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

net/publication/224694940

Simulation of Image Enhancement Techniques Using Matlab

Conference Paper · April 2007

DOI: 10.1109/AMS.2007.92 · Source: IEEE Xplore

CITATIONS READS

24 5,140

3 authors, including:

Atul Bansal Rochak Bajpai

GLA University GLA University
52 PUBLICATIONS 1,362 CITATIONS 18 PUBLICATIONS 97 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Atul Bansal on 03 June 2015.

The user has requested enhancement of the downloaded file.

 Scientific Bulletin of the Electrical Engineering Faculty – no. 1 / 2009

SIMULATION OF IMAGE ENHANCEMENT TECHNIQUES USING MATLAB

Nicoleta ANGELESCU1, Iulian UDROIU1, Ioan TACHE2
1
Valahia University of Targoviste, Electrical Engineering Faculty, 18-24 Unirii Blvd., 130082 Targoviste, Romania
2
Politehnica University Bucharest, Electronics, Telecommunications and Information Technology Faculty, 313
Splaiul Independentei., Bucharest, Romania
E-mail: [email protected]

Abstract: This paper emphasis on the importance of perception of the image through modification of
simulation techniques in the field of Digital Image intensity functions, image spectral content, or a
Processing. Simulations are carried out for combination of these functions [3]. Several different
implementation of various basic Image enhancement
classes of methods are available to enhance the image.
techniques on MatLAB in spatial domain. Through this
simulation, it is concluded that the spatial domain We first present standard gray-level mapping
analysis is easier to implement as compared to frequency techniques [5] that allow us to change the perception
domain and also it is concluded that one can understand of the image by a careful remapping of the intensity.
the complex procedures of Image Processing provided The second class of methods involves modification of
he/she is able to analyze the results and compare it with the histogram of the image. The goal is to apply a
the theoretical results mentioned in various books. point wise transformation of the intensity in such a
Keywords: image processing, image enhancement, way that the histogram of the transformed image has a
simulation tehiques.
predefined shape. Finally, the third class of methods
includes linear filtering techniques that can enhance
1. INTRODUCTION
features by applying a combination of low- and high
Simulation is a virtual representation of the reality. pass filters. The selection of a particular method
It may also be defined as the process of knowing the depends on the features that one needs to detect or
characteristics & exhibiting behavior of a particular measure in the image. We then addressed the problem
physical system. Sometimes a learner finds it quite of edge detection. The detection of edges, or object
difficult to understand any physical system behavior boundaries, remains one of the most fundamental
by just reading it from the written material but once he problems in image understanding and computer vision.
is able to see the things actually happening on the In order to solve the edge detection problem, we need
computer system the things really change. That’s why to take into account the noise that is present in the
the very important real life techniques of image image and to balance two conflicting goals: canceling
enhancement such as basic gray level techniques, the noise and minimizing the spatial smoothing.
using arithmetic & logical operations, using spatial
filtering and also in the frequency domain various 2.1. Spatial Domain
filters like Low Pass Filters, High Pass filters have
been simulated on MatLAB and studied. In this paper, In this section Image Enhancement in spatial
various Image Enhancement techniques their domain [1] has been discussed. The various basic gray
procedures, applications and requirements have been level transformations using mathematical operations
discussed. Also various simulated results of these are discussed below:
techniques have been discussed. In this paper it has
been tried to conclude that the simulation of these 2.2. Grayscale Transformation
complex techniques makes the learner much more
comfortable to understand them if he is able to Grayscale transformations can increase or decrease
visualize various simulation results. the intensity range. They are often used to recover
valuable visual information from images that are
2. IMAGE ENHANCEMENT TECHNIQUES overexposed, underexposed, or have a very small
dynamic range. They can also be used when the intensity
The principal objective of Image Enhancement is to range is too large to be displayed on a certain medium
process an Image so that the result is more suitable such as a terminal, printer, or film. Grayscale
than the original image for a specific application. The transformation algorithms are often a part of
goal of image enhancement is to improve the visualization software.

