International Journal of Science, Engineering and Technologies (IJSET), vol.2, no. 1, June.

2015

Improved Color Edge Detection by Fusion of

Hue, PCA & Hybrid Canny
Gurpreet kaur1 and Vijay kumar Banga2

Abstract Edge detection is the name for a set of

mathematical methods which target at classifying points
in a image at which the image intensity varies sharply or,
has discontinuities. This paper has focused on detecting
the edges for the Hue factor in Human visual system
(HSV). Hue is usually represented by position
significance, so the existing edge detection techniques has
been incapable to correctly perceive edges of hue factor in
HSV color plane. As a consequence, the most popular
approaches of color edge detection typically neglected the
role of hue factor, thus misplaced some edges triggered by
hue variations. This research work has proposed a new
color edge detection technique based on the mixture of
hue factor and principal component analysis to resolve the
problems with existing methods. The algorithm has been
designed and implemented in MATLAB. The
performance metrics has shown the efficiency of the
proposed algorithm over the existing ones.

The four steps of edge detection

Filtering- Filter image to improve performance of the
edge detector w.r.t noise.
Smoothing- suppress as much noise as possible, without
destroying the true edges.
Enhancement- applies a filter to enhance the quality of
the edges in the image (sharpening).
Detection- determine which edge pixels should be
discarded as noise and which should be retained (usually,
thresholding provides the criterion used for detection).
Localization- determine the exact location of an edge
(sub-pixel resolution might be required for some
applications, that is, estimate the location of an edge to
better than the spacing between pixels). Edge thinning and
linking are usually required in this step.

Keywords:- Edge Detection, Canny Edge Detector, PCA

Wesolkowski et al. (2000) [1] has presented a

comprehensive color edge comparison across a variety
color spaces. The edge detectors studied are the Sobel
operator, the modified Roberts operator, the vector
gradient operator, and the 3x3 difference vector operator.
Pratts Figure of Merit is used for quantitative evaluation.
The best quantitative and qualitative results were obtained
with the Sobel operator on the hlh2h3 space. The results
in the hlh2h3 space were generally better than results
obtained in any other color space. Dikbas et al. (2007) [2]
has proposed a color edge preserving grayscale
conversion algorithm that helps detect color edges using
only the luminance component. The algorithm calculates
an approximation to the first principal component to form
a new set of luminance coefficients instead of using the
conventional luminance coefficients. This method can be
directly applied to all existing grayscale edge detectors for
color edge detection. Processing only one channel instead
of three channels results in lower computational
complexity compared to other color edge detectors. To
save computation time and to design a hardware

I. INTRODUCTION
Edge detection refers to the process of identifying and
locating sharp discontinuities in an image. The
discontinuities are abrupt changes in pixel intensity which
characterize boundaries of objects in a scene. Edge
detection is a terminology in image processing that refers
to algorithms which aim at identifying edges in an image.
It is encountered in the areas of feature selection and
feature extraction in Computer Vision. An edge detector
accepts a digital image as input and produces an edge map
as output. The edge map of some detectors includes
explicit information about the position and strength of the
edges and their orientation.
Gurpreet kaur, Dept. of Electronics and communication engineering
from Amritsar College of Engineering and Technology,University of
PTU, Punjab 143001 ( e-mail: gurpreetkaurhohangmail.com)
Dr.Vijay Kumar Banga Principal (Prof. ECE) Amritsar College of
Engineering
and
Tecnology,
Punjab
143001
(email:[email protected])

