a hybrid edge-enhanced motion adaptive deinterlacer
DESCRIPTION
A Hybrid Edge-Enhanced Motion Adaptive Deinterlacer. By Marc Ramirez. Frame 1. Problem Statement & Motivation. Field A. Field B. Field C. Canon GL1 “Frame Mode” ~ 320 Vertical Lines Deinterlace from 60i into a 30p sequence - PowerPoint PPT PresentationTRANSCRIPT
A Hybrid Edge-Enhanced A Hybrid Edge-Enhanced Motion Adaptive DeinterlacerMotion Adaptive Deinterlacer
By Marc Ramirez
• Canon GL1 “Frame Mode” ~ 320 Vertical Lines• Deinterlace from 60i into a 30p sequence
• High Quality, Noise Reduction, Edge & Detail Preservation, Moderate Complexity
• MC Recursive, VT Median, EDDI, BOB• Best Approach Depends on Material
Field A Field B Field C
Frame 1
Problem Statement & MotivationProblem Statement & Motivation
Initially Proposed MethodInitially Proposed Method
Simonetti
de Haan
Actual MethodActual Method
Current DV Capture Cards Import 60i Sequences as Field-Merged 30p
1) Store First Three Fields (A,B,C) from Captured Frames1 & 2
2) Look at the SAD of Fields A & C, If <= Thresh1 -> Keep Original Frame *
Ex. Mounted Camera Recording a Stationary Object; White Background
3) Detect Edges
*Could Potentially Cause Problems
Two Types of Edge Detection
• EDDI Horizontal Emphasis
• Canny Method in Matlab Edge Function
- Smoothing By Gaussian Convolution
- 2D Derivative
- Ridge Tracking of Gradient Magnitude
4) Interpolate Along Found Edges
- Step Through Known Lines Only
- Pick a Test Block of Correct Length
- Use SAD to Determine Best Match
- If <= Thresh2 -> Interpolate
- Use Nearest Neighbor if Between Pixels
Known
Known
Known
5) Fill In Static Areas
IF
AND
Fill In With Previous or Average of Pixel P&N
3ThreshLUMLUM NP
32
)(
2
)(Thresh
LUMLUMLUMLUM EBNP
44/ ThreshLcLaKcKaJcJaIcIa
6) Detect If Slow Pixel Motion
7) Use Median Filter on Small Window B = SUM/|DIFF| for ( 4 Combinations)
Med{E[A,F] E[B,E] E[C,D] E[G,H] lowB}
8) Spatially Interpolate Remaining High-Motion Pixels
• 4 Tap Vertical Filter for Better Frequency Response
• Might Also Include a Horizontal Component
Conclusion/Future ChangesConclusion/Future Changes
• Overall the Implementation is Less Computationally Expensive than MC with Pretty Nice Results
• The Algorithm Tries to Use the Proper Method Based on Simple Motion Detection
• Many Threshold Parameters -> Difficult to Set the Correct Thresholds for All Cases
• Could Later Implement EDDI Correctly on the Final Image
• Future Method Could Incorporate Motion Estimation
• Implement a Plug-in For Virtual Dub or AVISynth
ReferencesReferences
[1] R. Simonetti, S. Carrato, G. Ramponi and A.Polo Filisan, 'Deinterlacing of HDTV Images for Multimedia Applications', in Signal Processing of HDTV, IV, E. Dubois and L. Chiariglione, Eds., Elsevier Science Publishers, 1993, pp. 765-772.
[2] G. de Haan and E.B. Bellers, ‘Deinterlacing -- An overview', Proceedings of the IEEE, Vol. 86, No. 9, Sep. 1998, pp. 1839- 1857.
[3] G. de Haan and R. Lodder, `De-interlacing of video data using motion vectors and edge Information', Digest of the ICCE'02, Jun. 2002, pp. 70-71.
[4] G. de Haan, `Video processing for multimedia systems', ISBN: 90-9014015-8, Eindhoven Sep. 2000.
[5] Y. Wang, J. Ostermann, and Y.Q. Zhang, ‘Video Processing and Communications’ Prentice Hall, 2002, ISBN 0-13-017547-1.