“Non-uniform motion deblurring for camera shakes using image registration” by Cho, Cho, Tai and Lee

  • ©Sunghyun Cho, Hojin Cho, Yu-Wing Tai, and Seungyong Lee




    Non-uniform motion deblurring for camera shakes using image registration



    This paper presents a novel blind motion deblurring method for dealing with non-uniform blurs caused by camera shakes. While there are recent works for non-uniform motion deblurring [Whyte et al. 2010; Joshi et al. 2010], those approaches either limit the freedom of camera motions or require special hardware. Our method is based on a novel representation of motion blurs, which models the blur effects using a set of homographies [Tai et al. to appear]. This representation can fully describe the motions of camera shakes in 3D world, which cause non-uniform motion blurs. Our main contribution is the blind motion deblurring algorithm associated with this representation. We solve the ill-posed non-uniform point spread function (PSF) estimation problem by transforming it into a well-posed image registration problem that estimates homographies. To improve the robustness of our method, we use two input images for deblurring. Our method is experimented with both synthetic and real world examples, and produces superior deblurring results compared to previous methods.


    1. Baker, S., and Matthews, I. 2004. Lucas-kanade 20 years on: A unifying framework. International Journal of Computer Vision (IJCV) 56, 3, 221–255.
    2. Joshi, N., Kang, S. B., Zitnick, L., and Szeliski, R. 2010. Image deblurring with inertial measurement sensors. ACM Trans. Graphics 29, 3, 30:1–30:9.
    3. Tai, Y.-W., P. Tan, and Brown, M. to appear. Richardson-lucy deblurring for scenes under a projective motion path. IEEE Trans. Pattern Analysis Machine Intelligence.
    4. Whyte, O., Sivic, J., Zisserman, A., and Ponce, J. 2010. Non-uniform deblurring for shaken images. In Proc. CVPR 2010.

ACM Digital Library Publication:

Overview Page: