See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/305985620

A Review on the Image Sharpening Algorithms Using Unsharp Masking

Article · August 2016

CITATIONS READS

25 9,956

1 author:

Archana J.N
Atria Institute of Technology
6 PUBLICATIONS 31 CITATIONS

SEE PROFILE

All content following this page was uploaded by Archana J.N on 08 August 2016.

The user has requested enhancement of the downloaded file.

ISSN 2321 3361 © 2016 IJESC

Research Article Volume 6 Issue No. 7

A Review on the Image Sharpening Algorithms Using Unsharp

Masking
Archana J.N1, Dr. Aishwarya .P2
Research Scholar1, Professor2
Department of Computer Science,
Bharathiar University, Coimbatore, India1, Atria Institute of Technology, Bangalore, India2

Abstract:
Digitized images usually faces problem of lack of quality, which includes specifically problem with contrast and occurrence of
shading and artefacts which is caused due to the deficiencies in focusing, lighting and various other constraints. Hence the image
quality has to be improved using the basic image enhancement techniques. Image enhancement is the processing of an image to
bring out specific features explicitly. The available methods for image contrast enhancement concentrates on the features of the
image to be processed and eliminating the user characteristics .Image sharpening is one of the prominent image enhancement
techniques applied in every field where images are ought to be understood and analysed In this paper different approaches of
Image sharpening using Unsharp masking(UM) are compared. These algorithms are analysed with different filtering techniques..
The results show that majority of these algorithms are very sensitive to the enhancement factor and can be used for image
sharpening.

Keywords: Image enhancement,Image sharpening, Unsharp Maskin, Discrete Wavelet Transform,Geographic data visualization.

I. INTRODUCTION valuable information suffers from severe background damage.

Image enhancement is considered as one of the most important Few examples of background damages are varying contrast,
techniques in image processing. The main aim of image ancient document age and the documents have degraded over
enhancement emphasizes to enhance the quality and visual time due to storage conditions and the quality of the written
appearance of an image, or to provide a better transform parchment. Image processing offers a few techniques to make
representation for future automated image processing. Many these documents readable. There are few enhancement
images like medical images, satellite, aerial images and also methods which can be applied in such situations. The
real life photographs suffer from poor and bad contrast and enhancement methods are:(a) Image enhancement methods
noise [1].The aim of image enhancement is to raise the using binarization method or thresholding method, (b) Image
dynamic series of the chosen features so that they can be enhancement methods using binarization method or
identified easily and hence it can improve the information in thresholding method and other methods, (c) Image
images for human observers. The characteristic information enhancement methods using other methods only[2].All these
content of the data will not be increased, but it increases the methods are used very often to restore documents. In[3] an
dynamic range of the chosen features so that they can be approach named Histogram Equalization is mentioned.
identified very easily .For example medical image analysis Histogram equalization uses the process of distributing the
and the analysis of images from satellites. Image sharpening gray levels within an image so that every gray level is equally
refers to any enhancement technique which concentrates on likely to occur. Hence histogram equalization increases the
the edges and the fine details of an image. Image sharpening is brightness and contrast of dark and low contrast images. This
mostly used in printing and photographic sectors for approach focusses prominently on contrast enhancement. In
increasing the limited contrast and sharpening the images. In situation were contrast cannot be increased histogram
general if method will be implemented, actual requirements equalization fails. Image sharpening is one among the image
will be obtained. Filtering of images will be done followed by enhancement approach performed on images. In general the
sharpening to get enhanced visual effect for images. Image edges and fine details of images should be preserved since
sharpening methods may be classified into two classes [1]. they are composed basically by high frequency modules. The
1. Spatial domain methods chances for these modules to be degraded in case of these high
2. Frequency domain methods. frequency being removed or attenuated is more. Hence, the
The spatial domain techniques are implemented to the image high frequency modules of an image needs to be enhanced in
plane itself and are based on direct operation of pixels in an order to improve the visual quality of the image. Image
image. Enhancement operations are based on the operations on sharpening refers to any enhancement technique that
the image pixels. Some of the spatial domain image highlights edges and fine details in an image. Image
enhancement operations include gray level transformation, sharpening is widely used in printing and photographic
histogram processing, basic spatial filters and unsharp industries for increasing the local contrast and sharpening the
masking. In the frequency domain methods, the image is first images [4]. Image sharpening can be applied in Medical
transferred in to frequency domain. It means that, the Fourier imaging also like in face recognition [5] where public security
transform of the image is computed first. Image enhancement should be ensured. Usually in face recognition techniques
techniques can be applied to restore the contents of old which were available earlier there were chances for the images
documents. Very often the old documents used for storing to be undefined and the identification rate also gets reduced. In

