We show that the union of PSFs corresponding to 2D linear motions occupy a wedge of revolution in Fourier domain, and that the orthogonal parabolic motion paths approach the optimal bound up to a multiplicative constant
Motion blur removal with orthogonal parabolic exposures
Computational Photography, pp.1-8, (2010)
Object movement during exposure generates blur. Removing blur is challenging because one has to estimate the motion blur, which can spatially vary over the image. Even if the motion is successfully identified, blur removal can be unstable because the blur kernel attenuates high frequency image contents. We address the problem of removing ...更多
下载 PDF 全文
- Motion blur can severely limit image quality, and while blur can be reduced using a shorter shutter speed, this comes with an unavoidable tradeoff of increased noise.
- One source of motion blur is camera shake.
- A second source of blur is an object movement in the scene.
- This type of blur is harder to control, and it is often desirable to remove it computationally using deconvolution.
- One needs to estimate the blur kernel, which depends on motion.
- Since objects in the scene can move independently, the blur kernel can vary over the image.
- While single-image based blur estimation techniques have been proposed [7, 8, 11–
- Motion blur can severely limit image quality, and while blur can be reduced using a shorter shutter speed, this comes with an unavoidable tradeoff of increased noise
- For reference we show an image taken with a static camera with 500ms exposure, synchronized to the first shot of the orthogonal parabolic camera
- We present a two-exposure solution to removing spatially variant 2D constant velocity motion blur
- We show that the union of PSFs corresponding to 2D linear motions occupy a wedge of revolution in Fourier domain, and that the orthogonal parabolic motion paths approach the optimal bound up to a multiplicative constant
- We assume that objects move at a constant velocity within the exposure time, which is a limitation shared by most previous work that deals with object motion
- Our image reconstruction takes into account occlusions by allowing some pixels to be reconstructed from a single image, but a full treatment of occlusion for deconvolution remains an open challenge
- The authors built a prototype camera, different from Levin et al , consisting of a sensor, two motion stages and their controllers.
- The authors mounted a light-weight camera sensor on two motion stages, where each can move the camera sensor along orthogonal axes (See Figure 4(a)).
- The authors could replace the motion stages with the image stabilization hardware.
- The authors incur a 100ms delay for switching the control from one motion stage to another, which can be reduced by using an improved hardware
- Figure 5 illustrates the deblurring pipeline.
- The authors capture two images with the detector undergoing a parabolic motion in orthogonal directions.
- The authors estimate a motion map, shown colored using the velocity coding scheme of the inset.
- The authors use the motion map to reconstruct the image.
- For reference the authors show an image taken with a static camera with 500ms exposure, synchronized to the first shot of the orthogonal parabolic camera.
- The refer-
- A flutter shutter camera: In a flutter shutter camera  (Figure 1 second column), the motion spectrum k f is constant along ωx, ωy and is modulated along ωt : kf 2 = m 2, where mis the Fourier transform of the shutter code.
- At low-to-mid frequencies the spectral power does not reach the upper bound.The authors present a two-exposure solution to removing spatially variant 2D constant velocity motion blur.
- The comprehensive study of solutions relying on an arbitrary number of exposures is, an important open question which requires careful modeling of the noise characteristics and the per-shot time overhead
- This research is partially funded by NGA NEGI-1582-040004, by ONR-MURI Grant N00014-06-1-0734, by gift from Microsoft, Google, Adobe, Quanta and T-Party, and by US-Israel Binational Science Foundation
- The first author is partially supported by Samsung Scholarship Foundation
- A. Agarwala, M. Dontcheva, M. Agrawala, S. Drucker, A. Colburn, B. Curless, D. Salesin, and M. Cohen. Interactive digital photomontage. ACM TOG (SIGGRAPH), 2004.
- A. Agrawal and R. Raskar. Optimal single image capture for motion deblurring. In IEEE CVPR, 2009.
- A. Agrawal, Y. Xu, and R. Raskar. Invertible motion blur in video. ACM TOG (SIGGRAPH), 2009.
- M. Ben-Ezra and S. K. Nayar. Motion-based motion deblurring. IEEE TPAMI, 26:689 – 698, 2004.
- J. Chen, L. Yuan, C.-K. Tang, and L. Quan. Robust dual motion deblurring. In IEEE CVPR, 2008.
- S. Cho, Y. Matsushita, and S. Lee. Removing non-uniform motion blur from images. In IEEE ICCV, 2007.
- S. Dai and Y. Wu. Motion from blur. In IEEE CVPR, 2008.
- R. Fergus, B. Singh, A. Hertzmann, S. T. Roweis, and W. T. Freeman. Removing camera shake from a single image. ACM TOG (SIGGRAPH), 2006.
- S. W. Hasinoff, K. N. Kutulakos, F. Durand, and W. T. Freeman. Time-constrained photography. In IEEE ICCV, 2009.
- J. Jia. Single image motion deblurring using transparency. In IEEE CVPR, 2007.
- N. Joshi, R. Szeliski, and D. J. Kriegman. PSF estimation using sharp edge prediction. In IEEE CVPR, 2008.
- D. Kundur and D. Hatzinakos. Blind image deconvolution. IEEE signal processing magazine, pages 43 – 64, 1996.
- A. Levin. Blind motion deblurring using image statistics. In NIPS, 2006.
- A. Levin, P. Sand, T. S. Cho, F. Durand, and W. T. Freeman. Motion-invariant photography. ACM TOG (SIGGRAPH), 2008.
- R. Ng. Fourier slice photography. ACM TOG (SIGGRAPH), 2005.
- P. Perez, M. Gangnet, and A. Blake. Poisson image editing. ACM TOG (SIGGRAPH), 2003.
- R. Raskar, A. Agrawal, and J. Tumblin. Coded exposure photography: Motion deblurring using fluttered shutter. ACM TOG (SIGGRAPH), 2006.
- A. Rav-Acha and S. Peleg. Two motion-blurred images are better than one. PRL, 26:311 – 317, 2005.
- Q. Shan, L. J. Jia, and A. Agarwala. High-quality motion deblurring from a single image. ACM TOG (SIGGRAPH), 2008.
- Y.-W. Tai, H. Du, M. S. Brown, and S. Lin. Image/video deblurring using a hybrid camera. In IEEE CVPR, 2008.
- L. Yuan, J. Sun, L. Quan, and H.-Y. Shum. Image deblurring with blurred/noisy image pairs. ACM TOG (SIGGRAPH), 2007.