@IRISET 2015

II. LITERATURE SURVEY

International Journal of Science, Engineering and Technologies (IJSET), vol.2, no. 1, June. 2015

implementable real-time algorithm, the proposed method

avoids eigenvector decomposition by making use of
power iteration. The conversion enables the edge detector
to detect some edges of the grayscale image that are not
detected using regular grayscale image. Chen et al. (2010)
[3] to improve the efficiency and the performance of the
color edge detection, a novel color edge detection
algorithm has proposed. An improved Kuwahara filter is
used to smooth the original image first. After edge
detection with each channel independently in RGB color
space, an adaptive threshold selection method is applied to
predict the optimum threshold value and an edge thinning
algorithm is used to extract accurate edge. Edge detection
is anessential tool in image processing and computer
vision. Compared with gray image, color image provides
more edge information of objects. However, the current
color edge detection algorithms acquired so much time to
compute and they are hardly used in real-time system.
Technion et al. (2010) [4] has proposed a new type of
edges on surfaces, termed relief edges. Intuitively, the
surface can be considered as an unknown smooth
manifold, on top of which a local height image is placed.
Relief edges are the edges of this local image. We show
how to compute these edges from the local differential
geometric surface properties, by fitting a local edge model
to the surface. Relief edges manage to capture the 3D
features. They have been utilized to draw edges on
scanned objects for artifact illustration in archaeology.
Edge detection on surfaces, on the other hand, has
received much less attention. The most common edges on
surfaces are ridges and valleys, used for processing range
images in computer vision, as well as for nonphotorealistic rendering in computer graphics. This paper
has extended the definition of edges from images to
surfaces, for which image edges are a special case. These
edges, termed relief curves, use local intrinsic surface
properties together with a rough approximation of the
base surface to produce superior results. Dezert et al.
(2011) [5] has presented an algorithm is based on the
fusion of local edge detectors results expressed into basic
belief assignments thanks to a flexible modeling, and the
proportional conflict redistribution rule developed in
DSmT framework. The purpose of DSmT is to overcome
the limitations of DST mainly by proposing new
underlying models for the frames of discernment in order
to fit better with the nature of real problems, and
proposing new efficient combination and conditioning
rules. The basic belief assignment (bba) associated with
the edge of a pixel in each channel of the image is defined

@IRISET 2015

according to its gradient magnitude, and one can easily

model the uncertainty about our belief it belong or not to
an edge. PCR5 and DS rules have been applied in this
work to combine these bbas to get the global bba for final
decision-making. Other rules of combination of bbas
could also have been used instead but they are known to
be less efficient than PCR5 or DS rules in high and low
conflict cases respectively. XIAO et al. (2011) [6] has
proposed a multi-scale edge detection algorithm which
took soft threshold method to implement detail
enhancement and noise reduction of the true color image.
Firstly, obtaining the true color images at different scales
through wavelet multi-scale edge detection algorithm,
then based on the improved soft threshold filter function,
selecting appropriate threshold of the obtained image
edges to perform noise reduction while enhance the edge
details of the reservation; and finally, carrying out the
weighted 2-norm fusion of edges of different-scale-image.
Wang and Yan (2012) [7] has presented a new edge
detection approach based on vector morphological
operators in color image processing. A new vector
ordering in RGB color space has proposed. And then by
analysing the characteristics of the noise contaminating
image, vector morphological operators has proposed and
these operators are applied in color edge detection.
Kamboj et al. (2012) [8] has proposed a method for edge
detection of color images with automatic threshold
detection. The proposed algorithm extracts the edge
information of color images in RGB color space with
fixed threshold value. The algorithm works on three
channels individually and the output is fused to produce
one edge map. The algorithm uses the improved
Kuwahara filter to smoothen the image, Sobel operator is
used for detecting the edge. A new automatic threshold
detection method based on histogram data is used for
estimating the threshold value. Edge detection reduces
the amount of data needed to process by removing
unnecessary features. Edge detection in color images is
more challenging than edge detection in gray-level
images. Compared with gray image, color image provides
more edge information of objects. However, the current
color edge detection algorithms acquired so much time to
compute and they are hardly used in real-time system. In
order to improve the efficiency and the performance of the
color edge detection. Stock and Koch (2013) [9] has
presented the automatic acceleration of a color edgedetection algorithm from a GA description. From a very
concise GA description of the color edge detection
algorithm, the synthesized accelerator achieved not only a

