Modeling 3D Humans from Uncalibrated Wide Baseline Views

In order to achieve highly precise, accurate and reliable tracking of humans, a 3D human model that is very similar to the targeted subject is needed. In this paper, we present a new method to precisely construct the external shape of the whole human body. In this work, we make use of a set of images of the subject seen in stanza position from different views, and a regular generic anthropometrical 3D model. The specific 3D model is obtained by the deformation of the generic one controlled by the set of images. We developed a two-stage iterative technique that makes use of human shape characteristic points and limb outlines that work together with the generic 3D model to yield our final customized 3D model. The first stage is an iterative camera pose calibration and 3D characteristic point reconstruction-deformation algorithm that gives us an initial customized 3D model. The second stage refines the initial customized 3D model by de-formation via the silhouette limbs information. From our results, we demonstrate that our system is able to construct a good quality human model.