene. I will improve automated capabilities to create highly variable textures depicting natural surfaces. Finally, I will improve texture and image editing capabilities by considering a vector representation of the image content. Improvements to texture will have impact throughout commercial computer graphics. Texture is everywhere the tool of choice for depicting surface detail. In commercial applications, much texture is constructed by artists; tools that assist texture manipulation can increase artists' productivity, while better automated synthesis methods can free artists to work elsewhere.