130 likes | 477 Views
Motion Estimation Final Project. Adviser: 葉家宏 Presenter: 張晏豪. Outline. Introduction Proposed motion estimation Comparison Demo conclusion. Introduction. Motion is continue and need more than 2 frames to display The motion vector in previous frame is very close to current frame.
E N D
Motion EstimationFinal Project Adviser:葉家宏 Presenter:張晏豪
Outline • Introduction • Proposed motion estimation • Comparison • Demo • conclusion
Introduction • Motion is continue and need more than 2 frames to display • The motion vector in previous frame is very close to current frame
Difference of motion vector Δmv2 Δmv1 … … … Facro block … … … … Frame Frame Frame Δmv : Motion vector of current frame minus motion vector of next frame
Mean of Δmv Size of frame:176x144 Size of Macro block:16x16 Each frame have 99 macro blocks Average motion vector= /number of macro block of a frame /number of frame
Proposed method -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 Step1: First frame employs full search algorithm, we can get the point with minimum SAD. The point searched from the first frame will be the initial center point at next frame 7 6 5 4 3 2 1 0 -1 -2 -3 -4 -5 -6 -7 Search window
Proposed method -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 7 6 5 4 3 2 1 0 -1 -2 -3 -4 -5 -6 -7 Step2: Pick 9 points in the shape of diamond,if the minimum SAD is in the center, then evaluate SAD at 4 more points around it. The point with minimum SAD will be the initial center point at next frame Search window
Conclusion • Proposed method • Reduce coding time • Slow motion has better PSNR than high motion