Abstract: This paper proposes a novel method for extracting L-system of trees from a single image, which is applied in modeling 3D trees. Firstly, the user sketches the visible trunk and crown of the tree, and recognizes the 2D skeletons of visible branches through image processing. Then, the 3D skeletons can be reconstructed according to the distribution laws of branches, from which a set of L-system rules and geometric parameters will be extracted. And the other invisible branch structures will be recovered by an iterative growth of L-system rules with the constraint of the silhouette. The proposed method, according to the experiments, turns out to be in lower cost and easier with maintaining the model quality compared with previous methods for extracting rules.

Key words: computer graphics, image-based tree modeling, L-system, lightweight 3D modeling

摘要： 介绍了一种新颖的从单张树木图片中提取树木的L-system规则的算法，并将其应用于三维树木建模。用户首先在图片上勾画出树木的主要可见枝干和树冠轮廓，通过图像处理的方法识别出树木的可见枝干的二维骨架；然后依据树木枝干的分布规律对骨架进行三维重建，并抽取其L-system生长规则和几何参数。所得L-system规则在树冠轮廓的约束下，经过迭代生长可以重建树木的分支结构。实验证明，相比以往的规则提取方法，该方法在保持模型质量的前提下，成本更低，方法更加简便。

关键词: 计算机图形学, 基于图像的树木建模, L-system规则, 轻量化3D建模