Registration of Large Motion Blurred CMOS Images
[Technical Report, Final Report]
INDIAN INSTITUTE OF TECHNOLOGY MADRAS
Pagination or Media Count:
In the first part of this work, they addressed the coalesced effect of rolling shutter and motion blur that occurs frequently in moving CMOS cameras. First model was generated for a general image formation framework for a 3D scene following a layered approach in the presence of rolling shutter and motion blur. They then developed an algorithm which performs layered registration to detect changes. This algorithm includes an optimization problem that leverages the sparsity of the camera trajectory in the pose space and the sparsity of changes in the spatial domain. The second part of the work addressed the problem of registering images taken from global shutter and rolling shutter RScameras and reveal the constraints on camera motion that admit registration, change detection, and rectification. Their analysis encompassed degradations arising from camera motion during exposure and differences in shutter mechanisms. They also investigated conditions under which camera motions causing distortions in reference and target image can be decoupled to yield the underlying latent image through RS rectification. In the final part of the work, a model for RS blind motion deblurring that mitigates many of the associated issues significantly was proposed. Comprehensive comparisons with state-of-the art methods reveal that their approach not only exhibits significant computational gains and unconstrained functionality but also leads to improved performance.