Robust space-variant super-resolution
Kaggere V. Suresh, Ambasamudram Narayanan Rajagopalan · 2006
In this paper, we address the problem of reconstruction of a high-resolution image from a number of sub-pixel shifted and space-variant blurred low-resolution observations. We use a maximum a posteriori-Markov random field (MAP-MRF) based approach for recovering the high-resolution image. The high-resolution image is modeled as MRF and is estimated by using a deterministic algorithm called iterated conditional modes (ICM) which maximizes the local conditional probabilities sequentially. We show that modeling by MRF lends robustness to errors in estimation of motion and blur parameters. A discontinuity adaptive prior image model is used to preserve edges. We also derive the exact posterior neighborhood structure in the presence of warping, blurring (space-variant) and down-sampling operations. The locality of the posterior distribution is critical for computational efficiency. Experimental results are given to demonstrate the effectiveness of our method.