基于分片的特征敏感的曲面重构.docVIP

基于网格变形的从图像重建三维人脸 董洪伟 江南大学物联网工程学院, 无锡,中国 214122) 摘 要: 从图像重建高质量三维人脸一直是计算机视觉和图形学的一个重要研究问题。不同于传统的基于立体匹配的窄基线多视几何和数据驱动的人脸形变方法，本文提出一种结合网格变形技术和立体视觉原理的从图像重建高质量三维人脸模型方法，给定对不同视角拍摄的几幅人脸图像，基于健壮图像特征如SIFT的稀疏立体匹配和一种更可靠的从运动恢复结构技术获得可靠的相机外部参数和稀疏三维点，在此基础上，提出一种结合几何细节保持和图像一致性约束的三维人脸变形算法重建三维人脸，通过对模板人脸的网格变形使得变形人脸在多幅图像中的可见投影具有一致性的图像颜色强度。基于模板的人脸变形可以有效解决三维模型成像中的遮挡问题，采用健壮估计法消除噪声和离群点对目标函数收敛性的影响，对目标函数的多次非线性优化求解进一步改进了人脸重建的质量。采用合成人脸图像和真实人脸图像重建三维人脸的实验结果表明，本文算法可以从几幅宽基线图像重建高质量的三维人脸模型。 关键词: 人脸重建；网格变形；基于图像的人脸重建；多视立体重建 分类号: TP391　　 文献标识码: A Mesh deformation based 3D facial modeling from images DONG HongWei (School of Information Engineering, Jiangnan University, Wuxi, China 214122) Abstract: High quality 3D facial modeling from images is an important problem in computer vision and graphics. Unlike the traditional narrow-baseline multi-view geometry and data driven shapeblending facial modeling from images, we propose a new high quality facial modeling from images technique which deforms a template mesh model using Laplacian mesh deformation technique and color consistency metric de?ned by all input images. Given a few facial images taken from different viewpoints, we first do sparse stereo matching from matched robust image feature points such as SIFT and use a robust structure from motion technique to recover camera extrinsic parameters and a few 3d matched points. Then an image based mesh deforming framework is proposed which deform a 3D facial template to make their visible images have consistent color intensities. Template based deformation avoid the difficulty due to occlusion in image data. Robust estimation reduce the influence of image noise, outliers and surface highlights. Multiple nonlinear optimization of object functions improve the quality of facial modeling. Experimental results using synthetic and real images for facial modeling show that our method provides stable and quality facial reconstruction results on wide-baseline images. Key words: facial modeling; mesh deformation; im