Abstract: Reconstruct colored 3D-model which can show the surface texture is an important issue in the field of computer vision. A registration method based on range-image is proposed to create a 3D-color model. This registration algorithm uses both geometric and texture information during point-matching stage in pair-wise registration and weighting update in global registration. Experimental results show that this approach provides a well registered 3D-color model.

Key words: 3D reconstruction, range images, texture, 3D registration

摘要： 重建包含模型真实纹理的彩色三维模型是计算机视觉研究领域的一个重要问题，提出了一种基于2.5维深度图像的三维配准算法，同时利用原始数据中的几何以及纹理信息，对多幅深度图像进行两两配准和全局配准。新算法在两两配准阶段将纹理信息用于初选两幅图像的匹配点对，在全局配准阶段将纹理信息用于计算残差，并据此更新对应点对的权重值用于下一次迭代计算。该算法克服了以往算法在纹理、几何信息的参数尺度量化等方面的问题。

关键词: 三维重建, 深度图像, 纹理信息, 三维配准