Deskewing by space-variant deblurring

Karthik Seemakurthy    Subeesh Vasu    A.N. Rajagopalan    

Abstract

Skew and motion blur are significant challenges when camera and scene of interest are in two different media. Skew occurs due to spatially varying refraction on a dynamic water surface, whereas motion blur results from multiple intensities impinging on the imaging sensor during camera exposure time due to time varying refraction. In this paper, we introduce the notion of virtual depth map which we assign to a planar scene when observed through a dynamic water surface and transform the deskewing problem into one of space-variant deblurring from a single image within an alternating minimization framework. Since the nature of the virtual depth map can change during exposure due to change in wind properties, we also propose a shot detection framework to identify segments of frames from the captured video which conform to a single virtual depth map. While the overall wave motion can be arbitrary, within each segment the nature of wave is modeled as an exponentially decaying periodic wave.

Examples

Synthetic

Blurred image	Tian et al. ICCV 2009	Oreifej et al. CVPR 2011	Xu et al. CVPR 2013	Ours

Real

Blurred image	Tian et al. ICCV 2009	Oreifej et al. CVPR 2011	Xu et al. CVPR 2013	Ours

---

Technical Paper, Supplementory, and Poster

Karthik Seemakurthy, Subeesh Vasu, A.N. Rajagopalan, "Deskewing by space-variant deblurring", British Machine Vision Conference (BMVC), 2016    Paper       Supplementary       Poster       BibTex    References [1] Yuandong Tian and Srinivasa G Narasimhan. "Seeing through water: Image restoration using model-based tracking", ICCV 2009. [2] Omar Oreifej, Guang Shu, Teresa Pace, and Mubarak Shah. "A two-stage reconstruction approach for seeing through water", CVPR 2011. [3] Li Xu, Shicheng Zheng, and Jiaya Jia, "Unnatural l0 sparse representation for natural image deblurring," CVPR, 2013.