Combination of l 1 and l 2 norms for image deconvolution problems
Kemal Özkan, Erol Seke · International Conference on Electrical and Electronics Engineering · 2011
Maintaining a balance between fidelity, over-smoothing and stability in iterative restoration of noisy and blurred images is a difficult task that researchers need to handle. We proposed a very simple yet effective method to combine l 1 and l 2 norms in order to both reduce over-blurring around edges and perform effective denoising within flat areas in images simultaneously and generate visually pleasing results. Experiments performed with tuned parameters for all compared approaches shown that the proposed combination is also the fastest to converge. A strategy for updating feedback parameters within iterations is also experimentally developed and used.