Adaptive blind deconvolution and denoising of motion blurred images

Abstract

Blur kernel estimation is the vital step in the deblurring process for images. Though, the kernel has no unique solution, deblurring is a severely ill posed problem. The kernel may be point, linear or non linear. In most of the applications, we consider the blur kernel as a linear one, because it is easy to estimate. In this paper, we consider non linear blur kernels, which have more than one motion components. Then, a blind deconvolution technique using piecewise linear model is introduced to estimate the unknown kernels. Furthermore, a denoising technique based on wavelet multiframe decomposition is put together with this deblurring approach, in order to improve the PSNR of the deblurred images. The experimental result shows that, blind deconvolution technique along with wavelet multiframe denoising can improve the PSNR ratios and also the algorithm is very efficient in identifying various blur kernels accurately.

Cite this paper

@article{Raj2016AdaptiveBD, title={Adaptive blind deconvolution and denoising of motion blurred images}, author={N. R. Nelwin Raj and Athira S. Vijay}, journal={2016 IEEE International Conference on Recent Trends in Electronics, Information & Communication Technology (RTEICT)}, year={2016}, pages={1171-1175} }