Bachelor Thesis Preview: Extracting Point Features for Symmetry Detection
Daniel Mewes
Max-Planck-Institut für Informatik - D4
Talk
Since winter term 2008/2009, I am a bachelor student of computer science (math minor) at Saarland University.
Since October 2009, I am a Student Assistent at the Statistical Geometry Processing group at MPII Saarbruecken.
I am now writing my bachelor thesis under the supervision of Michael Wand.
Automatic understanding of the structure of 3D point data with respect to symmetries in the scene has many practical applications (e.g. data compression, simultaneous editing etc.). The problem of symmetry detection is computationally complex however, even when limiting detection to the class of rigid symmetries only. Therefore, many existing algorithms utilize features to reduce the number of potential symmetries to probe to only the most promising ones.
In my Bachelor thesis, I explore ways of improving the quality of point features extracted from imperfect 3D point data with respect to their usability for subsequent symmetry detection. To achieve this goal, I propose and evaluate a framework which allows incorporating algorithms that "guess" about the scene's structure early in the feature extraction phase in order to improve the set of extracted feature points.
This talk gives an overview of my thesis topic and some preliminary results.