International Journal of Engineering Science and Computing, July 2016 8729 http://ijesc.org/
this approach mentioned there are much improvement on masking there are still some artifacts in smoother area because
image clarity and complexity as well. of amplification. A variation of the basic UM scheme is
proposed here that contains an adaptive filter in the correction
II. IMAGE SHARPENING PRINCIPLE: path. This approach enhances images whose dynamic range
Image sharpening involves the addition of a signal that is matches with the available dynamic range of a CRT monitor.
proportional to a high- pass filtered version of the original The reason for using the adaptive filter is to emphasize the
image to the original image [3]. The critical factor here is the medium contrast details in the input image more than large-
option of the high pass filtering operation. In the conventional contrast details such as abrupt edges so as to avoid overshoot
approach linear filters were used to implement the high pass effects in the output image. Smooth area will not be
filter. In case of the original image being corrupted with noise considered for sharpening using this filter, and hence the
the linear approaches may provide us with unexpected and overall system is more robust to the presence of noise in the
improper results. input images than traditional approaches. So this approach
provides twin purposes of avoiding noise amplification as well
as unnecessary overshoot in the detail regions for image
enhancement [10].

3.3 Image Sharpening using Unsharp Masking and

Wavelet Transform
The basic idea of image sharpening is to add to the input
signal a high pass filtered version of the original signal itself.
Wavelet coefficients provide multiresolution high frequency
details of an image. Using this concept, a wavelet based
approach was used for image sharpening in [11].Image
Fig(1) Image Sharpening principle
sharpening is introduced to improve image contrast and
brightness. In another method a combination of both Discrete
The Fig(1) illustrates this procedure called an unsharp
Wavelet Transform (HAAR) and Unsharp Masking technique
masking on a one-dimensional signal. As shown in the is used [12]. The wavelet coefficients provide high frequency
Fig(1),the original image is first filtered by a high-pass filter coefficients of the image such as edge information. Unsharp
which removes the high-frequency components, and then a masking technique is used to sharpen the image here. This
scaled version of the high-pass filter output is added to the approach is used to obtain edge information. This algorithm
original image and hence a sharpened image of the original uses the correlation between different wavelet coefficients and
will be produced. hence high frequency coefficients are considered to be edge of
an image. When 2D-DWT is applied to the image the
III. IMAGE SHARPENING ALGORITHMS USING approximation coefficients and detail coefficients such as
UNSHARP MASKING: Vertical, Horizontal and Diagonal represented as CA, CV,CH
3.1 Image Sharpening using unsharp masking and CD respectively. As the next stage, each of these
Image sharpening is done using unsharp masking for contrast
coefficients are processed with UM. finally the new
enhancement. The basic concept of UM is to blur the original
coefficient’s CA’, CV’, CH’ and CD’ are obtained. Image
image first, then subtract the blurred image from the original sharpening was calculated using the amount of rise in the
image itself. As the final stage add the difference to the value parameter. It was observed that there was a 7.47% rise
original image. The linear unsharp filtering approach is used in the value parameter in the original image and it was
to enhance the noisy image using the high pass filter. Unsharp observed that after using this approach there was 30.59% rise
masks are very applicable for sharpening images. But too in the value parameter [12].
much sharpening can also leads to artificiality in the image A simple image sharpening algorithm (WUM) which uses
losing its natural look. This method has two major drawbacks both Wavelet and UM approaches is mentioned in [11]. In this
such as the contrast in the darker area is enhanced much approach WUM was focused on grey scale images. Here the
deeper than the lighter area. Next problem is the method also
edge information of the original image was taken from a list of
enhances the noise and digitization effects. Due to these issues wavelet coefficients. The wavelet coefficients were obtained as
the images looses its originality in most cases [6].Among all
a result of FDWT applied on the input image. The wavelet
the image sharpening approaches used this UM approach is coefficients obtained ranges from different frequencies. Only
the easiest and simplest. the high frequency coefficients were selected, and the sub
bands with low frequency information were ignored. The
3.2. Image Sharpening using adaptive Unsharp selection of the high frequency coefficients is critical because
masking(UM) some of these frequencies might contain noise. The presence of
Many approaches were suggested for reducing the noise noise in these coefficients hampers the output quality. The
sensitivity of the conventional UM technique. They are usage utilization of WUM using DWT principles is done for
of a quadratic filter as a local mean weighted adaptive high Geographic data visualization [13].The data being visualized in
pass filter[6],and[8].The output noise obtained using these a map could be enhanced using image sharpening techniques.
approaches were less than that was there in the conventional Three geographical images were considered for sharpening. A
unsharp masking approach. A polynomial operator function is
three level DWT was applied to each of these three images and
used in[7][9] where the sharpening action is controlled by the
the sharpness was enhanced using Tenengrad measure.
output of an edge sensor which reduces the work of the high Unsharp masking for contrast enhancement of satellite
pass filter in situation where the processing mask is not images are discussed in [14]. Here nature of wavelet transform
located across an edge in the image. Hence, the system is less that separates the original image into low and high frequency
prone to a noisy input signal. Although the solution used in the sub-band images as low and high pass filter are considered.
above approaches reduce the noise sensitivity of Unsharp Specifically, a number of high frequency sub-band images

