纹理合成算法研.pdf

纹理合成算法研究 摘 要 基于样图的纹理合成技术给定一个小的样 图来生成大尺寸的纹理图，使其 具有相似的视觉纹理特征。该技术在视频游戏、虚拟现实、计算机辅助设计等 方面有广泛的应用前景，并成为计算机图形学研究的热点之一。本文在之前的 研究基础上对纹理合成进一步研究。 论文首先回顾了基于样图的纹理合成技术的发展历程，并对经典算法进行 了介绍，分析了各算法的思想、特点以及不足。接着对二维纹理合成进行深入 研究，相邻 Wang Tile 拐角处的匹配问题是获得更好合成效果的关键。为了解 决这一问题，提出了一种改进 Wang Tile 的新方法。该方法使用一个新的纹理 块来替换边界匹配纹理块的中心，并采用 Image Quilting 算法查找替换路径， 从而生成 Tile 集合，继而合成大块纹理。所绘制的纹理效果克服了匹配问题， 而且在某种程度上增强了纹理多样性。最后，我们深入研究了 EM 算法的全局 优化纹理合成思想，提出了一种实体纹理生成思想，并将其应用于实体表面纹 理生成。 关键词： 纹理合成；Wang Tile ；实体纹理；全局优化 Research on Texture Synthesis Algorithm Abstract Texture synthesis from Sample (TSFS) uses a given sample texture to synthesize large images with similar visual characteristics. It has been widely applied in many areas such as video games, virtual reality, CAD and so forth, and has been one of the hottest research spot of computer graphics. Base on previous research efforts on TSFS, this dissertation does a further research on texture synthesis. Firstly, we review the research results of TSFS and introduce typical algorithms, and analyze the principle, specialty and deficiency of these algorithms. Secondly, after studying on the technology of 2D texture synthesis deeply, we find that matching problem at the corners of adjacent Wang Tile is the key issue of achieving graceful synthesis results. To solve the issue, an improved Wang Tile creation technique is proposed. By replacing the center of edge-matching texture tile with a new texture tile and using the Image Quilting path-finding algorithm to find the instead path, this algorithm generates Tile set and synthesizes a large te