Robust Framework For Digital Image Denoising And Deblurring
Loading...
Date
2012-06
Authors
Toh, Kenny Kal Vin
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
Image restoration concerns improving visual quality of a captured image that goes beyond the
achievable limit of camera. Recent advancement in imaging and multimedia technology has
advocated the interests of image restoration through software, of which applications permeate
consumer photography as well as different industries. Unfortunately, the captured images
often suffer from degradations, such as blurring, noise, unpleasant artifacts, and more, due
to limitations of the imaging system. Despite considerable efforts have been channeled to
advance the state-of-the-art methods, surprisingly, these methods are often slow and only
designed for handling specific degradation model. As such, the existing methods usually fail
when applied to degraded real images. Based on this motivation, a robust framework is proposed
to address the main issues related to designing practical image restoration methods,
namely, visual restoration quality and computational complexity. The robust framework has
several advantageous properties: (1) the proposed framework is robust towards the presence
of data uncertainties, (2) it is spatially adaptive to the radiometric structures of the image
data, and (3) it is exceedingly robust in capturing the local structural information even
in noise-ridden images. By capitalizing on the advantages of this robust framework, three
novel image restoration methods have been developed in this work. The first method, termed
as Augmented Variational Series and Histogram-based Clustering (AVSHC), is a switchingscheme
filter that is capable to remove any kind of impulsive noise on color or monochrome
images. Then, two variants based on a robust method, called Locally Adaptive Bilateral Clustering (LABC), are proposed for image denoising, mild deblurring, and sharpness enhancement.
Description
Keywords
Image restoration concerns improving , visual quality of a captured image