中文摘要：

近年来, 图像空间的线绘制技术在绘制质量和算法鲁棒性方面取得了长足的进步. 但是, 图像空间的线绘制算法不能生成特征线的空间几何信息, 因而应用面不如物体空间的线绘制技术广泛. 针对现有物体空间线绘制算法的绘制效果和绘制速度还远不如图像空间算法, 提出一种物体空间线绘制算法. 在预处理阶段, 采用网格模型的高斯差分算子计算法向拉普拉斯; 在实时绘制时, 通过计算法向拉普拉斯与视线方向的点积, 从而快速计算特征线的三维几何. 在该算法框架下, 进一步提出各向异性的高斯差分算子, 在计算法向拉普拉斯时对于特征线的梯度方向和切线方向做不同的滤波处理, 从而更好地反映特征线的走向. 实验结果表明, 文中算法比现有的其他物体空间线绘制算法更加鲁棒, 对于包含几何噪声和非均匀网格化的三维模型, 其生成的线绘制结果更加光顺, 能更清晰地揭示模型的形状特征; 在实时绘制效率方面, 能实时生成包含上百万三角面片的网格模型的线绘制结果.

英文摘要：

In recent years, image-space line drawing techniques have made considerable progress in terms of quality and robustness. However, image-space methods do not generate the feature lines geometry in object space, which limits their application in many aspects. On the other hand, the rendering quality and efficiency of current object-space line drawing algorithms are inferior to the image-space methods. In this paper we present a new ob-ject-space line drawing algorithm. We use the Difference-of-Gaussian (DoG) operator for mesh surfaces to pre-compute the Laplacian of normal for each vertex. At runtime, the feature lines are efficiently extracted by com-puting the dot product of the Laplacian of normal to the viewing direction. Taking the direction of features into consideration, we further employ the anisotropic DoG operator, which filters differently in the gradient and tan-gent directions at each point. Experimental results show that, our method is more robust than the original Lapla-cian line drawing algorithm, as well as other object-space algorithms. For models with geometric noises and ir-regular tessellation, our results are smoother and cleaner. In addition, our method has the same run-time efficiency as the original Laplacian lines, and reaches real-time performance for models with millions of triangles.