Reconstruction of the surface of the Sun from stereoscopic images

Papyrus

http://hdl.handle.net/1866/7228

Abstract

This thesis concentrates on the stereoscopic reconstruction of environments containing transparent objects. The data used to test the algorithms is graciously provided by NASA through the STEREO mission. This thesis proposes a new spherical rectification technique as well as a dense reconstruction algorithm without making any prior assumptions on the shape or transparency of objects inside the scene. Firstly, the camera parameters are estimated, following a refinement step to get seamless alignment between images. In the next step the images get rectified in order to be able to restrict the search space to 2D rather than the full 3D. Afterwards the density at along each epipolar line gets estimated. The reconstruction of transparent scene is still largely an open problem and there are no general methods to deal with transparency. The applications of such an algorithm are numerous, ranging from reconstruction of smoke trails inside wind tunnels, optimal design of combustion chambers, augmented reality, etc.

Documents

PDF

---