Upload
iaeme-publication
View
57
Download
2
Embed Size (px)
Citation preview
http://www.iaeme.com/IJCIET/index.asp 1 [email protected]
International Journal of Computer Engineering & Technology (IJCET)
Volume 6, Issue 10, Oct 2015, pp. 01-12, Article ID: IJCET_06_10_001
Available online at
http://www.iaeme.com/IJCET/issues.asp?JType=IJCET&VType=6&IType=10
ISSN Print: 0976-6367 and ISSN Online: 0976–6375
© IAEME Publication
___________________________________________________________________________
COMPARISON ANALYSIS OF SENSITIVITY
OF NOISE B/W VARIOUS EDGE
DETECTION TECHNIQUE BY ESTIMATING
THEIR PSNR VALUE
Reena Jangra
M. Tech student in C.E, IIET (JIND) under KUK, India
Abhishek Bhatnagar
Asstt. Prof. in C.S.E, IIET (JIND) under KUK, India
ABSTRACT
Edge (corner) are boundaries (borders) between different textures. Edge
(corner) also may be defined as discontinuities in image (picture) intenseness
from one pixel (picture element) to other. Edge (corner) for a image (picture)
are always important characteristics that recommend a suggestion for a
higher frequency. Edge (corner) detection (identification) is a image (picture)
processing technique for finding boundaries (borders) of objects within
images (pictures). It works as a result of detecting discontinuation in
brightness (clarity). Edge (corner) detection (identification) is used for image
(picture) segment & data mining in areas such as image (picture) processing,
computer vision, & machine vision. Common edge (corner) detection
(identification) algorithms include Sobel, Canny, Prewitt, Roberts & fuzzy
logic methods.
Key words: Edge Detection, Canny Edge Detection, Sobel, Robert, Prewit,
PSNR.
Cite this Article: Reena Jangra and Abhishek Bhatnagar. Comparison
Analysis of Sensitivity of Noise B/W Various Edge Detection Technique by
Estimating Their PSNR Value. International Journal of Computer
Engineering and Technology, 6(10), 2015, pp. 01-12.
http://www.iaeme.com/IJCET/issues.asp?JType=IJCET&VType=6&IType=10
1. INTRODUCTION
The points at which image (picture) brightness (clarity) changes sharply are typically
structured into a set of curved line sectors termed edge (corner). same problem of
finding discontinuation in 1D signals is known as step detection (identification) &
problem of finding signal discontinuation with the time is known as change detection
Reena Jangra and Abhishek Bhatnagar
http://www.iaeme.com/IJCIET/index.asp 2 [email protected]
(identification). Edge (corner) detection (identification) is a fundamental tool in image
(picture) processing, particularly in areas of feature detection (identification) &
feature extraction.
Edge (corner) can be identity in a image by detecting correlation b/w image pixel
neighbor.
If a pixel (picture element)’s gray-level value is related to those around it, there is
most likely not a edge (corner) at that point.
If a pixel (picture element)’s has neighbors with extensively irregular gray levels, it
may close to at edge (corner) point.
Figure 1 [Original image ]
2. EDGE DETECTIONMETHODS
May are implemented with sector mask & based on distinct approximations to
differential operators.
Differential operations measure rate of change in image (picture) brightness (clarity)
function.
Some operators are return orientation information & others return information about
the existence of a edge (corner) at each point.
2.1 Roberts Operator
Mark edge(corner) point only
No information about edge(corner) orientation
Work most excellent with binary images(pictures)
Primary disadvantage:
High responsive to noise
A small number of pixel(picture element)s are used to approximate gradient
Comparison Analysis of Sensitivity of Noise B/W Various Edge Detection Technique by
Estimating Their PSNR Value
http://www.iaeme.com/IJCIET/index.asp 3 [email protected]
Implementation of Robert
a1=imread('tulips.jpg');
b1=im2double(a);
[m1,n1]=size(a);
L(1:m1,1:n1)=0
for i=1:m1-2;
for j=1:m1-2;
L(i,j)=-1*b1(i,j)+0+0+1*b1(i+1,j+1);
end;
end;
M(1:m1,1:n1)=0
for i=1:m1-2;
for j=1:m1-2;
M(i,j)=0-1*b1(i,j+1)+1*b1(i+1,j)+0;
end;
end;
figure;
subplot(2,2,1)
imshow(L)
title('Robert Gx1');
subplot(2,2,2)
imshow(M)
