| With the continuous development of digital geometry applications,the demand for 3D models is increasing day by day,but in the process of using 3D scanning equipment to obtain 3D models,it is inevitable to introduce different levels of noises,so 3D mesh model denoising is must step for geometric processing.The main goal of the mesh denoising method is to retain or restore the features of the model while removing the noises on the model,without introducing additional noises or deformations.Due to the large differences in the shape of the 3D models and the difficulties of distinguishing the noise levels,existing algorithms often use multiple steps or multiple iterations for processing,resulting in a high time complexity,and often require a lot of parameter adjustments and manual terminations of the iterations.Faced with these problems,this paper mainly studies the efficient adaptive and feature-preserving denoising methods for the mesh model.In this dissertation,we jointly analyze the over polished model reflecting the structural characteristics and the noisy model containing comprehensive information,and an adaptive algorithm based on linear interpolation is proposed.Through adaptive linear interpolation,feature information is gradually extracted from the noisy model and added to the smooth model to achieve denoising results that maintain features of varying levels.We also propose the judgment conditions for selecting the number of iterations,thus avoiding the manually operations.Furthermore,An adaptive bilateral filtering algorithm is proposed.We utilize the Laplace operator to describe the features of different levels.And we propose to set the variance parameter adaptively based on the Laplace operator.Thus different triangles are endowed with different bilateral filters based on the geometric characteristics.Such settings can help better maintain the characteristics of the model.And this also enable the modifications of the existing local and global bilateral filtering algorithms.Finally,we verify the effectiveness of the proposed algorithm through simulation experiments.Experimental results show that the two denoising algorithms and termination criteria proposed in this paper can obtain adaptive feature-preserving denoising results,and are robust to non-uniform sampling and large noises,and can remove noises while maintaining the detailed characteristics of the model and reduce the time cost,the overall performance is better than the existing denoising methods. |
