真实感三维模型的纹理融合

Texture Blending of 3D Photo-Realistic Model

提出一种消除物体表面纹理接缝,重建高精度真实感三维模型的纹理融合方法.首先将不同视场的深度图像匹配到同一坐标系下,经过深度图像融合获得物体完整的几何模型;然后确定深度图像和纹理图像的映射关系,并定义复合权重进行纹理融合来获取整个纹理映射图,并进行模型的纹理映射.该方法可消除物体表面重叠区域由于光照变化、几何模型的非完全漫反射以及拍摄视场的改变等外界因素引起的色彩差异,重构真实感三维模型;且对物体的几何形状、拍摄视场没有限制,能够自由拍摄,并实现高精度三维纹理模型.实验结果表明,文中方法是有效和鲁棒的.

A method of texture blending for eliminating texture artifacts and p textured model is proposed. In this method, the range scans of multi-view common coordination system, and integrated to establish a 3D object model. roducing a photo-realistic are first registered to a A set of 2D photographs are aligned with 3D model through an accurate calibration. Then a eomposited weight function is defined and used to remove the artifacts within the overlapping areas, which are given rise to from lighting change, non-lambertian object surface, etc. The experiment results show that the proposed method can produce high-quality photorealistic textured model without any constraint on model shape.