International Journal of Engineering Science and Computing, July 2016 8730 http://ijesc.org/
consist of horizontal, vertical, and diagonal coefficients that overall nonlinear behaviour of the enhancement system is very
contain detail of information. Taking inverse wavelet transform easily controlled by a single parameter only. Fuzzy networks
of each sub-band image separately except low frequency one, can effectively model conflicting tasks such as noise removal
the calculation of high frequency information in horizontal, and edge highlighting. Combining fuzzy networks in the same
vertical, and diagonal image can be done. All of them are processing architecture is also very easy. In this approach the
scaled by the scaling factor in each one separately. Adaptive enhancement system includes three networks operating on
algorithm is implemented to results the suitable scaling factor different subsets of input data. However, no complicated
to obtain the enhanced image corresponding with the given tuning of fuzzy set parameters is necessary because the overall
criterion based on variance of each area smooth and detail area. nonlinear behavior is very easily controlled by one parameter
Based on the experimental results obtained as a result of this only. Results of computer simulations have shown that the
approach it is observed that it achieves well to high enhance in proposed method outperforms state-of-the-art techniques in
detail area and low in smooth area. the enhancement of noisy image data.

3.4 Image Sharpening using Unsharp masking the depth 3.6 Rational Unsharp Masking Technique: Unsharp
buffer: masking was introduced in image processing to make the
This approach is applied to enhance the perceptual features of image crisper[18].As shown in the fig(1) in unsharp masking
images that contain depth information [15].Like UM, the approach a high pass filter is used to sharpen the image and
difference between the original depth buffer content and a also a scaling factor is used. But still it has few drawbacks
low-pass filtered copy is used to determine information about such as noise sensitivity and excessive overshoot on sharp
spatially important areas in a scene. Depending on these data details. Linear Unsharp masking approach used for image
the contrast, color and the other factors of the image are enhancement is modified with few changes here. A control
enhanced. The objective of this approach is to improve the term is expressed as a rational function of the local input data
perception of complex scenes by introducing additional depth in this approach which is the basic parameter used for image
cues. The basic principle used here is, a mask is created by enhancement. The noise amplification or noise sensitivity as
subtracting a low-pass filtered copy from the original image mentioned earlier as a drawback of unsharp masking which is
which effectively produces a high-pass filter. This final high caused during image sharpening can be eliminated as a result
frequency signals are added to the original image to obtain a of this parameter. Also the overshoot effects on the sharp
sharpening or a local contrast enhancement. For producing edges can be reduced at the end of image enhancement.
more effect, a larger filter kernel size can be used which
results in an enlarged radius of influence. The approach can be 3.7 Nonlinear Unsharp Masking Algorithm For
applied for conventional image data without considering depth Mammography:
information, and as a result, enhances all local contrasts in an A new unsharp masking (UM) scheme, called nonlinear UM
image. Still, by applying this technique to the depth buffer, (NLUM), is used for mammogram enhancement. The NLUM
which means we compute the difference between the original provides few benefits for users which are 1) to embed
