3D reconstruction: Why should the accuracy always be presented in the pixel unit?

摘要

This paper analyses the error presentation of parallax-based techniques (mainly stereoscopy and structured light). They are usually presented using an absolute (mm) or a relative (%) scale. These results are hard to compare between different systems as they are system-dependent. This paper presents results using the pixel unit which avoids the influence of geometric parameters. Moreover it is apt at evaluating whether the system under-performs or is similar compared to theoretical accuracy.