Sparse representation based iterative incremental image deblurring
Haichao Zhang, Yanning Zhang · 2009
Inspired by the observation that in image restoration, parametric models are extremely specific while pixel-level models are too loose, tending to under or over fit the underlying image respectively, in this paper, we proposed an ‘intermediate-language’ based method for image deblurring. The solution space is represented at a level higher than the pixel-grid level, while retain an enough degree of freedom (DOF), thus avoids the common local under or over fitting problem. Considering the sparseness property of images, a sparse representation based incremental iterative method is established for blurry image restoration. Comprehensive experiments demonstrate that the framework integrating the sparseness property of images significantly improves the deblurring performance.