and the low-pass filtered depth buffer. Using this an additional different types of filters into the nonlinear filtering operator;
fact is identified which is, the spatial relation between the 2)to choose different linear or nonlinear operations for the
objects in the scene.In case of an image that contains a blue fusion processes that combines the enhanced filtered portion
object in front of a blue background, using the conventional of the mammogram with the original mammogram; and 3) to
unsharp mask the image cannot be altered due to the lack of allow the NLUM parameter selection to be performed
color contrast. This approach enables to alter the colors in this manually or by using a quantitative enhancement measure to
scenario due to the additional depth information. obtain the optimal enhancement parameters. Also a new
enhancement measure method, called the second-derivative-
3.5 Image Enhancement technique combining Sharpening like measure of enhancement is used here, which is shown to
and noise reduction: have better performance than other measures in evaluating the
There are various reasons for using fuzzy logic in image visual quality of image enhancement. The comparison and
processing which are as follows: firstly, there are various evaluation of enhancement performance demonstrate that the
conditions by which an image has to travel through, which NLUM can improve the disease diagnosis by enhancing the
includes issues like projection of images in various fine details in mammograms with no a priori knowledge of the
dimensions, digitization of images. Secondly, the ambiguity image contents [19].
associated to boundaries and non-homogeneous regions and In this approach [20] Unsharp Masking (UM) algorithm using
few descriptions like edges, contrast enhancement are very a non-linear enhancement function is introduced. The
common factors in fuzzy. Thirdly, image processing with algorithm tries to overcome the issues with the conventional
deficiencies in data is also a problem and only using sufficient approach. Here the combination of the conventional UM with
understanding of the situation the problem can be solved. The the non-linear enhancement function is implemented. The
basic solution to these issues can be obtained using human conventional UM algorithm is extremely sensitive to noise
language interpretation. But human natural language is because of the presence of the linear high pass filter. The
imprecise and has a fuzzy nature. By considering all the improved high pass filter used in the proposed work provides
factors above, we can conclude that fuzzy set theory is a high frequency components of the image which are insensitive
useful mathematical tool for handling the ambiguity or to noise which reduces the noise sensitivity of the UM
uncertainty[16].]The approach explained in[17] concentrates algorithm. The input image is simultaneously processed using
on the contrast enhancement. It is based on a multiple output the improved high pass filter and the non-linear enhancement
system that accepts fuzzy models in order to avoid the noise function; both the images are then combined to get the final
increase during the sharpening of the image details. It is enhanced image. The Nonlinear enhancement function
considered to provide better performance considered to the overcomes the problem found in the normal enhancement
other approaches implemented .There is no complicated approaches used in medical images. When the fine detail of
tuning of fuzzy set parameters in this approach. Also the ROI in a medical image is enhanced, the noise present in the

International Journal of Engineering Science and Computing, July 2016 8731 http://ijesc.org/
background also will be enhanced. Hence the Nonlinear edges of the masses, but at the same time suppresses the
enhancement function is used which not only enhances the background noise also.

