APPROACHES

Approach

There can be many appraches to tackle this problem. Two basic categories are :

• Caliberated approach - where the exact location of camera and its internal features are known a priori. This appraoch is taken when large number of camera are used simultaneouly .
• Uncaliberated appraoch - where none of internal nor external features are known precisely.

Our approch

We are going for the uncaliberated approach. Although the problem is a bit involved, we still worked on it as it provides more realistic scenario. We are using two images to find matches between the world points projected onto the two image planes by estimating the fundamental matrix ( which is simply algebraic representation of epipolar geometry ). Finally we find the co-ordinates ( upto a scale ) of points projected on the image planes in the world frame using the projection matrices for the two camera positions . Earlier we had thought of reconstructing the entire 3-d geometry of polyhedral objects using 10-15 images of the same object taken frm all sides of the object but due to shortage of time & complexity of the problem we restricted ourselves to two views but compensated the problem by being able to reconstruct not only polyhedral objects but any scene upto the level of finding the depth map.