International Journal of Science, Engineering and Technologies (IJSET), vol.2, no. 1, June. 2015

speed-up of 11.8x against highly optimized software

running on a current-generation processor, but also beat a
programmable GA accelerator ASIC by a speed-up of
350x even when constraining the FPGA to run at the same
clock rate. Using the hardware back-end, Gaalop
synthesizes an algorithm specific accelerator architecture
from the CLU Script DSL. It easily allows
experimentation with different number formats, including
support for the automated optimization of fixed-point
representations. Geometric Algebra (GA) is a branch of
mathematics that generalizes complex numbers and
quaternions. One of the advantages of the framework is,
that it allows intuitive description and manipulation of
geometric objects. While even complex operations can be
described concisely, the actual evaluation of these GA
expressions is extremely compute intensive. However, it
has significant fine-grained parallelism, which makes it a
profitable target for hardware implementation. Hao et al.
(2013) [10] has studied that the premise of obtaining the
clear object contour in traditional Canny operator is to set
appropriate parameters, does not have the adaptive ability.
An adaptive Canny edge-detection method is proposed
which Based on Canny theory. Adopt the 3*3
neighborhood instead of canny algorithm in 2*2
neighborhood to calculate the calculation gradient. Then,
the maximum between-class variance (Otsu) method is
used to obtain the high and low thresholds. Fu et al.
(2013) [11] has compared the two improved methods,
which are improved Sobel operator and improved wavelet
transform using the multi-scale morphological filtering,
subjective visual have achieved better results. However
there are advantages and disadvantages in objective
evaluations. So improvement is further done by using two
improved methods with the wavelet transform fusion
technology. The experimental results has shown that the
fused image has increased significantly in information
entropy and the average gradient compared to the
improved Sobel operator, and it also has improved the
peak signal to noise ratio and the distortion degree
compared to the improved wavelet edge detection method.
The fused image can concentrate the advantages of the
two improved methods together and make complementary
advantages. Eventually, the good de-noising effect and
complete edge are achieved. Abid et al. (2013) [12] has
proposed a new method for image edge detection based on
multilayer perceptron (MLP). The method is based on
updating a MLP to learn a set of contours drawn on a 33
grid and then take advantage of the network
generalization capacity to detect different edge details

@IRISET 2015

even for very noisy images. The method is applied first to

Gray scale images and can be easily extended to color
ones. The method works well even for very low contrast
images for which other edge operators fail. Franchini et al.
(2013) [13] has proposed a hardware implementation of
an edge detection method for color images that exploits
the definition of geometric product of vectors given in the
Clifford algebra framework to extend the convolution
operator and the Fourier transform to vector fields. The
proposed architecture has been prototyped on the Celoxica
Field Programmable Gate Array (FPGA) board. The
proposed hardware architecture allows for an average
speedup ranging between 6x and 18x for different image
sizes against the execution on a conventional generalpurpose processor. Clifford algebra based edge detector
can be exploited to process not only color images but also
multispectral gray-scale images. Lei and Fan (2014) [14]
has proposed a novel color edge detection method based
on the fusion of hue component and principal component
analysis to solve the problems. First, a novel
computational method of hue difference is defined, and
then it is applied to classical gradient operators to obtain
accurate edges for hue component. Moreover, complete
object edges can be obtained by using the edge fusion of
the first principalcomponent and hue component of color
image with low-computational complexity. The proposed
gradient operators are found to be very effective to obtain
better edge results for color images.
III. PROPOSED METHODOLOGY
A. Input color image
Firstly, we input a coloured
algorithm.

image to the proposed

B.To Apply dynamic thresholding

The dynamic thresholding method is used for binarizing
gray level images. The procedure consists of computing
the average gray value of a square of fixed size
surrounding the pixel for each pixel of an image. The
pixel value is then set to black or white, according to
whether its initial value was larger than the surrounding
average or not.
Firstly MATLABs rgb2gray function is used to convert
spectrogram RGB image to gray scale image with pixel
values ranging from 0 to 255; where 0 represents a fully
black pixel and 255 represents white pixel. Then im2bw
function is used with a threshold to convert gray scale

