International Journal of Management, Technology And Engineering ISSN NO : 2249-7455

Zero crossing Image Edge Detection For MRI

Scanning
Venkata Raghavendra1, E.Lakshmi Prasad2
Department of Electronics and Communication Engineering
1,2*
Dept. Of E.C.E., Jawaharlal Nehru Technological University Anantapur College of Engineering, Anantapur, India
[email protected]
Abstract—Edge detection is a process of identifying the sharpness of image discontinuities. It is usually characterized by the
boundary of an image and their internal edges occur on the boundary of the two regions. Several types of research are done for
extracting the edges from an image by using the following techniques such as Sobel, canny, Prewitt, Roberts etc., these edge
detection methodologies are also used for medical image process. In the proposed method, edge extraction is done for finding
the lesions (Precious) regions in PET, MRI, CT scan images. Edge detection methods are implemented on FPGA for MRI scanned
images to achieve better accuracy. Zero-crossing-based edge detection framework is designed to identify the edges for MRI
scanned images. Zero crossing able to identify the significant scale space edge points. So, our method is compared with state-of-
art criteria like Sobel, canny, Prewitt, Roberts. The above algorithm is designed and developed using Mat lab and implement the
proposed work on the FPGA board.

Keywords—Sobel, Prewitt, Canny, Robert, Zero crossing Edge Detections

I. INTRODUCTION

Edge detection is an image processing technique for locating the boundaries of objects within the image. This
approach is likely used for an image segmentation and data extraction for finding the edges in an image. The edges of an
image enclosed with the maximum data[1],[2]. So, the initial step is to determine the edges in an image through filtering by
enhancing the areas which include edges thereby developing the sharpness of the image. An edge shows the result in changes
of light, coloration, shade, and texture. These properties may be used to estimate the depth, size orientation and surface
properties of a digital image. An edge detection to determine in an image that requires to enable the filter the unwanted areas
by selecting the threshold factors. The detection of suitable modifications can be tuned up via noise and it relies upon at the
pixel, change in the threshold that defines an edge. If the image is damaged through the noise, then it is difficult to detect the
edges of an image. Edge detection is a process of collecting the characteristic detection and feature extraction. The cause of
detecting sharp adjustments in image brightness is to discover the adjustments in properties of the arena. It may be established
that the discontinuities in an image brightness likely correspond to:

• Discontinuities depth,
• Discontinuities in surface orientation,
• Changes in material properties and
• Variations in scene illumination.

At the initial stage, it is to be noticed that the region detector of an image is hard and speedy of related curves which
advocate the limits to an object. The limits of surface markings are added to the curves that corresponded to discontinuities in
surface orientation. Edge detection follows set of rules to determine the edges for an image, it reduces the quantity of data to be
processed and filters the masses less relevant data on the same time as retaining the structural reed genes of the image. If this
step is successful, the problem of identifying the facts for any image can be simplified. However, it is not generally viable to
acquire to find out the fine edges for the real-life image.

Volume 8, Issue VIII, AUGUST/2018 Page No:1043

International Journal of Management, Technology And Engineering ISSN NO : 2249-7455

II. RELATED WORK

Edge detection is one of the methods in image processing, it can find the edges of an image. It detects the boundaries
based on discontinuities in brightness. The following are the common edge segmentation or edge detection algorithms:

• Sobel edge detection

• Canny edge detection
• Prewitt edge detection
• Roberts edge detection

A. Sobel Edge Detection

The Sobel operator is derived into two 3x3 kernels [3], [4], those are convolved with the original image to calculate
derivative approximations one for horizontal and another for vertical. Assume A as the source image, Gx and Gy are two images
at each point contain the horizontal and vertical derivative approximations respectively.