55
Scientific Bulletin of the Electrical Engineering Faculty – no. 1 / 2009

5000
50 4000
4500
3500
100 4000
3000
3500
150
2500 3000

200 2000 2500

2000
250 1500
1500
1000
300 1000
500 500
350
0 0

400 0 50 100 150 200 250 0 50 100 150 200 250

450

500
Fig.3. Original image histogram and histogram
50 100 150 200 250 300 350 400 450 500
equalization
Fig.1. Original image and grayscale image
3. IMAGE FILTERING
2.3. Image Binarizations
Spatial Filtering [1]: The spatial filtering process
Image binarization converts an image of up to 256 consists of simply moving the filter mask from point
gray levels to a black and white image. The simplest to point in an image. At each point the response of
way to use image binarization is to choose a threshold filter is calculated using a predefined relationship.
value, and classify all pixels with values above this Here two smoothening spatial filters - mean and
threshold as white, and all other pixels as black. The median are discussed.
problem then is The goal of image filtering is to reduce impulse
how to select the correct threshold. In many cases, noise or Gaussian noise using a Median and Mean
finding one threshold compatible to the entire image is Filters. A 2-D median filtering operation is performed
very difficult, and in many cases even impossible by sliding a P x P window over the entire image and
therefore, adaptive image binarization is needed where replacing the center pixel by the median of the pixels
an optimal threshold is chosen for each image area. in that window.

v(m, n ) = ∑ ∑ a(k , l )y(m − k , n − l )

( k ,l )∈W
Where: y(m,n) , v(n,m) are the input image and
output image W is the window

v(m, n ) = ∑ ( ∑) y(m − k , n − l )
1
NW k ,l ∈W
Fig 2 Binarized image with a fixed threshold T= and with Obviously, P must be an odd integer in order to
arithmetical mean of gray level. define a center pixel. The filtered pixel values are
stored in another matrix, which forms the processed
2.4. Histogram equalization image. A mean filter works the same way as above
except that the filtered version of the center pixel is
The histogram indicates the number of times each the mean of the window. Mean and median filters are
grayscale value occurs. The normalized histogram is effective for reducing different types of noise.
given by the histogram divided by the total number of Experiments with the Roberts and Sobel edge
pixels in the image, and it is an approximation to the detectors indicate that we need to combine low pass
Probability Mass Function. filtering (or smoothing) with a derivative operator if
The goal in histogram equalization is to we want to generate edge maps that do not contain
approximate the grayscale-value distribution of an noisy spurious edges. The effect of low pass filtering
image to the uniform distribution. Histogram in the spatial domain is to blur the sharp edges, and
equalization is useful because it helps in spreading the therefore increase the uncertainty about the location of
grayscale values and allows us to see a larger range of the edges [4]. In fact, recall that we used a high pass
grayscale values. filtering technique (unsharp masking) to increase the
One can automatically remap the grayscale values contrast of the edges. As is often the case in the design
in order to "stretch the histogram and utilize all the of optimal filters, edge detection involves two
gray values that are available. This procedure is called conflicting goals: cancellation of noise and
histogram equalization. After histogram equalization, minimization of spatial smoothing.
the histogram of the transformed image is uniform.

56
 Scientific Bulletin of the Electrical Engineering Faculty – no. 1 / 2009

fundamental problems in image understanding and

computer vision. The importance of high-contrast lines
that outline objects was clearly stated by David Hubel
"light-dark contours are the most important
components of our perception."[6] In fact, the visual
cortex uses 2-D filters that designed td specifically
detect black lines against a white background or white
lines against a dark background, or simply light-dark
discontinuities in the intensity. Contours provide such
important visual cues that the brain often makes
interpolation between fragments of contours in order
to reconstruct an object.