TABLE 1: Comparison of different Unsharp masking approaches for Image sharpening

Image References Filtering Approach used Results Advantages

Sharpening Technique
Algorithm used
1.Image [5][6] High pass Linear Unsharp There is a very Capability of taking into
Sharpening using filter used masking approach good background account the human visual
conventional Un variance achieved system response
sharp Masking at the expense of a Ability to sharpen images
smaller even in the presence of noise
sharpening effect nonlinear approach
2.Image [7][8][10] Linear High Filter controls the Linear UM and Medium contrast details
sharpening using pass filter contribution of the type IB operator enhanced.
adaptive Unsharp used sharpening path in provide good Due to Noise amplification
masking such a way that sharpening of the constraint, cannot be used
contrast enhancement image in the low for online applications.
occurs in high detail contrast details..
area and no /less
sharpening occurs for
smooth areas.
3.Image [11][12][13][1 High-pass Edge information 7.47% rise in the Improve image contrast and
sharpening using 4] filter obtained using the value parameter in brightness.
UM and Wavelet wavelet coefficients. the original and
Transform 30.59% rise in the
value parameter
after sharpening.
4.Image [15] High pass Like UM, the Operation relies Applied to images of any
sharpening using filter difference between solely on color kind, ranging from complex
unsharp masking the original depth and depth buffer landscape
the depth buffer buffer content and a operation. Data and technical artifacts,
low pass filtered copy to volume rendering,
is utilized to photograph, and video with
determine depth information.
information about
spatially important
areas in a scene.
5.Enhancement [16][17] Nonlinear Focusses on multiple Overall nonlinear Better performance
technique Filters output system that behaviour is compared to other
combining accepts fuzzy models controlled by enhancement methods
Sharpening and in order to avoid the single fuzzy set applied in images affected
noise reduction . noise increase during parameter. by Gaussian noise
sharpening. and no complicated tuning
of fuzzy set parameters.
Used in mobile applications
6.Rational [18] Squared band A control term Provides a good Noise amplification is
Unsharp pass filters represented as a noise insensitive avoided and overshoot
Masking rational function of sharpening action effects on sharp edges are
Technique the local input data is without artificial limited.
used. overshoot
artifacts.
7.Unsharp [19][20] Improved Algorithm combines In addition with Provides high frequency
Masking high pass the conventional UM the enhancement components of the image
Algorithm For filters with the non-linear of the edges of the which are insensitive to
Mammography enhancement image the noise.
Using Non- function . algorithm also
Linear suppresses the
Enhancement background noise
Function also.

International Journal of Engineering Science and Computing, July 2016 8732 http://ijesc.org/
 IV. CONCLUSION: Science and Software Engineering (CSSE 2008), Wuhen,
In this paper the different methods for Image sharpening using Hubei, China, Dec 2008, pp. 1053-1056, 2008.
the Unsharp Masking (UM) approaches are analyzed. It was
observed that all algorithms are very sensitive to the [12] SalonikaKansaall, Gurpreet Kaur, Image Sharpening
enhancement factor used for sharpening. using Unsharp Masking and Wavelet Transform,International
The adaptive approach supports the medium contrast details Journal of Advance Research in Computer Science and
enhancement The Wavelet based UM provides good Management Studies, Volume 2, Issue 6, June 2014
improvement in the value parameter percentage and also
improves the contrast and brightness. The Rational approach [13] Archana J.N, Aishwarya P, Geographic Data
provides a good noise insensitive sharpening action without Visualization using Wavelet based Unsharp Masking,
artificial overshoot artifacts. published in IEEE Conference proceedings,
ICCICCT,Nagercoil,Pages:490-494, 2015.
REFERENCES
[1] P. Janani, J.Premaladha and K. S. Ravichandran, Image [14] S. Chitwong, S. Phahonyothing, P. Nilas, and
Enhancement Techniques: A Study, Indian Journal of Science F.Cheevasuvit, Contrast enhancement of satellite image based
and Technology, vol 8(22), September 2015 on adaptive unsharp masking using wavelet transform,
ASPRS 2006 Annual Conference, Reno, Nevada, 2006.
[2] Mohd Sanusi Azmi, Khairuddin Omar, Mohammad
Faidzul Nasrudin, Khadijah,Wan Mohd Ghazali, Review on [15] Thomas Luft, Carsten Colditz, Oliver Deussen,, Image
image enhancement methods of old manuscript with the Enhancement by Unsharp Masking the Depth Buffer, ACM
damaged background”, published in IEEE Conference Transactions on Graphics Volume 25 Issue 3, Pages 1206-
Proceedings International Conference on Electrical 1213 July 2006
Engineering and Informatics (Volume:01 ), Page(s):62 - 67
,Aug. 2009 . [16] Zhiguo Gui, Yi Liu, An image sharpening algorithm
based on fuzzy logic, Optik International Journal for Light and
[3] Sargu, Shashi B., Rana, A Review of Medical Image Electron Optics, Volume 122, Issue 8, Pages 697–702 April
Enhancement Techniques for Image, International Journal of 2011
Current Engineering and Technology, Vol.5, No.2 ,pages
1282-1286,April 2015. [17].F. Russo, Image Enhancement Technique combining
Sharpening and Noise Reduction, IEEE Transactions on
[4]http://www.nptel.ac.in/courses/117104069/chapter_8/8_32. Instrumentation and Measurement ,Volume 51,Issue 4,pp 824-
html 828,August 2002.