title('Robert Gy1');
N=M+L;
subplot(2,2,3)
imshow(N)
title('Robert Gx1+Gy1');
subplot(2,2,4)
imshow(b1)
title('Original Image');
Reena Jangra and Abhishek Bhatnagar
http://www.iaeme.com/IJCIET/index.asp 4 [email protected]
Figure 2 [Edge Detection using Robert]
2.2 Prewit Operator
The masks are as follows
Edge(corner) Magnitude =
Edge(corner) Direction =
Implementation of prewit
%PREWIT
N(1:m1,1:n1)=0
for i=1:m1-2;
for j=1:m1-2;
N(i,j)=-1*b(i,j)-1*b(i,j+1)-
1*b(i,j+2)+0+0+0+1*b(i+2,j)+1*b(i+2,j+1)+1*b(i+2,j+2);
end;
end;
O(1:m1,1:n1)=0
for i=1:m1-2;
for j=1:m1-2;
22 yx
x
y1tan
111
000
111
y
101
101
101
x
Comparison Analysis of Sensitivity of Noise B/W Various Edge Detection Technique by
Estimating Their PSNR Value
http://www.iaeme.com/IJCIET/index.asp 5 [email protected]
O(i,j)=-1*b(i,j)+0+1*b(i,j+2)-1*b(i+2,j)+0+1*b(i+1,j+2)-
1*b(i+2,j)+0+1*b(i+2,j+2);
end;
end;
figure;
subplot(2,2,1)
imshow(N)
title('Prewit Gx1');
subplot(2,2,2)
imshow(O)
title('Prewit Gy1');
Z=N+O;
subplot(2,2,3)
imshow(Z)
title('Prewit Gx1+Gy1');
subplot(2,2,4)
imshow(b)
title('Original Image');
Figure 3 [Edge detection using prewit]
Reena Jangra and Abhishek Bhatnagar
http://www.iaeme.com/IJCIET/index.asp 6 [email protected]
2.3 Sobel Operator
Looks for edge (corner) in both horizontal & vertical directions, then merge
information into a single particular metric.
The masks are as follows:
Edge(corner) Magnitude =
Edge(corner) Direction=
Implementation of sobel
%SOBEL
P(1:m1,1:n)=0
for i1=1:m1-2;
for j=1:m1-2;
P(i1,j)=-1*b(i1,j)-2*b(i1,j+1)-
1*b(i1,j+2)+0+0+0+1*b(i1+2,j)+2*b(i1+2,j+1)+1*b(i1+2,j+2);
end;
end;
R(1:m1,1:n)=0
for i1=1:m1-2;
for j=1:m1-2;
R(i1,j)=-1*b(i1,j)+0+1*b(i1,j+2)-2*b(i1+1,j)+0+2*b(i1+1,j+2)-
1*b(i1+2,j)+0+1*b(i1+2,j+2);
end;
end;
figure;
subplot(2,2,1)
imshow(P)
title('Sobel Gx1');
subplot(2,2,2)
imshow(R)
title('Sobel Gy1');
22 yx
x
y1tan
121
000
121
y
101
202
101
x
Comparison Analysis of Sensitivity of Noise B/W Various Edge Detection Technique by
Estimating Their PSNR Value
http://www.iaeme.com/IJCIET/index.asp 7 [email protected]
Y=P+R;
subplot(2,2,3)
imshow(Y)
title('Soble Gx1+Gy1');
subplot(2,2,4)
imshow(b)
title('Original Image');
Figure 4 [Edge detection using Sobel Operator]
3. PROPOSED WORK
Canny edge (corner) detector have highly developed algorithm originate from
previous work of Marr & Hildreth. It is a best possible edge (corner) detection
(identification) technique as offer good detection (identification), clear response &
good localization.
It is widely used in modern image (picture) processing techniques with a great or
advance achievement in this field. Objective of research is to High light benefit of
canny edge (corner) detection (identification) over traditional edge (corner) detection
(identification) schemes.
On analyzing all these edge (corner) detection (identification) techniques , it is
found that canny gives best possible edge (corner) detection (identification)
.Following are some points throwing light on reward of canny edge (corner) detector
as compared to additional or others detectors discussed here in this research:
Reena Jangra and Abhishek Bhatnagar
http://www.iaeme.com/IJCIET/index.asp 8 [email protected]
Figure 5 [Canny based Edge detection]
4. OBJECTIVE OF RESEARCH
1. Less Sensitive to noise: As compared to others operators like Prewitt, Robert &
Sobel canny edge (corner) detector is less responsive to noise. Its uses Gaussian filter
that may removes noise at a large level as compared to different filters. LoG operator
is also greatly responsive to noise as discriminate twice in contrast to canny operator.
2. Remove streaking problem: standard operators’ like Robert uses only one
thresholding technique but its consequences is that it outcome the edge detection into
streaking. Streaking means, if edge (corner) gradient just above & just below set
threshold limit it removes valuable part of connected edge (corner), & leave
disconnected final edge (corner).
3. Adaptive in nature: slandered operators have unchanging kernels so may not be
adapted to a given image (picture) while performance of canny algorithm depends
only on variable or adjustable.