International Journal of Science, Engineering and Technologies (IJSET), vol.2, no. 1, June. 2015

image to binary image. Dynamic thresholding is

calculated by :
= 0.5 [min( ) + max( )] ......(1)
Where,
= 0.5, is a threshold value and
is a input
image.
C. To Apply dynamic thersholding to get binarized image
Binarized images are images that have been quantised to
two values, usually denoted 0 and 1,but often with pixel
values 0 and 255, representing black and white. Binary
images are used in many applications since they are the
simplest to process, but they are such an impoverished
representation of the image information that their use is
not always possible. However, they are useful where all
the information you need can be provided by
the silhouette of the object and when you can obtain the
silhouette of that object easily. Binary images are
typically obtained by thresholding a grey level image. As
shown below equation
, & is binarized image.
Binarized image is calculated by the given equation:
If ( , ) is greater then then,
( , ) = ( , )+

( , ) = ( , )

(, )

1)..(7)

D. To apply canny on binarized image , &

=[
( ,
)]..(8)
=[
( ,
)]..(9)
=
,
..(10)

)]..(11)
=[
( ,
Where,
, & is output of the binarized image , &
from the improved canny edge detector.
E. To Apply edge fusion
Fusion is the process of combine significant information
from two or more images into a particular image. The
consequential image will be more useful than any of the
input images.
Apply improved canny on the step IV all the binarized
images edges are detected and in this step apply edge
fusion to get fused image by the given equation:

( (, )

1) .....................(2)

D = D + D + D + D ...........................................(12)
Where ,
D is improved canny to apply on the different binarized
image .

( (, )

1) .......................(3)

F. Then find the Final edge detected image

Otherwise,
( , ) = ( , )

Where ,
( , ) is binarized image, ( , ) is input image,
maximum dynamic limit, shows an error and
, & is the variable of sub image.

The output is
the final edge detected image.
Mathematically, the resultant image is given by :
is

(, )

1)....................(4)

(, )

1)....................(5)

Otherwise,
( , ) = ( , )

(, )

........................................................(13)

The proposed technology has been design and

implemented in metlab using image processing tool
box. Subsequent section contain experimental result
and performance evaluation existence and proposed
technique.
V. EXPERIMENTAL SET-UP

To calculate
:
If ( , ) is greater then then,
( , ) = ( , )+

IV. RESULTS AND DISCUSSIONS

To calculate
:
If ( , ) is greater then then ,
( , ) = ( , )+

1)...(6)

Here is some representation of proposed algorithm to

detect edges in a color image. Firstly input image is
shown below:-

Otherwise ,

@IRISET 2015

International Journal of Science, Engineering and Technologies (IJSET), vol.2, no. 1, June. 2015

Fig: (c): Hue Detected Edge Image

Fig (a): shows the input image
Figure (b) shows the hue image of given input image, the
hue image describes the true color of an image. From
color image to grey-scale image, leads to the outcome that
a few edges are missed. In addition, mainly of the missing
edges consequence from hue changes. Also the pixel
values are same in grey scale images. So to detect missed
edges we apply hue analysis to color image.

Figure (d) shows the PCA image of the given input image,
the PCA describes the major component information of a
part in an image. The three color components are
condensed into one containing a main component of
information. So to detect missed edges we apply PCA
analysis to color image.

Fig (d): PCA Image

Fig (b): Hue Image
Figure (c) shows the hue detected edge image, we can
present a superior edge detection representation for color
image once the problem of edge detection of hue
component is solved.

Fig (e): PCA Edge Detected Image

Figure (e) shows the PCA detected edge image, we can
present a superior edge detection representation for color
image once the problem of edge detection of PCA
component is solved.

@IRISET 2015

International Journal of Science, Engineering and Technologies (IJSET), vol.2, no. 1, June. 2015

Figure (h) gives the final result of edge detection in a

color image by fusion of hue component, PCA and
improved canny based edge detection.

