Upload
gyala
View
54
Download
0
Tags:
Embed Size (px)
DESCRIPTION
NORTHWESTERN UNIVERSITY. Motion from Blur. Shengyang Dai and Ying Wu EECS Department, Northwestern University. Discrete samples over time. Integral over time. Optical flow estimation. Motion from blur. ?. http ://vision.middlebury.edu/flow / - PowerPoint PPT Presentation
Citation preview
Motion from BlurShengyang Dai and Ying Wu
EECS Department, Northwestern University
NORTHWESTERNUNIVERSITY
http://vision.middlebury.edu/flow/Baker, Scharstein, Lewis, Roth,
Black, Szeliski, ICCV’07(courtesy to Simon Baker)
Motion from blur
Optical flow estimation
?
Discrete samples
over time
Integral over time
LiteratureTask Input Extra info
Q. Shan, W. Xiong, and J. Jia, CVPR 07 Rotational motion blur
Single image
User interaction
A. Levin, NIPS 06
Multiple / local invariant linear motion estimation and
segmentation
Known single blur direction
S. Cho, Y. Matsushita, and S. Lee, ICCV 07 Two images
XL. Bar, B. Berkels, M. Rumpf, and G. Sapiro,
ICCV 07Sequence
Space-variant linear motion estimation from blurred image(s)
Our workTask Input Extra info
Q. Shan, W. Xiong, and J. Jia, CVPR 07 Rotational motion blur
Single image
User interaction
A. Levin, NIPS 06
Multiple / local invariant linear motion estimation and
segmentation
Known single blur direction
S. Cho, Y. Matsushita, and S. Lee, ICCV 07 Two images
XL. Bar, B. Berkels, M. Rumpf, and G. Sapiro,
ICCV 07Sequence
Our work
1. Global parametric form motion from blur (e.g., affine / rotational motion)
2. Multiple / local motion estima-tion and segmentation from blur
3. Non-parametric motion from blur
Single image X
Space-variant linear motion estimation from blurred image(s)
)2
()2
(b
pb
pbpb III
+
-b
pb bI
P
pbI
Motion blur constraint
=( ) ·
)2
()2
(b
pb
pbpb III
Motion Blur vs. Optic Flow
)2
()2
(b
pb
pbb III
tII m
• Lucas and Kanade, 81• Horn and Schunck, 81• Barron, Fleet, Beauchemin, IJCV 94• Black and Anandan, CVIU 96• Baker and Matthews, IJCV 04• Baker, Scharstein, Lewis, Roth, Black,
Szeliski, ICCV 07• ……
?
channel image representation
Spectral matting Levin, Rav-Acha, and Lischinski, CVPR 07
Unsupervised image layer decomposition
(courtesy to Anat Levin)0 1
Alpha motion blur constraint
)2
()2
(b
pb
pbb III )2
()2
(b
pb
pbb 1 bb0
1
b b
0 bb 0 b
• Use to replace the original image
• Assumptions: most pixels have 0 / 1 alpha values
• Observation: mostly when
,b II ,b
0 bwhen
motion blurred imagebI
1 bb
Alpha motion blur constraint
1
0
b1 bb
Alpha motion blur constraint
b
Alpha channel of the blurred imageb
Motion Blur vs. Optic Flow
1 bb )2
()2
(b
pb
pbb III
tII m
• Lucas and Kanade, 81• Horn and Schunck, 81• Barron, Fleet, Beauchemin, IJCV 94• Black and Anandan, CVIU 96• Baker and Matthews, IJCV 04• Baker, Scharstein, Lewis, Roth, Black,
Szeliski, ICCV 07• ……
?
Hough transform
b
1 bb
Techniques for optical flow
• Lucas-Kanade• Horn-Schunck• Robust estimation• RANSAC• Multiple resolution
• Parametric• Piecewise smooth• Segmentation• LK meets HS• ……
Techniques for motion from blur
• Lucas-Kanade• Horn-Schunck• Robust estimation• RANSAC• Multiple resolution
• Parametric• Piecewise smooth• Segmentation• LK meets HS• ……
Tasks
Global affine motion blur
Global rotational motion blur
Multiple blur model estimation and segmentation
Non-parametric motion field estimation
Experiments – affine motion
Experiments – rotational motion
Experiments – segmentation
Segmentation result
Experiments – non-parametric motion
Applications
Ground truth image
Synthesized affine blurred image
(PSNR: 21.66dB)
Deblurred result (PSNR: 24.75dB)
Image deblurringModified Richardson-Lucy iteration
average estimation error: 0.43average motion vector: 15.43
Affine blurred image Our deblur result with• space-variant blur• modified RL iteration
Matlab deconvlucy with • space-invariant blur• original RL iteration
time Input
Deblurred
Blur synthesis
Motion synthesis
Blur / motion synthesis
Blur synthesis
Input Output
Motion synthesis
Input Output
Motion synthesis
Summary• Contributions
– A local linear motion blur constraint– Connection between motion blur and optic flow– Space-variant motion estimation from blur– Applications on deblurring and blur/motion synthesis
• Limitations– May not hold for heavily textured region– Rely on robust matting
• Future work– Integrating more algorithms from optic flow
Motion from BlurShengyang Dai and Ying Wu
EECS Department, Northwestern Universityb
1 bb
Thanks! Questions?