Abstract
Deblurring images of
moving objects captured from a
traditional camera is an ill-posed problem due to the loss of high
spatial frequencies in the captured images. Recent techniques have
attempted to engineer the motion point spread function (PSF) by either
making it invertible using coded exposure, or invariant to motion by
moving the camera in a specific fashion. We address the problem
of optimal single image capture strategy for best deblurring
performance. We formulate the problem of optimal capture as maximizing
the signal to noise ratio (SNR) of the deconvolved image given a scene
light level. As the exposure time increases, the sensor integrates more
light, thereby increasing the SNR of the captured signal. However, for
moving objects, larger exposure time also results in more blur and
hence more deconvolution noise. We compare the following three single
image capture strategies: (a) traditional camera, (b) coded exposure
camera, and (c) motion invariant photography, as well as the best
exposure time for capture by analyzing the rate of increase of
deconvolution noise with exposure time. We analyze which strategy is
optimal for known/unknown motion direction and speed and investigate
how the performance degrades for other cases. We present real
experimental results by simulating the above capture strategies using a
high speed video camera.
Paper (Preprint)
Motion Blur Datasets and
Matlab/C
codes
Related Papers in Motion/Focus Deblurring
SIGGRAPH 2006 Coded exposure for motion deblurring
SIGGRAPH 2007 Coded
aperture and a new theory of light
field capture
SIGGRAPH 2009 Invertible
Motion Blur in Video
CVPR 2007
Simultaneous motion
deblurring and super-resolution
CVPR 2009
New codes for coded
exposure deblurring to help in PSF Estimation
Figure 1. (Left) We
consider object motion in terms of magnitude and
direction. If both magnitude and direction are known, trivial capture
strategy is to move the camera in that direction with that magnitude.
Both coded exposure and MIP will also work well. If direction is
unknown and magnitude is known, coded exposure should be used (left
col). As magtnitude changes, the performace of coded exposure degrades
gradually. If direction is known and magnitude is unknonw, MIP should
be used However, as the direction changes, MIP performace
degrades sharply.
Key Results:
1. For traditional camera, increasing exposure time degrades SNR on moving objects, if photon noise is taken into accout. But for coded exposure and MIP, one can increase exposure time without SNR degradation.
2. If the direction of the motion is known exactly, but magnitude is not known, one should use the idea of motion invariant photography (MIP) to move the camera along that direction in accelerating fashion.
3. Coded exposure works for any motion direction. If magnitude is known, but direction is unknown, coded exposure gives best results.
4. Performance Generalization
a. For Coded Exposure: If the magnitude of motion (blur size) differs from code size, no deconvolution artifacts are produced. Only the minimal resolvable blur size is increased. For example, if a code of 50 is used and the blur is 100 pixels, one can get deblurring results where the object is blurred only 2 pixels, without any deconvolution artifacts. Thus, performace of coded exposure degrades gradually.
b. For MIP: If the direction of motion differs from camera motion direction, MIP starts producing deconvolution artifacts and the PSF does not remain invariant any more.
c. For MIP: As the magnitude of motion increases from assumed maximum in MIP, the PSF becomes more flatter (closer to a box function) and thus MIP performace degrades rapidly.
References
1. Coded Exposure: Ramesh Raskar, Amit Agrawal and Jack Tumblin, Coded Exposure Photography: Motion Deblurring using Fluttered Shutter, SIGGRAPH 2006
2. MIP: A. Levin, P. Sand, T. S. Cho, F. Durand, W. T. Freeman. Motion-Invariant Photography. SIGGRAPH 2008.
1. For traditional camera, increasing exposure time degrades SNR on moving objects, if photon noise is taken into accout. But for coded exposure and MIP, one can increase exposure time without SNR degradation.
2. If the direction of the motion is known exactly, but magnitude is not known, one should use the idea of motion invariant photography (MIP) to move the camera along that direction in accelerating fashion.
3. Coded exposure works for any motion direction. If magnitude is known, but direction is unknown, coded exposure gives best results.
4. Performance Generalization
a. For Coded Exposure: If the magnitude of motion (blur size) differs from code size, no deconvolution artifacts are produced. Only the minimal resolvable blur size is increased. For example, if a code of 50 is used and the blur is 100 pixels, one can get deblurring results where the object is blurred only 2 pixels, without any deconvolution artifacts. Thus, performace of coded exposure degrades gradually.
b. For MIP: If the direction of motion differs from camera motion direction, MIP starts producing deconvolution artifacts and the PSF does not remain invariant any more.
c. For MIP: As the magnitude of motion increases from assumed maximum in MIP, the PSF becomes more flatter (closer to a box function) and thus MIP performace degrades rapidly.
References
1. Coded Exposure: Ramesh Raskar, Amit Agrawal and Jack Tumblin, Coded Exposure Photography: Motion Deblurring using Fluttered Shutter, SIGGRAPH 2006
2. MIP: A. Levin, P. Sand, T. S. Cho, F. Durand, W. T. Freeman. Motion-Invariant Photography. SIGGRAPH 2008.