𝐺x = [𝑃7 + 𝐶𝑃8 + 𝑃9] – [𝑃1 + 𝐶𝑃2 + 𝑃3] (1)

𝐺y = [𝑃3 + 𝐶𝑃6 + 𝑃9] – [𝑃1 + 𝐶𝑃4 + 𝑃7] (2)
With the value clasp constant c=2. Therefore, the equation can be expressed as
𝐺𝑥 = {[(𝑥 + 1, 𝑦 − 1) + 2(𝑥 + 1, 𝑦) + 𝑓(𝑥 + 1, 𝑦 + 1 [(𝑥 − 1, 𝑦 − 1) + 2(𝑥 − 1, 𝑦) + 𝑓(𝑥 − 1, 𝑦 + 1) ] } (3)
and
𝐺𝑦 = [(𝑥 − 1, 𝑦 + 1) + 2(𝑥, 𝑦 + 1) + 𝑓(𝑥 + 1, 𝑦 )] − [ 𝑓(𝑥 − 1, 𝑦 − 1) + 2𝑓(𝑥, 𝑦 − 1) + 𝑓(𝑥 + 1, 𝑦 − 1)] (5)

At each point in the image, the resulting gradient approximations can be combined, and it gives the gradient magnitude, Using:

G = √𝐺𝑥2 + 𝐺𝑦2 (6)

Using this equation-6, gradient’s direction can be calculated:

𝐺
𝜃 = 𝑎𝑟𝑐𝑡𝑎𝑛 𝑎𝑟𝑐 tan ( 𝑦 ) (7)
𝐺𝑥

B. Canny Edge Detection:

The Canny edge detection is also one kind of methods in image processing [5]. Canny edge detection operator is a multi-
stage algorithm that finds a wide range of edges in the image. This method is used to determine the image edges in a variety of
directions such as horizontal, vertical and diagonal edges in the blurred image. Canny edge detection consists of four filters to
detect the boundaries of an image. The edge detection operator returns a value for the first derivative in the horizontal direction
(GX) and the vertical direction (Gy). From this the edge gradient and direction can be determined:

G = √𝐺𝑥2 + 𝐺𝑦2 (8)

Angle (

C. Prewitt Edge Detection:

Prewitt method is also like Canny and Sobel for detecting edges in an image. It can detect two types of edges:

• Horizontal edges
• Vertical edges

Volume 8, Issue VIII, AUGUST/2018 Page No:1044

International Journal of Management, Technology And Engineering ISSN NO : 2249-7455

Edges are estimated based on the difference between adjacent and corresponding pixel intensities of an image. All the masks
are used for detection which is also known as derivative masks[6],[7]. All the derivative masks consist of following properties:

• Opposite sign should present in the mask.

• Sum of the mask should be equal to zero.
• More weight means more edge detection.

Prewitt operator provides with two masks one for detecting edges in the horizontal direction and another for detecting edges
in a vertical direction.

D. Roberts Edge Detection:

The Roberts cross operator is an edge detection operator in image processing and computer vision for edge detection.
The state of affairs remained one of the essential element detectors and stayed initially proposed by Lawrence Roberts in 1963
[8]. Robert edge detection properties are:

• The proposed edges should be well-defined

• The background should contribute as little noise as possible
• The intensity of edges should correspond as close as possible.

Therefore, the gradient magnitude is given as:

G = √𝐺𝑥2 + 𝐺𝑦2 (5)

The direction of the gradient can also be defined as:

𝐺𝑦 (𝑥,𝑦)
𝜃(𝑥, 𝑦) = arctan ( ) (6)
𝐺𝑥 (𝑥,𝑦)

III. METHODOLOGY

A. Zero Crossing Edge Detection

The Zero crossing edge detector select the specific values in the Laplacian [9],[10] of an image wherein the change of
the Laplacian passes through zero. The points in which the Laplacian changes the edge sign. Generally, the elements are often
upward push up at `edges' in images i.e. whereas the Points in which intensity values change swiftly. However, in addition to
that arise at places that aren't as clean to accomplice with edges. It is higher to expect zero detail detectors as some shape of
function detector in the edge of a specific element detector. Zero crossings constantly lie on closed contours, and so the output
from the zero crossing detectors is usually a binary image with single pixel thickness showing the positions of the zero crossing
points.

