A recursive resolution-enhancement using multiframe SRR based on meridian filter with Meridian-Tikhonov regularization

The real noise model corrupting the observed images is unknown and usually random statistical model. Consequently, classical SRR (Super Resolution Reconstruction) algorithms using median (L1) and mean (L2) filtering structures may degrade the reconstructed image sequence rather than enhance it. The mathematical analysis [1] demonstrates that the meridian filtering structure exhibits more robust characteristic than that of median (L1) and mean (L2) filtering structures. For applying on images that are corrupted by any noise models at several noise power, a recursive resolution-enhancement using a multiframe SRR is proposed. The stochastic framework (using maximum a posteriori or maximum likelihood estimator) has been applied to the proposed SRR algorithm. The Meridian filter is used for removing outliers in the data and for measuring the difference between the projected estimating of the HR image and each LR image. Due to the ill-pose condition, Tikhonov and Meridian Tikhonov regularization are compulsively incorporated to remove artifacts from the final answer and improve the rate of convergence. In experimental section, numerical experiments are carried out on synthetic data by using the proposed SRR algorithm. Both of the peak signal-to-noise ratio (PSNR) and virtual images are used to measure the quality of an image. The performance of proposed methods compared with other SRR algorithms based on LI and L2 norm is demonstrated on several noise models (such as Noiseless, AWGN, Poisson Noise, Salt&Pepper Noise and Speckle Noise) at different noise power.