VI. PERFORMANCE EVALUATION

(a) Root Mean Square Evaluation
Table 1 RMSE values of many images
Fig (f): Fusion of Hue and PCA Edge Image

Images

Figure (f) shows fusion of hue and PCA edge image

combine significant information from two or more images
into a particular image and gets clearer results to detect
edges in color image.

1. jpg
2. jpg
3. jpg
4. jpg
5. jpg
6. jpg
7. jpg
8. jpg
9. jpg
10. jpg
11. jpg
12. jpg
13. jpg
14. jpg
15. jpg
16. jpg
17. jpg
18. jpg
19. jpg
20. jpg

Fig (g): Improved Canny Based Edge Detection.

Figure (g) shows improved canny based edge describes.
The improved canny detection detects fine edges in an
image. It preserves edges of an image even in low
intensity images

Existing Edge
Detection
125.2078
120.4907
106.3626
149.8299
168.1220
139.4274
192.5643
135.8713
162.9663
116.1163
153.2123
90.2940
144.9931
151.3968
121.0331
133.6114
137.3244
144.2186
116.1034
213.1525

Proposed Edge
Detection
120.0840
115.3661
103.2256
147.6897
161.0595
132.3054
186.4110
127.8087
149.8312
109.0129
148.0882
86.1887
140.8033
148.3175
117.7808
129.4241
133.0474
139.9236
112.9181
210.0117

Table 1 shows the result of RMSE for many images, the

two corresponding values for each image shows the
difference between the existing and proposed Edge
Detection. Here in this the values are decreased for each
image from existing to proposed Edge Detection.

Fig (h): Final Edge Detected Image

@IRISET 2015

International Journal of Science, Engineering and Technologies (IJSET), vol.2, no. 1, June. 2015

Graph 1: RMSE Evaluation

Graph 1 shows the various images on X-axis and rate on
Y-axis, the two graphs indicate the results according to
the values of existing and proposed Pratt of previous
RMSE table and results shows the Pratt figure of merit for
RMSE.
(B) Comparison of PFOM values of edge images by
different methods
Table 2: PFOM values of many images
Images
1. jpg
2. jpg
3. jpg
4. jpg
5. jpg
6. jpg
7. jpg
8. jpg
9. jpg
10.jpg
11.jpg
12.jpg
13.jpg
14. jpg
15. jpg
16. jpg
17. jpg
18. jpg
19. jpg
20. jpg

@IRISET 2015

Existing Edge
Detection
0.61730
0.6022
0.4145
0.5349
0.4408
0.4811
0.5068
0.4122
0.5743
0.6001
0.5289
0.4786
0.5270
0.7578
0.5306
0.5958
0.5254
0.6154
0.5929
0.6274

Proposed Edge
Detection
0.6563
0.6803
0.6293
0.6639
0.7551
0.5281
0.6725
0.7265
0.6694
0.7173
0.5870
0.7279
0.7313
0.7810
0.6143
0.6777
0.6582
0.7555
0.6738
0.7434

Graph 2: PFOM Evaluation

VII. CONCLUSION AND FUTURE SCOPE
Edge detection is a fundamental tool used in most image
processing applications to obtain information from the
frames as a precursor step to feature extraction and object
segmentation. By conducting the survey of different
research papers on the new edge detection techniques it
has been found that the most of the existing work has been
based on gray images so effect of true color has been
neglected, moreover the use of hybrid canny for edge
detection has also been neglected. The significant
improvement of the algorithm over the existing edge
detectors has been proved with the help of various
parameters. This work has not considered any artificial
intelligence based algorithm, so in near future, fuzzy
logic, genetic algorithm and particle swarm optimization
edge detection can be included.