5. PEAK NOICE RATIO
The psnr function implements following equation to calculate Peak Signal-to-Noise
Ratio (PSNR):
PSNR=10log10
(peak val2/MSE)
where peak val is either specified by user or taken from rage of image(picture) data
type
function psnr= PSNR(X1,Y1)
%Calculates Peak-to-peak Signal to Noise Ratio of two images(pictures) X & Y
[M1,N1]=size(X1);
m1=double(0);
X1=cast(X1,'double');
Comparison Analysis of Sensitivity of Noise B/W Various Edge Detection Technique by
Estimating Their PSNR Value
http://www.iaeme.com/IJCIET/index.asp 9 [email protected]
Y1=cast(Y1,'double');
for i1=1:M1
for j1=1:N1
m1=m1+((X1(i1,j1)-Y1(i1,j1))^2);
end
end
m1=m1/(M1*N1);
psnr1=10*log10(255*255/m1);
psnr1
5.1 ROBERT
>>imgr=imread(‘Robert.jpg’)
5.2 PREWIT
>>imgr=imread(‘prewit.jpg’)
Reena Jangra and Abhishek Bhatnagar
http://www.iaeme.com/IJCIET/index.asp 10 [email protected]
5.3 SOBEL
>>imgs=imread(‘sobel.jpg’)
5.4. OLD CANNY (SINGLE THRESHOLD)
>>imgoc=imread(‘oldcanny1.jpg’)
Comparison Analysis of Sensitivity of Noise B/W Various Edge Detection Technique by
Estimating Their PSNR Value
http://www.iaeme.com/IJCIET/index.asp 11 [email protected]
5.5 CANNY (DOUBLE THRESHOLD)
6. TABLE OF PSNR ANALYSIS WITH DIFFERENT
TECHNIQUE
Technique Test1 Test2 Test3
Robert 20.0619 19.981 20.0729
Prewit 20.4084 20.4544 19.9521
Sobel 20.3162 20.4245 20.4147
Oldcanny 20.0191 20.0229 20.0354
canny 20.0816 20.0886 20.0959
Graph: Comparison analysis b/w all technique
Reena Jangra and Abhishek Bhatnagar
http://www.iaeme.com/IJCIET/index.asp 12 [email protected]
7. FUTURE SCOPE & CONCLUSION
In this research we have studied & calculate different edge (corner) detection
(identification) techniques. We observed that canny edge (corner) detector offer a
confident positive result as compared to others with several optimistic points.
It is a lesser amount of responsive to noise, frank with nature, resolute dilemma of
streaking, offer good localization & detects sharper edge (corner) than others. It is act
as most favorable edge (corner) detection (identification) technique hence a big
amount of work & upgrading on this algorithm is continue &has been done so
advance development are probable in future as a enhanced canny algorithm may
distinguish edge (corner) in color image (picture) without transforming in gray image,
improved canny algorithm for automatic mining (extraction) of moving (unstable)
object in image (picture) guidance.
It finds realistic application in Runway Detection (identification) & Tracking for
Unmanned Aerial Vehicle, in brain MRI image, cable insulation layer measurement,
Real-time facial look identification, edge (corner) detection (identification) of river
regime, Automatic several Faces Tracking & Detection(identification).
Canny edge (corner) detection (identification) technique is helpful in license plate
restructuring system that act as a significant component of intelligent traffic system
(ITS), finds convenient application in traffic organization, public protection &
military sector. In addition it finds fur there additional application & researches in
field of medical area as in ultrasound, x –rays etc.
REFERENCES
[1] The Technology of Night Vision by Harry P. Montoro, ITT Night Vision
http://www.photonics.com/EDU/Han dbook.aspx?AID=25144
[2] A. Marion An Introduction to image Processing, Chapman and Hall, 1991
[3] Azeema Sultana, Dr. M. Meenakshi, “Design and Development of FPGA based
Adaptive Thresholder for image Processing Applications” ,on line access
[4] Gerhard X. Ritter; Joseph N. Wilson, “ Handbook of Computer Vision
Algorithms in image Algebra” CRC Press, CRC Press LLC ISBN:0849326362
Pub Date: 05/01/96
[5] N. Nacereddine, L. Hamami, M. Tridi, and N. Oucief, “Non-Parametric
Histogram-Based Thresholding Methods for Weld Defect Detection in
Radiography “, online access.
[6] Otsu,N., "A Threshold Selection Method from Gray-Level istograms,"IEEE
Transactions on Systems, Man, and Cybernetics, Vol. 9, No. 1, 1979, pp. 62-66.
[7] Elham Ashari , Richard Hornsey, “ FPGA Implementation of Real-Time
Adaptive image Thresholding” ,online access
[8] http://en.wikipedia.org/wiki/Digital_image_processing
[9] J. Canny, “A Computational Approach to Edge Detection,” IEEE Trans. Pattern
Analysis and Machine Intelligence, PAMI-8, 6, November 1986, 679–698.
[10] Ms. Sonali Meghare and Prof. Roshani Talmale, Developing and Comparing an
Encoding System Using Vector Quantization & Edge Detection. International
Journal of Computer Engineering and Technology, 4(3), 2013, pp. 503 – 511.