Image-based Rendering and Modeling

In this class project, we describe a flexible approach for constructing mosaics of architectural environments from a sparse set of uncalibrated views. The main contribution in this work is the use of environment constraints in order increase the efficiency and level of automation of the mosaic construction process. The observation that in architectural environments, the majority of lines is aligned with the principal orthogonal directions of the world coordinate frame, will be exploited in different stages of the mosaic construction pipeline. The automated detection of vanishing directions will enable us to partially calibrate the camera an estimate the relative orientation of the camera with respect to the scene from a single view. These initial estimates will facilitate efficient feature matching, computation of displacements between the views as well as alignment of multiple views. While the approach described here will be presented in the context of rotational mosaics, the alignment and matching techniques are applicable for general displacements, where the constraints of man-made environments are present and the displacement between the views is large.

Multiple Views

Mosaic

[Back to my research page]