Modeling of complex 3D objects is a painful and time-consuming task. The
work presented here focuses on the reconstruction of a 3D model of existing objects starting from
photographs, using a paradigm called Image-Based Modeling.
Automatic approaches for Image-Based Modeling, as developped in Computer Vision for instance,
have shown their limits, as some ambiguities can only be solved by a human interpretation.
This paper describes the different tasks of such a computer assisted Image-Based Modeling process that are used in our software environment called
Reality. This solution is both iterative (tasks may be
repeated to obtain more precision) and constrained (the 3D model is generated according some geometric constraints provided by the user).
We also present several possibilities for the initial tasks of the process, depending on the given information. By an example, we also show
that Reality is able to reconstruct a 3D model without any prior information about about the scene's geometry or about the
camera's characteristics.
