2002 | Bruno Lévy, Sylvain Petitjean, Nicolas Ray, Jérôme Maillot
This paper presents a novel method for automatic texture atlas generation for 3D models, focusing on minimizing artifacts and improving texture mapping efficiency. The main contribution is a least-squares conformal map (LSCM) parameterization method that minimizes angle distortions and geometric property norms, ensuring uniformity and independence from mesh resolution. The method is robust and can handle large charts with complex borders, avoiding triangle flips and ensuring consistent orientation. The paper also introduces a segmentation method to decompose the model into charts with natural shapes and a packing algorithm to optimize texture space usage. The approach is validated through experiments on scanned and modeled datasets, demonstrating its effectiveness in generating texture atlases with fewer discontinuities and better texture mapping performance compared to existing methods.This paper presents a novel method for automatic texture atlas generation for 3D models, focusing on minimizing artifacts and improving texture mapping efficiency. The main contribution is a least-squares conformal map (LSCM) parameterization method that minimizes angle distortions and geometric property norms, ensuring uniformity and independence from mesh resolution. The method is robust and can handle large charts with complex borders, avoiding triangle flips and ensuring consistent orientation. The paper also introduces a segmentation method to decompose the model into charts with natural shapes and a packing algorithm to optimize texture space usage. The approach is validated through experiments on scanned and modeled datasets, demonstrating its effectiveness in generating texture atlases with fewer discontinuities and better texture mapping performance compared to existing methods.