Location: MIC-Arena, Maria-Goeppert-Str. 3, 23562 Lübeck
Title: Semi-global Matching for 3D Image Registration
Abstract: Semi-global matching (SGM) is popular technique for estimating pixel disparities in a stereo image to reconstruct the 3D geometry of a recorded scene. Stereo vision has become a powerful component for scene understanding tasks in computer vision and robotics applications. It has been adapted very quickly by the industry due to its real-time capability and it yields very robust performance when combined with a census data term.
This talk gives a proper introduction to the concept of SGM and presents how this technique can be extended to make it applicable for 3D image registration. Another topic will be an overview of the census data term and how this technique can be applied in variational approaches.