[5]Xie Lanchi; Guo Jingjing; Li Zhihui ,A Novel Unsharp [18]Giovanni Ramponi, Andrea Polesel, Rational Unsharp
Mask Sharpening Method in Preprocessing for Face Masking, Journal of Electron Imaging ,Volume 7 Issue 2,pp
Recognition, IEEE Conference Publications, Fifth 333-338,April 1998.
International Conference on Instrumentation and
Measurement, Computer, Communication and Control [19] Karen Panetta, Yicong Zhou; Sos Agaian,, Hongwei Jia,
(IMCCC), Pages: 378 - 381, 2015 Nonlinear Unsharp Masking for Mammogram Enhancement ,
IEEE Transactions on Information Technology in
[6]G. Ramponi, N. Strobel, S. K. Mitra, and T. Yu, Nonlinear Biomedicine, Volume:15 Issue:6 , Page(s):918 – 928,Nov
Unsharp masking for contrast enhancement, ,Journal of 2011.
Electronic Imaging 5(3),353-366,July 1996.
[20]Siddharth,Rohit Gupta,Vikrant Bhateja,A new Unsharp
[7] A. Polesel, G. Ramponi, and V. J Mathews,Image masking algorithm for Mammography using Non-linear
enhancement via adaptive unsharp masking, IEEE transactions Enhancement Function, Advances in Intelligent and Soft
on image processing, 9(3), 2000. Computing, Volume 132,pp 779-786,2012.

[8] S. K. Mithra and H. Li, A new class of nonlinear filters for Author Details
image enhancement, in Proc. IEEE Int. Conf. Acoustics, Archana J.N received the M. Tech degree in Computer
Speech, Signal Processing, Toronto, Ont., Canada, pp. 2525– Science and Engineering from Anna University, Chennai,
2528.May 1991, India, in 2005 and, currently pursuing her research in Image
Sharpening in Bharathiar University, Coimbatore She is
[9] G. Ramponi,,A cubic unsharp masking technique for working as Assistant Professor in Computer Science and
contrast enhancement,Signal Process., vol. 67, pp. 211–222, Engineering Department at Atria Institute of Technology,
June 1998. Bangalore affiliated to Vishvesharaya Technological
University(VTU). Her research interests include Image
[10] G. Ramponi,, V. John Mathews ,Image Enhancement via processing , Mobile Ad hoc Networks.
Adaptive Unsharp Masking, IEEE transactions on image
processing, vol. 9, no. 3, March 2000 Dr.Aishwarya P received her P.hD in Computer Science
and Engineering from M.S University, Tamil Nadhu. She is
[11]L. Ying, N.T Ming, and L.B Keat, A wavelet based image working as Professor and Head of the Department of CSE at
sharpening algorithm, International Conference on Computer Atria Institute of Technology, Bangalore affiliated to
Vishvesharaya Technological University(VTU),

International Journal of Engineering Science and Computing, July 2016 8733 http://ijesc.org/

View publication stats