基于特征保持的高效网格去噪方法

发布时间：2018-03-28 13:09

本文选题：网格去噪　切入点：特征保持　出处：《浙江大学》2017年硕士论文

【摘要】：网格去噪用于从输入的带有噪声的网格中获取整洁并高质量的结果。因为通过仪器扫描得到的模型不可避免地在一定程度上受到噪声污染,会给后续的网格应用带来不可预测的后果,网格模型的去噪算法越来越受到重视。近些年来,学者们提出了多种类型的网格去噪算法,并取得了一定的成果。但是,这些网格去噪算法在不同程度上存在着局限性,主要有两点:(1)不能保持包含大噪声的模型的几何特征;(2)算法虽然取得了理想的视觉效果,但计算效率低下。针对上述两点,本文提出了一种新的网格去噪方法,不仅高效,同时能很好的保持模型的几何特征。本文的去噪方法分为四步:初始顶点预处理、过滤面法线、过滤顶点法线和顶点更新。在预处理阶段,引入一种快速顶点过滤算法重新计算顶点坐标,从而大幅度减少噪声的干扰。针对预处理后的网格模型,使用一种非标准化的双边过滤方法,高效的对面法线进行平滑。在处理顶点法线这一步,首先根据过滤后的面法线对顶点进行分类,然后针对不同类型的顶点,分别计算出顶点法线。最后进行顶点更新,基于上述步骤获得的面法线和顶点法线,使用一种高效迭代的方法对顶点进行更新。通过定性和定量比较结果可知,本文的方法可以超越文中选取的现有的最先进方法,尤其是在效率上,本文方法的计算效率非常高。
[Abstract]:Mesh denoising is used to obtain clean and high quality results from input meshes with noise, because the models obtained by instrument scanning are inevitably contaminated by noise to some extent. In recent years, many kinds of mesh denoising algorithms have been proposed by scholars, and some achievements have been made. These mesh denoising algorithms have some limitations in varying degrees, including two points: 1) can not maintain the geometric features of the model with large noise, although the algorithm has achieved ideal visual effect, but the computational efficiency is low. In view of the above two points, In this paper, a new mesh denoising method is proposed, which is not only efficient, but also good at preserving the geometric characteristics of the model. The denoising method is divided into four steps: initial vertex preprocessing, filtering surface normality, and so on. Filtering vertex normals and vertex updates. In the preprocessing stage, a fast vertex filtering algorithm is introduced to recalculate vertex coordinates so as to greatly reduce noise interference. Using a non-standardized bilateral filtering method, efficient opposite normals are smoothed. In this step of processing vertex normals, vertices are first classified according to the filtered surface normals, and then for different types of vertices. The vertex normals are calculated respectively. Finally, vertex updating is carried out. Based on the surface normal and vertex normals obtained from the above steps, an efficient iterative method is used to update vertices. The method in this paper can outperform the most advanced methods selected in this paper, especially in the efficiency, the computational efficiency of this method is very high.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【相似文献】