Krishan Kant Lavania et al.

/ International Journal on Computer Science and Engineering (IJCSE)

Image Enhancement using Filtering

Techniques
Krishan Kant Lavania
Department of Computer Science and Engineering
Arya Institute of Engineering and Technology
Jaipur (Rajasthan), India
[email protected]

Shivali
Department of Computer Science and Engineering
Arya Institute of Engineering and Technology
Jaipur (Rajasthan), India
[email protected]

Rajiv Kumar
Department of Information Technology
Gurgaon Institute of Technology and Management
Gurgaon (Haryana), India
[email protected]

Abstract— Image enhancement is used to improve the quality of an image. In this paper, two novel
algorithms CB (Center-to-Boundary) and BB (Boundary-to-Boundary) filters have proposed. These
algorithms have implemented for mask, but in this paper, authors have chosen only which
are convoluted with various poorly contrast images. The performance of these two algorithms are
analyzed and compared with the average filter on the basis of four parameters - eme, mse, rmse, and
psnr. The experimental results are carried on more than 1000 images that prove that these two
algorithms enhance poor quality images very effectively.

Keywords- Image Processin; Image Enhancement; CB Filter; BB Filter;

I. INTRODUCTION
Image enhancement [1,2,7,8,13] is the processing of image to enhance certain feature of an image. Image
enhancement is basically improving the interpretability or perception of information in images for human
viewers and providing better input for other automated image processing techniques. The principal objective of
image enhancement is to modify attributes of an image to make it more suitable for a given task an\\\\d a
specific observer. During this process, one or more attributes of the image are modified. The choice of attributes
and the way they are modified are specific to a given task. Moreover, observer-specific factors, such as the
human visual system and the observer's experience, will introduce a great deal of subjectivity into the choice of
image enhancement methods. Image enhancement is used in the following cases: Removal of noise from image,
Enhancement of the dark image and highlight the edges of the objects in an image. The result is more suitable
than the original image for certain specific applications. Processing techniques are very much problem oriented
[3,9,10,12]. For example, best techniques for enhancement of X-ray image may not be best for enhancement for
microscopic images.
In this paper, two novel filters CB (Centre-to-Boundary) and BB (Boundary-to-Boundary) filters have
developed. Both these filters are used to remove the noisy pixels and transform them into good one. They are
based on calculating the distance between image pixels and their neighbors. They are used to enhance abnormal
pixels values to remove noise and enhance the appearance of an image. The distance between the image pixel
and its neighbors will be calculated to minimize the effectiveness of noisy pixels. The distance values represent
the relation between good pixels and noisy pixels, which satisfy the assumption, that “far noisy pixels have less
effect on surrounded good ones”[11,12,15].
In section II and III, authors defined the CB (Centre-to-Boundary) and BB (Boundary-to-Boundary) filter which
remove the noisy pixels and transformed them into good pixels. They are implemented for . In section
IV, the performance of these two filters for 3 3 mask are evaluated and compared with average filter. The
graphs and various images that are obtained from these filters have also displayed. Finally, paper concludes in
section V.

ISSN : 0975-3397 Vol. 4 No. 01 January 2012 14

 Krishan Kant Lavania et al. / International Journal on Computer Science and Engineering (IJCSE)

II. CB(CENTER-TO-BOUNDARY) FILTER

In CB filter, pixels are scanned from Center to Boundary to calculate the distance between them by using
Newton-Raphson’s formula. The Newton-Raphson method is a powerful technique for solving equations
numerically. The Newton-Raphson formula is based on the equation (1) as given below:
1 
Where
1 2
2 1 3
= Central pixel
= Neighboring pixel of
Equation (ii) is chosen experimentally and this is the fitness criterion. This formula is applied on 1000 images
and results are found very satisfactory. The central pixel is assumed to be 1. Then use 4-connectivity chain code
and apply the Newton Raphson’s method using fitness criterion 1, The distance is calculated
between centralized pixel and its neighbors and then apply the same process until all the pixels are scanned. Do
not scan those pixels that are already scanned. The table 1 has shown for 3×3 CB filter.
TABLE 1. 3×3 MASK FOR CB FILTER

