基于有序图像的快速三维重建

Fast 3D Reconstruction Base on Ordered Image

如何提高重建速度一直是三维重建中的一个关键问题。传统的三维重建方法由于需要对输入的所有图像做任意两图像之间的匹配,故不适合大场景的重建。提出一种基于有序图像的快速三维重建方法,仅需对有序图像序列中的相邻图像做匹配,减少匹配阶段时间,从而提高重建速度。实验结果表明,提出的方法与传统的全匹配方法相比,能达到相同水平的重建效果,但在重建速度上具有明显的优势。

How to improve the computational speed is a key issue in 3D reconstruction. The traditional approach is too computational expensive to handle large scale scene 3D reconstruction, since it needs pairwise matching between all input images. Proposes a novel method for fast 3D reconstruction, which is based on ordered input images, thus the matching is only need to be performed between neighbor images. This approach can largely reduce the computational time. Experiment show that, compared with traditional methods, the proposed method has an obvious advantage in speed, while obtains similar reconstruction result.