1. MOTION ESTIMATION AND VIDEO COMPRESSION
By,
Jarjit Tandel
Waseem Khatri
Sidhesh Khapare

2. Outline Introduction Motion Estimation Motion Compensation Algorithm
Block Estimation Algorithm
Compression
Results
Conclusion
References

3. Introduction Motivation Background Understand Motion Estimation
Reconstruction of Video Using Motion Compensation
Background
A Video sequence consist of series of frames.

4. What is Motion Estimation
Predict current frame from previous frame
Determine the displacement of an object in the video sequence
Types of Motion Estimation:
Horn and Schunck
Three Step Search Block Motion Method
Hierarchical Block Motion

5. What is Motion Compensation
Reconstruction of video file
Reference frame is used to predict current frame using motion vectors.

6. Proposed Algorithm Input Color Video Extract frames ‘k’ and ‘k+1’
3-step motion estimation
Obtain motion vectors
Forward motion estimation
Predicted video frame -
Predicted video frame
Original frame ‘k+1’ + + +
Reconstructed frame ‘k+1’
Quantized error
Prediction error

7. Proposed Algorithm Input Color Video Extract frames ‘k’ and ‘k+1’
3-step motion estimation
Obtain motion vectors
Forward motion estimation
Predicted video frame -
Predicted video frame
Original frame ‘k+1’ + + +
Reconstructed frame ‘k+1’
Quantized error
Prediction error

8. Three Step Search Method
Input RGB
Video
Extract
Frames
Divide each
Frame into
Blocks of
size 16X16
Divide each
block into
9 equal
parts
Calculate
MSE
Select block
With lowest
MSE/MAD
Divide the
selected
Block into
9 equal
parts
Video Frame
Draw line
connecting
Center of
frame to
this point
Select block
With lowest
MSE/MAD
Calculate
MSE
Divide the
selected
Block into
9 equal
parts
Select block
With lowest
MSE/MAD
Calculate
MSE
16 X 16 Block

9. Block Representation Input Color Video Extract frames ‘k’ and ‘k+1’
3-step motion estimation
Obtain motion vectors
Forward motion estimation
Predicted video frame -
Predicted video frame
Original frame ‘k+1’ + + +
Reconstructed frame ‘k+1’
Quantized error
Prediction error

10. Predicting Next Frame Frames ‘k’ and ‘k+1’ Motion Vectors
Predicted Frame ‘k+1’

11. Block Representation Input Color Video Extract frames ‘k’ and ‘k+1’
3-step motion estimation
Obtain motion vectors
Forward motion estimation
Predicted video frame -
Predicted video frame
Original frame ‘k+1’ + + +
Reconstructed frame ‘k+1’
Quantized error
Prediction error

12. Prediction Error Calculation
Frame 60
Frame 61
Prediction error + -
Predicted Frame

13. Results Motion Vectors Predicted frame Color video
3-step motion estimation
Forward motion estimation
Motion Vectors
Predicted frame
Color video
Extracted frames ‘k’ and ‘k+1’ -
Predicted frame +
Frame ‘k+1’ + +
Quantized error
Reconstructed video frame
Prediction error

14. Conclusion Advantages:
Simplicity: Simple geometric transformation of pixel co-ordinate.
Easy to implement in hardware
Limitations:
Fails for zoom, rotational motion, and under local deformations.

15. References
[1] H. Gharavi and M. Mills, “Block-matching motion estimation algorithms: New results,” IEEE Trans. Circ. and Syst., vol. 37, pp , 1990.
[2] V. Seferidis and M. Ghanbari, “General approach to block-matching motion estimation,” Optical Engineering, vol. 32, pp , July 1993.
[3] M. Bierling, “Displacement estimation by hierarchical block-matching,” Proc. Visual Comm. and Image Proc., SPIE vol. 1001, pp , 1988.
[4] B. K. P. Horn and B. G. Schunck, “Determining Optical Flow,” Artif. Intell., vol. 17, pp , 1981.
[5] S. V. Fogel, “Estimation of velocity vector fields from time varying image sequences,” CVGIP: Image Understanding, vol. 53, pp , 1991.
[6] T. S. Huang, ed., Image Sequence Analysis, Springer Verlag, 1981.
[7] A. V. Oppenheim and R. W. Schafer, “Discrete - Time Signal Processing,” Prentice Hall Signal Processing Series, 1989.
[8] A. M. Tekalp, “Digital Video Processing,” Prentice Hall Signal Processing Series, 1995.
[9] D. E. Dudgeon, “Multidimensional Digital Signal Processing,” Prentice Hall Signal Processing Series, 1996.
[10] K. Sayood, “Introduction to Data Compression,” Morgan Kaufmann Publishers, 2006.

16. Thank You