Journal of Engineering Science and Technology Review 8 (5) (2015) 41-48
Review Article
A Review of Image Denoising Methods
I. Irum*, M. A. Shahid, M. Sharif and M. Raza
COMSATS Institute of Information Technology Wah Cantt, Pakistan
Received 10 August 2015; Accepted 19 December 2015
___________________________________________________________________________________________
Abstract
Image Denoising is one of the fundamental and very important necessary processes in image processing. It is still a
challenging and a hot problem for researchers. Images are one of essential representations in every field like education,
agriculture, geosciences, aerospace, surveillance, entertainment etc by means of electronic or print media. Images can get
corrupted by noise, there has been a great research effort which made solutions for this problem, a number of methods
have been proposed. An overview of various methods is given here after a brief introduction. These methods have been
categorized on the bases of techniques used.
Keywords: Derivative Based Denoising, Fuzzy Based Denoising, Mathematical Morphology Based Denoising, Median Based
Denoising, Nonlinear Denoising methods, s Statistical Modeling Based Denoising
__________________________________________________________________________________________
1. Introduction
Various sources let the digital images to be corrupted by
poor contrast and noise, theses sources include image
transmission, acquisition, compression [1-5], quantization,
illumination conditions [6], malfunctioning instruments, ill
positions etc. These sources directly degrade the visual
quality of image during processing of image [7]. Process of
image denoising or image restoration is targeted to estimate
the original image from corrupted image. It is still the most
fundamental, largely unsolved and widely studied problem
[8]. Image has important information and certain details, as
communication through visual images is an integral part of
modern life. Images and videos are used everywhere to
communicate as a visual source. Noise affects image
features seriously like edges, thin lines, and loss of image
details provides degradation of spatial resolutions. Therefore
noise removal from corrupted images is very important and
necessary before further processing on them like
segmentation [9-12], feature matching, edge detection,
feature extraction [13-15], feature detection [16] of image
details used for face recognition [17-38] etc, These denoised
images can be used for face detection [39-41], content based
image retrieval [42-48], medical image reconstruction [49],
understanding Morphology of medical images [50] with its
applications [51], medical image enhancement [52], image
rendering [53] etc.
A huge number of methods have been proposed in this
context over the past decades in image processing. A review
of these methods has been given here covering the
representative methods which gave better performance in
this area. These methods have been categorized on the bases
of their natures of techniques used in. These categories are
Median Based, Statistical Modeling Based, Derivative Based,
Fuzzy Logic Based and Mathematical Morphology Based
image denoising methods. These categories are discussed
one by one in upcoming section of rest of the paper and
conclusion is given at the end.
2. Median Based Image Denoising Methods
Median Based Filters or Denoising Methods are the corner
stones of image cancellation methods in modern image
processing. Tukey [54] first introduced the Median Filter,
after its inception, tremendous efforts have been made for
optimization, improvement and refinement over the years.
Standard median filters presented in [55] treated all the
pixels of the image whether corrupted or uncorrupted. [56]
To overcome this drawback Weighted Median Filters
(WMF) [57] and Switching Median Filter (SMF) [58] were
proposed. WMF reduced the smoothing effect, preserved the
image sharpness and treated the entire pixels like standard
median filter by giving higher weights to the central pixel
[59-60]. Weighted Order Statistics Median Filters (WOSFs)
were proposed in [61]. Design of WMF admitting negative
and complex weights presented in [62-65]. Steerability
concept was introduced in [66] and its application found in
[67]. Recently Dimitrios Charalampidis [68] proposed
steerable WMF inheriting the noise robustness and edge
preserving capabilities of WMF.
SMF reduced the number of pixels subjected to filtration
by identifying the corrupted and uncorrupted pixels and
leaving the uncorrupted pixels unchanged. In [69-80]
extensions of median filter and SMF have been presented.
Recent examples of noise adaptive approaches can be found
in [81-83]. The parameters that can be used for as input
function to adaptive approach can be window size, shape or
rank [84]. It created blurred images when applied to mix or
Gaussian Noise, Directional Weighted Median (DWM) [85]
Jestr
JOURNAL OF
Engineering Science and
Technology Review
www.jestr.org
____________
* E-mail address: ismairum@gmail.com
ISSN: 1791-2377 © 2015 Kavala Institute of Technology.
All rights reserved.