BLINDSUPER-RESOLUTIONIMAGE RECONSTRUCTION USINGA MAXIMUM A POSTERIORIESTIMATION
Lists
Only you see this
Abstract
Thispaper proposes anewalgorithm toaddress blind image super-resolution byfusing multiple blurred low-resolution (LR)imagestorender a high-resolution (HR)image. Conventional super-resolution (SR)imagereconstruction algorithms assumeeither theblurring during theimage formation process isnegligible ortheblurring function is knownapriori. Thisassumption, however, isimpractical as itisdifficult toeliminate blurring completely insome applications orcharacterize theblurring function fully. In viewofthis, we present a newmaximuma posteriori (MAP)estimation framework thatperforms joint blur identification andHR imagereconstruction. An iterative scheme basedonalternating minimization isdeveloped to estimate theblurandHR imageprogressively. A blurprior thatincorporates thesoft parametric blurinformation and smoothness constraint isintroduced intheproposed method. Experimental results showthat thenewmethod iseffective inperforming blind SRimagereconstruction wherethere is limited information about theblurring function. Indexterms-Image resolution, imagerestoration, MAP estimation, conjugate gradient methods.
BibTeX
@inproceedings{yap2006blindsuper,
author = {Kim–Hui Yap},
title = {{BLINDSUPER-RESOLUTIONIMAGE RECONSTRUCTION USINGA MAXIMUM A POSTERIORIESTIMATION}},
booktitle = {ICIP},
year = {2006},
url = {https://openreview.net/forum?id=VW8TahjyRN2},
}Loading related content...