4.1 Roberts Edge Detector

The Roberts edge detector [13] computes an

approximation to with the first-order finite difference
Fig.4 Filtered image with mean filter operators. The operator will detect edges that are
oriented along the diagonal y = x and the antidiagonal
From figure 4 and 5 we can observe that image y = -x. effect of increasing the threshold. Rather, the
suffer a lost of contrast and luminosity. With mean fundamental problem with this edge detector is that no
filter we obtained better results than the median filter. effort is made to smooth the image before computing
the derivatives.

4.2. Sobel and Prewit Edge Detector

The Sobel operator is used in image processing,

particularly within edge detection algorithms.
Technically, it is a discrete differentiation operator,
computing an approximation of the gradient of the
image intensity function. At each point in the image,
the result of the Sobel operator is either the
corresponding gradient vector or the norm of this
vector. The Sobel operator is based on convolving the
image with a small, separable, and integer valued filter
in horizontal and vertical direction and is therefore
relatively inexpensive in terms of computations. On
the other hand, the gradient approximation which it
produces is relatively crude, in particular for high
frequency variations in the image.
Fig.5 Filtered image with median filter The Sobel [7] and Prewitt [8] edge detectors are
obtained by smoothing the image before computing the
Filter 3x3 5x5 7x7 derivatives, in the direction perpendicular to the derivative.
Mean (PSNR) 19.24 dB 19.29 dB 19.18 dB The god of these methods is to enhance the
Median(PSNR) 19.06 dB 19.25 dB 19.18 dB perception of the image. These techniques rely either
Table 1. PSNR values for the window size equals to: on a point wise transformation of the intensity or on a
3x3, 5x5 and 7x7
combination of low and highpass filters.

4. EDGE DETECTION

Edge detection is a terminology in image

processing and computer vision, particularly in the
areas of feature detection and feature extraction, to
refer to algorithms which aim at identifying points in a
digital image at which the image brightness changes
sharply or more formally has discontinuities The
detection of object boundaries is one of the most Fig.6. Contour Image obtained with Roberts edge detector.

57
 Scientific Bulletin of the Electrical Engineering Faculty – no. 1 / 2009

learning Process. It can be concluded from the above

results that mean filter is more efficient for the same
size of median filter, and for edge detection we
found the best results using the Sobel contour
operator.

REFERENCES

[1] Rehman S. Solar radiation over Saudi Arabia and

Fig.7. Contour Image obtained with Prewitt edge detector. comparisons with empirical models. Energy 1998;
23(12): 1077-1082
[2] Rafael C. Gonzalez, Digital Image Processing
Second Ed., Pearson Education, 2004.
[3] Anthony Edward Nelson, Implementation of
image processing algorithms on FPGA hardware, MS
Diss., Nashville, TN, 2000.
[4] Sohi, D.S.; Devgan, S.S. Application to enhance the
teaching and understanding of basic image processing
techniques Proceedings of the IEEE Southeastcon
2000. 7-9 April 2000, Page(s): 413– 416.
Fig.8. Contour Image obtained with Sobel edge detector. [5] Petrou M. and P Garcia-Sevilla, "Image
processing: Dealing with Texture", 2006, John
Wiley & Sons, Ltd. ISBN: 0-470-02628-6[6] Pitas,
5. CONCLUSION ’’Digital Image Processing Algoritms”, Prentice Hall
Int.UK, 1993
The above discussion and various results obtained [7] D. H. HUBEL"E, ye, brain, and vision," Sdenttfic
lead to the conclusion that it is easier to comprehend American, pp. 8&87,1995.
the concepts of Digital Image Processing and [8] J.M,S. Pn~wm, "Object enhancement and extraction,"
specifically various image enhancement techniques Pidure Processing and Psychopidorics,pp. 7cS149,
with the help of their simulation on MatLAB. As it is 1970.
evident from the results, the learner can easily [9] Sobel ,Neighbourhood coding of binary images
for fast contour following and general array binary
visualize the effects of various techniques and also processing," Computer Vision, Graphics and
can analyze the effects on the images by changing Image Processing, 8, pp. 127-135, 1978.
various parameters. Hence the Simulation of Image
Processing Techniques is an aid to the individual for

58

View publication stats