1.67 2 1.67

2 1 2

1.67 2 1.67

A. Algorithm
Step 1: Choose any n×n matrix.
Step 2: Check whether the matrix is odd or even. If even, then exit otherwise proceed to step3.
Step 3: Initialize the central pixel of the matrix equal to 1 i.e 1.
Step 4: Scan from the Central pixel to all neighboring pixels by using 4-connectivity.
Step 5: Apply Newton-Raphson’s Method on step 4 to find out the value of neighboring pixels using formula:
4
Step 6: Repeat step 4 and 5 until all pixels are scanned (NOTE: Do not scan those pixels that are already
scanned.
Step 7: Exit

B. Results
The CB filter applied on the original image and results has displayed in Fig.1 [4,5,11].

Figure 1. Image enhancement using CB filter

III. BB(BOUNDARY-TO-BOUNDARY) FILTER

In BB filter, pixels are scanned from upper left Boundary to lower right Boundary pixels to calculate distance
between them by using Newton-Raphson’s formula using equation 1 as described above. The table 2 has shown
for 3×3 CB filter.

ISSN : 0975-3397 Vol. 4 No. 01 January 2012 15

 Krishan Kant Lavania et al. / International Journal on Computer Science and Engineering (IJCSE)

TABLE 2. 3×3 MASK FOR BB FILTER

1 2 1.67

2 1.67 1.62

1.67 1.62 1.62

A. Algorithm
Step 1: Choose any n×n matrix.
Step 2: Check whether the matrix is odd or even. If even, then exit otherwise proceed to step3.
Step 3: Initialize the upper left boundary pixel of the matrix equal to 1 i.e 1.
Step 4: Scan from the upper left boundary pixel to all neighboring pixels by using 4-connectivity.
Step 5: Apply Newton-Raphson’s Method to find out the value of neighboring pixels using formula:

′
5
Step 6: Repeat step 4 and 5 until all pixels are scanned (NOTE: Do not scan those pixels that are already
scanned.
Note: Do not scan those pixels that are already scanned.
Step 7: Exit

B. Results
The BB filter applied on the original image and results has displayed in Fig. 2 [4,5,11].

Figure 2. Image enhancement using BB filter

IV. PERFORMANCE EVALUATION AND COMPARISION OF CB AND BB FILTER WITH AVERAGE

FILTER
The performance of the CB and BB filter for 3 3 mask are evaluated and compared with average filter.
The four parameters, namely, eme, psnr, mse and rmse have chosen to measure the performance of these filters.
These algorithms can be applied on n×n masks. In this paper, only 3×3 mask has taken and it convolute with
various poor quality images. The results are implemented using MATLAB 7.4 using Image Processing Toolbox
(IPT) [4,5,13]. The results are shown as:
A. eme(Quality Measure Of Image Enhancement)
The following results based on eme parameters [14] have shown in table 3 given below:

TABLE 3. EME VALUES OF CB, BB AND AVERAGE FILTER FOR 3×3 MASK

Image CB Filter BB Filter AVG Filter

moon 13.0678 12.3028 12.4036
fish 1.5656 3.1460 2.7913
coin 3.8763 4.1736 4.1546
flower 7.9600 7.9816 7.7674
pout 2.6538 2.7055 2.5805

ISSN : 0975-3397 Vol. 4 No. 01 January 2012 16

 Krishan Kant Lavania et al. / International Journal on Computer Science and Engineering (IJCSE)

B. mse(Mean Square Error)

The following results based on mse parameters have shown in table 4 given below:

TABLE 4. MSE VALUES OF CB, BB AND AVERAGE FILTER FOR 3×3 MASK

Image CB Filter BB Filter AVG Filter

moon 0.1883 0.3361 5.8554
fish 1.0943 2.2286 10.3686
coin 0.7643 1.6088 16.3373
flower 0.4473 1.1854 5.9577
pout 1.2738 2.1309 5.9394

C. rmse(Root Mean Square Error)

The following results based on rmse parameters have shown in table 5 given below:

TABLE 5. RMSE VALUES OF CB, BB AND AVERAGE FILTER FOR 3×3 MASK

Image CB FILTER BB FILTER AVG FILTER

moon 0.4340 0.5797 2.4198
fish 1.0461 1.4928 3.2200
coin 0.8742 1.2684 4.0419
flower 0.6688 1.0888 2.4408
pout 1.1286 1.4598 2.4371

D. psnr(Peak-to-Signal-Noise Ratio)
The following results based on psnr parameters [6] have shown in table 6 given below:

TABLE 6. PSNR VALUES OF CB, BB AND AVERAGE FILTER FOR 3×3 MASK

Image CB Filter BB Filter AVG Filter

moon 55.3817 52.8660 40.4552

fish 47.7394 44.6505 37.9736
coin 49.2984 46.0658 35.9990
flower 51.6251 47.3922 40.3800
pout 47.0799 44.8451 40.3934

The graphs of CB, BB and average filters are shown for various images for 3×3 mask [4,13] and they are shown
as:

ISSN : 0975-3397 Vol. 4 No. 01 January 2012 17

 Krishan Kant Lavania et al. / International Journal on Computer Science and Engineering (IJCSE)

Figure 3. Comparison of eme values of various image using Figure 4. Comparison of mse values of various image using
3×3 filter for table 3×3 filter for table

Figure 5. Comparison of rmse values of various image using Figure 6. Comparison of psnr values of various image using
3×3 filter for table 3×3 filter for table

The results of various images using CB, BB and average filter are compared and they are shown in figure 7.

Figure 7. Results of various images using CB, BB and average filter

V. CONCLUSION
In this paper, two novel filters that are Center-to-Boundary (CB) filter and Boundary-to-Boundary (BB) filter
using Newton Raphson’s methods have developed. These two filters are compared with existing average filter

ISSN : 0975-3397 Vol. 4 No. 01 January 2012 18

 Krishan Kant Lavania et al. / International Journal on Computer Science and Engineering (IJCSE)

on the basis of four parameters- eme (Quality measure of image enhancement), mse (mean square error), rmse
(root mean square error) and psnr (Peak-to-Signal noise ratio). The performances of these two filters are
evaluated on more than 1000 images using MATLAB 7.4 and the results are found very satisfactory as
compared to average filter. From table 3, the following results are obtained. eme value of CB, BB and average
filter in case of mask are 13.0678, 12.3028 and 12.4036 respectively which shows that CB filter is better
than BB and average filter. As we increase the dimension of filters, the quality of image has decreased but still
the results of CB and BB filter are much better than average. As it is clear from table 4, the mse values of CB
and BB filter are much less than average filters that show that the image is of good quality in case of CB and BB
filter. As the dimension of filter has increased, the error between the original images and enhanced image will
increase but still the values in CB and BB filter are much lesser than average filter. Similar is the case with rmse
value in Table 5.The Table 6 shows that psnr values of CB and BB filter are much higher than average filter that
is necessary for good enhanced image. In case of increased filter dimensions, psnr value has decreased but still it
is good in case of CB and BB filter. Finally, this paper work is concluded with graph and bar chart that shows
the comparison of these three filters over eme, mse, rmse and psnr parameters. The results are implemented on
MATLAB 7.4 using Image Processing Toolbox (IPT). In short word, it has been concluded that performance of
CB and BB filter are much better than average filter.

