Speed up motion estimation
My project is speeding up motion estimation for various video processing applications. The programming is to be done in Matlab. My project supervisor has a meeting every week with me. He assigns me a particular task to be completed within that week.
Description- 1)Build a RavME Motion Estimator that uses motion from Compressed video as a kick start. Compare the below Measures of performance wrt to really good motion estimator(RGME) like EPIC
Are the vectors from RavME the same as from full search BM?
Are the vectors from RavME the same as ground truth or RGME?
Is the prediction error from RavME the same or better than full search BM?
Is the prediction error from RavME the same or better than RGME?
2)work on P – frame
Select candidate motion vectors at a site .. remember to scale depending on how far in the past they point to
For each candidate
Measure DFD
Measure Spatial smoothness
Pick the best
3)B – frames == same thing
Select candidate motion vectors at a site
Flip the vectors around. So if you have one vector pointing back, then just negate it and make it also point forward
Scale the vectors depending on the target frames
For each candidate
Measure DFD
Measure Spatial smoothness
Pick the best
4) then figure out for Iframe
The code should be simple without using any matlab functions
