Spatio-Temporal Video Filtering
There are many image and video processing algorithms (e.g. calculation of Optical-Flow, or Image Domain Warps) that are usually solved using large optimization problems. The solution of such large optimization problems in real-time is difficult and sometimes even infeasible. However the mathematical structure of some of these problems allows us to approximate their solution by using non-linear filtering in the spatial and temporal domain. These filters scale better in terms of computational complexity than the corresponding optimization problems, and therefore would allow to perform certain video processing steps more efficiently. In this project we would like to evaluate different filtering kernels with respect to hardware efficiency, and develop an efficient hardware architecture which is capable of filtering a high definition video stream in real time.
The project is quite open at the moment and is suitable for Master and Semester Theses.
- Scope: Semester or Master Thesis
- Looking for 1-2 Interested Students
- Supervisors: Michael Schaffner, Lukas Cavigelli
- VLSI I
- Introductory course in computer vision (optional)
- Interest in computer graphics / computer vision
- Matlab, VHDL and C++