• Introduction
• Proposal
• Projective Geometry
• Registration
• Global Registration
• Composing
• Implementation and Results
• Conclusion
• References
• Appendix

Proposal

HIGH-RES PANORAMAS USING IMAGE MOSAICING

Feeling strongly interested in the field of immersive environments, we would like to consider the image mosaicing technique, whose aim it is to stitch together multiple shots from the same place into one large reconstruction of the environment field . Our review of existing literature shows that this is still an active field of research and that so far many different methods have been proposed. Our goal is to to assess the efficiency of a selection of algorithms and to come up with a reasonably simple solution that yields visually compelling results. Indeed the problem can be tackled at different levels of complexity depending on the number of degrees of freedom that the camera movement has. We plan to restrict ourselves to a "single point-of-view" constraint, that is, we plan to take a sequence of shots from a digital camera mounted on a tripod. The expected result is a wide field of view, high-resolution, one piece panorama extending in both dimensions.

The main parts of our work will be focusing on the modelling of the camera model given the chosen constraint, the registration of the shots relatively to each other, and the blending between the images in the final projection. We will most likely use an eight parameter projective camera model [1]. But the main variations between image mosaicing methods lie in the registration technique. For that matter, we intend to implement a frequency domain phase-correlation method [2] and a feature-based registration technique. It might even prove beneficial to combine the two because the phase-correlation method can provide a good initial condition to the feature-based search. For the composing, we will first evaluate a simple bilinear blending method. If it proves to be unsatisfactory, we will also try a multi-resolution spline approach [3].

For performance, debugging, and reusability reasons we plan to implement our algorithms in C.

[1] Szeliski, R., Video Mosaics for Virtual Environments, IEEE Comp. Graph. and Applications, March 1996, pp 22-30.
[2] Davis, J., Mosaics of Scenes with Moving Objects, IEEE Proc. CVPR 1998.
[3] Burt, P.J., Adelson, E.H., A Multiresolution Spline With Application to Image Mosaics, ACM Transactions on Graphics, Vol. 2. No. 4, October 1983, pp 217-236.