High Quality Image Deblurring Scheme Using the Pyramid Hyper-Laplacian L2 Norm Priors Algorithm

In this paper, a very effective image deblurring algorithm is proposed. It combines the techniques of the pyramid structure, the Hyper-Laplacian model, the hybrid norm priors for gradients, and the half-quadratic penalty method. In image deblurring, it is always a tradeoff between the goals of making the edge part sharp and reducing the ringing and the noise effects in the non-edge part. However, using the proposed algorithm, the two goals can be achieved at the same time. Many state-of-art image deblurring algorithms take the gradient of the reconstructed image into account to reduce the effect of noise. In the proposed algorithm, since the pyramid structure and the hybrid norm are applied, the image deblurring performance can be further improved. Simulations show that the proposed algorithm can successfully reconstruct the original image and outperforms the state-of-art methods for image deblurring.