International Journal of Modern Engineering Research (IJMER)

www.ijmer.com Vol.2, Issue.2, Mar-Apr 2016 pp-552-554 ISSN: 2249-6645

Performance and Analysis of Edge detection using FPGA

Implementation
R. Ponneela Vignesh
Assistant Professor,Dept of ECE,Tamilnadu College of Engineering
Coimbatore,Tamilnadu.

Abstract II. Canny edge detection algorithm

Edge detection serves as a pre-processing step for The Canny edge detection algorithm is known to many as
many image processing algorithms such as image the optimal edge detector. The first and most obvious is low
enhancement, image segmentation, tracking and image/video error rate. It is the important that edges occurring in images
coding. The edge detection is one of the key stages in image should not be missed and that there are no responses to non-
processing and object recognition. This paper present a edges. The second criterion is that the edge points be well
Canny edge detection algorithm that results in significantly localized. A third criterion is to have only one response to a
reduced memory requirements, decreased latency and single edge.
increased throughput with no loss in edge detection This was implemented because the first two were
performance. This edge detection algorithm is based on not substantial enough to completely eliminate the possibility
MATLAB simulation and FPGA implementation. of multiple responses to an edge. Based on these criteria, the
canny edge detector first smoothes the image to eliminate
Keywords: Canny edge detector, MATLAB and FPGA and noise. It then finds the image gradient to highlight
regions with high spatial derivatives After the edge directions
I. Introduction are known, non maximum suppression now has to be applied
Edge detection is a basic operation in image processing, it Non-maximum suppression is used to trace along the edge in
refers to the process identifying and locating sharp the edge direction and suppress any pixel value that is not
discontinuities in an image, the discontinuities are abrupt considered to be an edge. This will give a thin line in the
changes in pixel intensity which characterize boundaries of output image.
objects in a scene. It is a very important first step in many The algorithm then tracks along these regions and
algorithms used for segmentation, tracking and object suppresses any pixel that is not at the maximum (non
recognition [1]. maximum suppression). The gradient array is now further
There are an extremely large number of edge reduced by hysteresis. Hysteresis is used to track along the
detection operators available, each designed to be sensitive to remaining pixels that have not been suppressed Hysteresis
edges, typically it reduces the memory size and the uses two thresholds and if the magnitude is below the first
computation cost[2] the edge detection algorithms are threshold, it is set to zero (made a non edge). To avoid this
implemented using software, with advance in very large scale hysteresis uses 2 thresholds, a high and a low. If the
integration(VLSI) technology, their hardware magnitude is above the high threshold, it is made an edge.
implementation become an attractive alternative, especially This model was based on a step edge corrupted by additive
for real-time applications, Many edge detection algorithms white Gaussian noise. Canny edge detection developed an
such as Robert detector, Prewitt detector, Kirsch detector, approach to derive an optimal edge detector based on three
Gauss-Laplace detector and Canny edge detector have been criteria related to the detection performance. The Canny edge
proposed[3], among these algorithms has be widely used in detection algorithm block diagram is shown in below.
the field of image processing because of its good
performance[3]. It operates on two rows of pixels at a time
this reduces the memory requirement at the expense of a
decrease in the throughput [2]. In order to reduce memory
requirements, decrease latency and increased throughput is
proposed in [1]. An absolute different mask edge detection
algorithm and its pipelined VLSI architecture for real-time
application. But the edge detector in offers a trade-off
between precision, cost and speed, and its capability to detect
edges is not as good as the Canny algorithm [4]. Fig 1 : Block of Canny edge detection
In this paper, a method based on non-uniform and
coarse quantization of the gradient magnitude histogram is The Canny edge detection algorithm consists of the
proposed. In addition, the proposed algorithm is mapped onto following steps:
reconfigurable hardware architecture. The original Canny  Calculating the horizontal gradient and vertical
edge detection algorithm computes the high and low gradient at each pixel location by convolving the
thresholds for edge detection based on the entire image image with partial derivatives of a 2D Gaussian
statistics which prevents the processing on individual block function. 
independently.

www.ijmer.com 552 | Page

 International Journal of Modern Engineering Research (IJMER)
www.ijmer.com Vol.2, Issue.2, Mar-Apr 2016 pp-552-554 ISSN: 2249-6645
 Smoothing the input image by Gaussian mask. The set as an edge pixel. If a pixel has a value above the low
 output smoothed the image.  threshold and is the neighbour of an edge pixel, it is set as an
 Computing the gradient magnitude and direction at edge pixel as well. If a pixel has a value above the low
 each pixel location.  threshold but is not the neighbour of an edge pixel, it is not
 Applying non-maximum suppression (NMS) to thin set as an edge pixel. If a pixel has a value below the low
 edge.  threshold, it is never set as an edge pixel.
 Computing the hysteresis high and low thresholds First the edge detection algorithm is implemented
based on the histogram of the magnitudes of the by applying a Gaussian filter on the whole image. Then the
 gradients of the entire image.  processed image data are used as the input to calculate the
gradient at each pixel level and these gradient values are used
 Performing hysteresis thresholding to determine the
edge map.  to calculate the direction and the magnitude at a certain pixel.

 The low pass filtering is achieved by taking the
average pixel values. 

