1 / 11

Short-term CV Motion Prediction and its use in long term motion compensation

Short-term CV Motion Prediction and its use in long term motion compensation. Stanford University EE398 Final Project Zhengyun Zhang, Dileep George Winter 2005. Overview. Problem Formulation Short-term Constant Velocity Predictor Results Conclusions. Problem Formulation.

marcel
Télécharger la présentation

Short-term CV Motion Prediction and its use in long term motion compensation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Short-term CV Motion Predictionand its use in long term motion compensation Stanford University EE398 Final Project Zhengyun Zhang, Dileep George Winter 2005

  2. Overview • Problem Formulation • Short-term Constant Velocity Predictor • Results • Conclusions

  3. Problem Formulation • New codecs (e.g. H.264/AVC) can use long-term motion compensation to increase coding gain • Computationally expensive, O(NxM2xD) • N macro-blocks • ~M length of average motion vector • D frames to search • Can we find a more efficient algorithm?

  4. Short-term Constant Velocity Model • Objects in motion tend to not change in velocity drastically • Short term velocity approximately constant • Predictor can follow this model

  5. Short-term CV: First Step Predictor t-2 Desired predictor from past t-1 t x-1 x x+1

  6. Short-term CV: Second Step Predictor t-N t-N+1 Predictor: b t-N+2 c=2b-a c a ... t x-1 x x+1

  7. Short-term CV:Motion Search • Search fixed window centered on predictor (x,y) • Same as brute-force search, except brute-force search always predicts (0,0) • Short term CV also keeps track of best predictor for each frame in the backbuffer while searching

  8. Results:Foreman (frames 1-399)

  9. Results:Mobile (frames 1-299)

  10. Conclusions • Works comparatively well for small search ranges, but not for larger ones • Good predictor, since it’s closer to optimal for very small search ranges • Combining with a better per-frame search algorithm should get better gain • e.g. steepest descent • Combining with decision-tree structure can reduce errors with large search ranges

  11. References • H. Chung, A. Ortega, A. A. Sawchuk, "Low complexity motion estimation for long term memory motion compensation," Proc. Visual Communications and Image Processing, VCIP 2002, San Jose, CA, January 2002. • A. M. Tourapis, “Enhanced predictive zonal search for single and multiple frame motion estimation”, VCIP 2002, pp.1069-1079 • T. Wiegand, G. J. Sullivan, G. Bjontegaard, A. Luthra, “Overview of the H.264/AVC Video Coding Standard,” IEEE Trans. CSVT, Vol. 13, No. 7, pp.560-576, July, 2003

More Related