基于多边形顶点法矢量的网格模型简化算法

在计算机图形学中，经常采用网格模型进行几何物体的描述，而网格模型的大数据量成为实时绘制的瓶颈，因此，必须对网格模型进行简化。目前的简化算法，主要是以网格模型几何误差的最小化为准则，而忽略了模型的视觉特征，为此提出了一种基于法矢量的模型简化算法，其简化准则是视觉特征的最优化。首先获取多边形顶点的平均法矢量，然后依据该法矢量确定简化门限。实验结果表明，当地景模型简化至95.4%时，仍然保持了令人满意的图象质量。该算法能够在保证高度真实感视觉效果的前提下，实现模型较大幅度的简化。

An Algorithm of Vertex Normal-Based Mesh Model Simplification

Model simplification is very important in the fields of virtual reality and real-time simulation. In computer graphics, objects and scenes are often described by mesh models, such as terrain model, Stanford's scanning models and so on, but, a complicated model often includes large-scale data, the large data scale is a bottle-neck of fast-speed rendering, so it is necessary to simplify mesh model. The approach of levels of detail(LOD) is good, there have been some efficient algorithms for LOD and model simplification, these algorithms' simplification criteria are minimizing the geometry errors. It has been found that the last destination for simplification is to obtain fast rendering speed as well as good visual effects, hence, visual effects can and should become simplification criteria. This paper gives an algorithm of normal-based mesh model simplification. First, this paper gives a mathematics model for curve, and concludes that grads can express a curve's visual effects. Second, this paper uses normal of mesh models to replace grads, if the normal is averaged three times or more, a mesh can simulate a cure well, and the averaged normal can replace grads to become simplification criteria. Finally, this paper defines an angle between a normal of given point of intersection of the mesh model and those around this point, and uses this angle to simply mesh models. A group of experimental results are also given, which illustrate that the above algorithm can realize model simplification effectively, and at the same time, keep the models highly accurate.