It can be shown that convolving an image with a

symmetric 2D Gaussian and then differentiating in the
direction of the gradient forms simple and effective
directional operator, which meets the three criteria mentioned
above
Suppose the G(x, y) is a 2D Gaussian and I(x, y) is the image,
then the smoothed image H(x, y) as
Fig 2 : Design flow of Canny edge detection
H(x, y) = G(x, y) * I(x, y)
In order to implement the Canny edge detector
Where the H(x, y) is denoted as horizontal gradient, the G(x, algorithm, some steps must be followed. The following steps
y) is denoted as gradient magnitude and I(x, y) as smoothed are executed sequentially:
image.  The first step is to filter out any noise in the original
The data obtains usually contains some spurious
 image before trying to locate and detect any edges. 
responses in the non maximal suppression (NMS). This is  The Gaussian filter mask is used exclusively in Canny
called the streaking problem and is quite common in the edge
 edge detection algorithm. 
detection problem. These streaking can be eliminated by
 The Gaussian smoothing can be performed using
using a threshold with hysteresis.
standard convolution methods. a convolution mask is
much smaller than the actual image 
The Canny edge detector will be carried out in  The larger width is the Gaussian mask and the
following four steps: lower is the detector sensitivity.
1. Image smoothing
 The localization error in the detected edges also
2. Gradient calculation and directional Non-Maximum
increases slightly as the Gaussian width is increased.
suppression
The Gaussian in implementation is shown below. 
3. Calculating thresholds
4. Thresholding with hysteresis.

1. Smooth the image with a two dimensional Gaussian. In

most cases the computation of a two dimensional
Gaussian is costly, so it is approximated by two one
dimensional Gaussians, one in the x direction and the
other in the y direction.
2. Take the gradient of the image. This shows changes in
intensity, which indicates the presence of edges. This
actually gives two results, the gradient in the x direction
and the gradient in the y direction.
3. Non-maximal suppression. Edges will occur at points the
where the gradient is at a maximum. Therefore, all
points not at a maximum should be suppressed. In order
to do this, the magnitude and direction of the gradient is Fig 3: Discrete approximation to Gaussian function
computed at each pixel. Then for each pixel check if the  The second step is to find the edge strength by
magnitude of the gradient is greater at one pixel's  taking the gradient of the image. 
distance away in either the positive or the negative  Then, the approximate absolute gradient magnitude
direction perpendicular to the gradient. If the pixel is not  (edge strength) at each point can be found. 
greater than both, suppress it.  It uses a pair of 3x3 convolution masks, one
4. Edge Thresholding. The method of thresholding  estimating the gradient in the x-direction (columns). 
used by the Canny Edge Detector is referred to as  The other convolution masks estimating the gradient
"hysteresis". It makes use of both a high threshold and a low in the y-direction (rows). They are shown below: 
threshold. If a pixel has a value above the high threshold, it is
www.ijmer.com 553 | Page
 International Journal of Modern Engineering Research (IJMER)
www.ijmer.com Vol.2, Issue.2, Mar-Apr 2016 pp-552-554 ISSN: 2249-6645
CONCLUSION
The result of the image can be obtained by using
MATLAB.Thus compared to other edge detection algorithm,
Canny edge detection algorithm use probability for finding
error rate localization and response in various images. Thus
even blurred image and images with noise can be detected
accurately using this Canny edge detection algorithm as there
The magnitude or edge strength of gradient is then is a specification of improving signal to noise ratio. It results
approximated using the formula: in a significant speedup and successfully detects the edges of
the images.
|G| = |Gx| + |Gy|
FUTURE WORK
The direction of the edge is computed using the gradient in
the x and y directions. Whenever the gradient in the x Thus as a future enhancement, the implementation of Canny
direction is equal to zero, the edge direction has to be equal edge detection algorithm can be perform on FPGA platform
to 90 degrees or 0 degrees, Formula for finding edge and tested in system C. Thus tested in system C which can
direction: overcome the major disadvantage of canny edge detection
algorithm of complex computation and time consuming. It is
-1 capable of supporting fast real time edge detection for images
θ = tan (Gy / Gx)
and videos.
If GY has a value of zero, the edge direction will equal to 0
degrees. Otherwise the edge direction is equal to 90 degrees. REFERENCES
[1] S. Varadarajan, C. Chakrabarti, L. J. Karam, and J.
III Simulation results M.Bauza, “A distributed psycho-visually motivated
The algorithm performance was tested using a variety of Canny edge detector,” IEEE ICASSP, pp.882-825,
images. The simulations results are obtained in MATLAB. Mar.2010.
Here the results are tested by two images. [2] D. V. Rao and M. Venkatesan, “An efficient
reconfigurable architecture and implementation of
(a). Finger print and edge detection algorithm using Handle-C,” ITCC,
vol. 2, pp. 843 – 847, Apr. 2004.
(b). Eye [3] W. He and K. Yuan, “An improved Canny edge
detector and its realization on FPGA,” WCICA, pp.
Simulation output: 6561 –6564, Jun. 2008.
[4] F. M. Alzahrani and T. Chen, “A real-time edge
detector algorithm and VLSI architecture,” Real-Time
Imaging, vol. 3, no. 5, pp. 363 – 78, 1997.
[5] J.F. Canny, “A computation approach to edge
detection,” IEEE Transactions on Pattern Analysis and
Machine Intelligence, vol. 8, no.6, pp. 769-798, Nov
1986.
[6] D. Marr and E. Hildreth, ”Theory of edge Detection,”
Proc. Royal Soc. of London, series B, vol. 207, pp.
187-217, 1980.
[7] D. Demigny, and T. Kamle,” A discrete expression of
Canny’s criteria for step edge detector performances
evaluation”, IEEE Transactions on Pattern Analysis
and Machine Intelligence, 19(6), pp. 1199-1211, 1997.
Fig 4: Matlab simulation result for Finger print image
[8] L. Torres, M. Robert, E. Bourennane, and M.
Paindavoine, “Implementation of a recursive real time
edgedetector using retiming techniques,” VLSI, pp.
811 –816, Aug. 1995.

Fig 5: Matlab simulation result for Eye image

www.ijmer.com 554 | Page