REFERENCES
[1] K. Jain, Fundamentals of Digital Image Processing. Englewood Cliffs, NJ: Prentice Hall, 1989.
[2] Aniati Murni [2000], Image Processing, class handouts, Faculty of Computer Science, University of Indonesia, Jakarta.
[3] Ashish Mehta,” A Literature Survey On image enhancement”, MHTASH002, Research Methods, Date: 07 August 2006.
[4] D-I-Y Matlab Tutorials http://pesona.mmu.edu.my/~johnsee/matlab
[5] E.S. Gopi, Digital Image Processing using Matlab, Senior Lecturer, Department of Electronics and Communication Engineering, Sri.
Venkateswara College of Engineering Pennalur, Sriperumbudur, Tamilnadu, SciTech Publication (India) Pvt. Ltd., 2007.
[6] “Image Quality Evaluation Based On Image Weighted Separating Block Peak Signal-to-Noise Ratio”, IEEE Int.Conf. Neural
Networks & Signal Processing, Nanjing, China, December 14-17, 2003.
[7] J.C. Russ, The Image Processing Handbook, CRC Press, Boca Raton, FL., 1992.
[8] K. R. Castleman (1979), Digital Image Processing. Prentice Hall, Englewood Cliffs, NJ
[9] Mr. Salem Saleh Al-amri1, Dr.N.V. Kalyankar2, Dr.S.D.Khamitkar,” Linear and Non-linear Contrast Enhancement Image”, IJCSNS
International Journal of Computer Science and Network Security, VOL.10 No.2, February 2010.
[10] Muhammad Shahzad, Shiraz Latif,” Efficient Image Enhancement Techniques”, Journal of Information & Communication
Technology Vol. 3, No. 1, (Spring 2009) 50-55.
[11] Raghad Jawad AHMED,” Image enhancement and noise removal by using new spatial filters”, U.P.B. Sci. Bull., Series C, Vol. 73, Iss.
1, 2011 ISSN 1454-234x.
[12] Raman Maini and Himanshu Aggarwal, “A comprehensive review of Image Enhancement techniques, journal of computing”, volume
2, issue 3, march 2010, issn 2151-9617

[13] R. C. Gonzalez, Richard E. Woods, Digital Image processing, Addision-Wesely, 2003 R. Gonzalez, R. Woods and S. Eddins “Digital
Image Processing Using Matlab”, 2004, Prentice Hall.
[14] S. S. Agaian, K Panetta, and A. M. Grigorian, “A new measure of image enhancement,” presented at the IASTED Int. Conf. Signal
Processing Communication, Marbella, Spain, Sep. 19-22, 2000.
[15] T.S. Huang, G.J. Yang, And G.Y. Tang, "A Fast Two Dimensional Median Filtering Algorithm", IEEE Trans. On Accustics, Speech,
Signal Processing, Vol. ASSP-27, No.1, Feb 1997.

AUTHORS PROFILE

Krishan kant lavania has done M.Tech. (Computer Science) from JNRV University,
Udaipur,M.Sc. ( Computer Science ) from Dr. B.R. Ambedkar University (Formerly Agra
University), Agra, B.Sc. ( Vocational ) Specialization in Computer Applications from DDIVE,
Agra; An Institute of Dr. B.R. Ambedkar University (Formerly Agra University), Agra. Currently
he is working as Associate Professor in Arya Institute of Engineering & Technology
He has published several National and International journal papers.

Shivali has done B.Tech from Kurushetra University, Kurushetra Haryana in 2006. She is pursuing
M.Tech from RTU, Jaipur. She obtained MBA from Symbiosis University, Pune. She has published
three papers in National Conference and one international journal on Image Processing. She is
member of CSI. Her area of interest is image morphing,

ISSN : 0975-3397 Vol. 4 No. 01 January 2012 19

 Krishan Kant Lavania et al. / International Journal on Computer Science and Engineering (IJCSE)

Rajiv Kumar obtained M. Tech (CSE) from Guru Jambeshwar University, Hissar, Haryana in 2005.
He is pursuing Ph.D. from Bharathiar University, Coimbatore, Tamil Nadu from 2008
onwards.During 2005-2009, he served as Lecturer in R V College of Engineering, Bangalore.
Currently, he is working as Assistant Professor in Gurgaon Institue of Technology and Management,
Gurgaon, Haryana. He has published 13 papers in National/International Conference and 2 papers are
in International Journal. He has participated in various programme/workshop/conference from
different organization like Infosys, Wipro, RVCE etc. he is member of CSI. His area of interest is
image processing, computer vision and pattern recognition.

ISSN : 0975-3397 Vol. 4 No. 01 January 2012 20