140 likes | 551 Views
FPGA BASED REAL TIME VIDEO PROCESSING. Characterization presentation. Presented by: Roman Kofman Sergey Kleyman Supervisor: Mike Sumszyk . AGENDA. Project Objectives. Algorithm. Based on the 2D non-linear Diffusion equation Iterative solution. Good Filtering – smoothes noises.
E N D
FPGA BASED REAL TIME VIDEO PROCESSING Characterization presentation Presented by: Roman Kofman Sergey Kleyman Supervisor: Mike Sumszyk
Algorithm • Based on the 2D non-linear Diffusion equation • Iterative solution. • Good Filtering – smoothes noises. • Keeps borders intact.
Matlab simulation Filtered image Original image dt=4 , 4 iterations
Schemes Explicit Semi implicit • Stable for small time steps. • Good results requires many iterations. • Simple implementation. • Stable for all time steps. • Good results after several iterations. • High computational and storage effort (Thomas).
Original image dt=0.5, 50 iterations dt=5, 5 iterations
Original image 10 iterations dt=1.1 20 iterations 50 iterations
Basic Block Diagram DVI IN DVI OUT Implementation with internal pipeline PROC MultiPORT On-board memory blocks Daughter board Daughter board ProcStarII board