Mixed impulse and Gaussian noise removal using detail-preserving regularization

摘要

Abstract. Over the years, numerous methods have been proposednseparately for restoring images corrupted by either impulse noise ornGaussian noise. Nevertheless, because of the distinct nature of bothntypes of degradation processes, not much work has been developed toneffectively remove mixed noise from images, a problem that is commonlynfound in practice. To alleviate this problem, we propose a two-stage ap-nproach based on impulse detectors and detail-preserving regularization.nWe employ the detectors to identify impulse noise, and then restorenthem and smooth the remaining Gaussian noise simultaneously basednon regularization framework. A novel error norm that can adaptively mim-nick traditional l1 and l2 norms is used in the regularization process. Thisnadaptivity enables our approach to be universally capable of removingnvarious degrees of impulse noise and mixed noise, while preserving finenimage details well. Extensive experiments have been conducted to testnthe proposed approach and shown its improvements over the algorithmsnexisting in the literature. © 2010 Society of Photo-Optical Instrumentation Engineers.nu0003DOI: 10.1117/1.3485756