REFERENCES
[1] Wesolkowski, Slawo, M. E. Jernigan, and Robert D.
Dony. "Comparison of color image edge detectors in
multiple color spaces." In Image Processing, 2000.
Proceedings. 2000 International Conference on, vol.
2, pp. 796-799. IEEE, 2000.
[2] Dikbas, Salih, Tarik Arici, and Yucel Altunbasak.
"Chrominance
edge
preserving
grayscale
transformation with approximate first principal
component for color edge detection." In Image
Processing, 2007. ICIP 2007. IEEE International
Conference on, vol. 2, pp. II-261. IEEE, 2007.
[3] Chen, Xin, and Houjin Chen. "A novel color edge
detection algorithm in RGB color space." In Signal

International Journal of Science, Engineering and Technologies (IJSET), vol.2, no. 1, June. 2015

Processing (ICSP), 2010 IEEE 10th International

Conference on, pp. 793-796. IEEE, 2010.
[4] Kolomenkin, Michael, Ilan Shimshoni, and Ayellet
Tal. "On edge detection on surfaces." In Computer
Vision and Pattern Recognition, 2009. CVPR 2009.
IEEE Conference on, pp. 2767-2774. IEEE, 2009.
[5] Dezert, Jean, Zhun-ga Liu, and Grgoire Mercier.
"Edge detection in color images based on DSmT."
In Information Fusion (FUSION), 2011 Proceedings
of the 14th International Conference on, pp. 1-8.
IEEE, 2011.
[6] Xiao, Feng, Mingquan Zhou, and Guohua Geng.
"Detail enhancement and noise reduction with true
color image edge detection based on wavelet multiscale." In Artificial Intelligence, Management Science
and Electronic Commerce (AIMSEC), 2011 2nd
International Conference on, pp. 1061-1064. IEEE,
2011.
[7] Wang, Lei, and Lingbo Yan. "Edge detection of color
image using vector morphological operators."
In Computer Science and Network Technology
(ICCSNT), 2012 2nd International Conference on,
pp. 2211-2215. IEEE, 2012.
[8] Kamboj, Anil, Kavita Grewal, and Ruchi Mittal.
"Color Edge Detection in RGB Color Space Using
Automatic Threshold Detection." International
Journal of Innovative Technology and Exploring
Engineering (IJITEE) ISSN: 2278-3075, Volume-1,
Issue-3, August 2012
[9] Stock, Florian, Andreas Koch, and Dietmar
Hildenbrand. "FPGA-accelerated color edge detection
using a Geometric-Algebra-to-Verilog compiler."
In System on Chip (SoC), 2013 International
Symposium on, pp. 1-6. IEEE, 2013.
[10] Hao, Geng, Luo Min, and Hu Feng. "Improved SelfAdaptive Edge Detection Method Based on Canny."
In Intelligent Human-Machine Systems and
Cybernetics (IHMSC), 2013 5th International
Conference on, vol. 2, pp. 527-530. IEEE, 2013.
[11] Fu, Zuo-Xian, Cai-Xia Deng, and Yuan-Yan Tang.
"A fusion edge detection method based on wavelet
transform and differential." InWavelet Analysis and
Pattern Recognition (ICWAPR), 2013 International
Conference on, pp. 17-21. IEEE, 2013.
[12] Abid, Sabeur, Farhat Fnaiech, and Ezzeddine Ben
Braiek. "A novel neural network approach for image
edge detection." In Electrical Engineering and
Software Applications (ICEESA), 2013 International
Conference on, pp. 1-6. IEEE, 2013.

@IRISET 2015

[13] Franchini, Silvia, Antonio Gentile, Giorgio Vassallo,

Filippo Sorbello, and Salvatore Vitabile. "A
Specialized Architecture for Color Image Edge
Detection Based on Clifford Algebra." In Complex,
Intelligent, and Software Intensive Systems (CISIS),
2013 Seventh International Conference on, pp. 128135. IEEE, 2013.
[14] Lei, Tao, Yangyu Fan, and Yi Wang. "Colour edge
detection based on the fusion of hue component and
principal component analysis." IET Image Processing
8, no. 1 (2014): 44-55.