The zero crossing detector is detecting the edges in the image by using the Laplacian of the Gaussian algorithm. The
result of the zero crossing operators is strongly inspired via the dimensions of the Gaussian used for the smoothing the image.
The core of the zero crossing detectors is the Laplacian of Gaussian which can clearly find the edges in an image. As defined
there, `edges' in images supply upward push to zero crossings within the LoG output.

However, zero crossings furthermore rise up in any place wherein the image depth gradient starts off advanced growing and this
can display the edges of an image. Zero crossing method is able to determine the edges in the image specifically at the very low
gradient in which the intensity gradient wobbles up and down spherical zero. Once the image has been LoG filtered, then discover
the zero crossings. This can be finished in several strategies.

Volume 8, Issue VIII, AUGUST/2018 Page No:1045

International Journal of Management, Technology And Engineering ISSN NO : 2249-7455

• The simplest way is set the edge of the LoG output at 0, to provide a binary image in which the boundaries among
foreground and facts edges represent the places of zero crossing factors. These limitations may be finding out in single
attempt e.g. the usage of some morphological operators. For example, to discover all boundary elements should mark
every foreground component that has as a minimum one historic beyond neighbor.
• Since the zero crossing method is determining the problems in among pixels which are eliminated by using LoG filtered
method. The possibility output example is an image grid this is spatially shifted half of a pixel within the path and half
of a pixel down, relative to the actual image graph. This sort of example is called a twin lattice. But this approach does
no longer localize the zero crossings more as it ought to be.
• A more appropriate approach is to carry out the shape of interpolation to estimate the placement of the zero crossing to
sub-pixel precision.

Input image

Canny edge detection

Sobel edge detection

Prewitt edge detection

Zero crossing edge detection

Figure 1: Flow of the proposed method

IV. RESULTS AND DISCUSSION

Figure 1: Original image

Figure1 represents the original MRI image that is considered as an input for the proposed approach. By taking this MRI input we
performed different edge detection techniques.

Volume 8, Issue VIII, AUGUST/2018 Page No:1046

International Journal of Management, Technology And Engineering ISSN NO : 2249-7455

Figure 2:sobel edge detection at different t values

Sobel edge detection at different t values is represented in figure2. The first image is taken on the value of t=0.06 and the second
one is taken at the fee of t=0.15. The 0.33 image is evaluated on the value of 0.90 and the final is evaluated at the value of -45 at
the value of t are 0.90.

Figure 3: Sobel edge detection

Sobel edge detection is represented in figure3. The Sobel operator performs a 2-D spatial gradient length on an image. Generally,
it is used to estimate the distance to locate the approximate absolute gradient charge at each factor in an input grayscale image.

Volume 8, Issue VIII, AUGUST/2018 Page No:1047

International Journal of Management, Technology And Engineering ISSN NO : 2249-7455

Figure 4: Prewitt edge detection

The Prewitt Edge detection is showed in figure4. This method is used to discover the edges through making use of a horizontal
and vertical filter out in the collection. Both filters are implemented and summed to get the final result.

Figure 5: Robert edge detection

The Roberts edge detection is computed and a couple of-D spatial gradient sizes on an image graph. It mainly highlights the
edges of excessive spatial gradient which often corresponding to edges. The input to the operator is a grayscale image as same
due to the truth the output. Pixel values at every trouble within the output represent the predicted absolute importance of the
spatial gradient of the input image at that element.

Volume 8, Issue VIII, AUGUST/2018 Page No:1048

International Journal of Management, Technology And Engineering ISSN NO : 2249-7455

Figure 6: Laplacian of Gaussian method

The Laplacian edge detection is showed in figure6. It calculates the second order derivatives in a single segment. This detector
is extremely sensitive to noise.

Figure 7: Canny edge detection

Canny edge detection showed in figure7, this method is consists of a multi-step algorithm that can detect edges with suppressed
Noise. The cause of edge detection is mainly to reduce the image data, the structural properties are to be used for image
processing.

Volume 8, Issue VIII, AUGUST/2018 Page No:1049

International Journal of Management, Technology And Engineering ISSN NO : 2249-7455

Figure 8: zero-cross method

The zero crossing detectors showed in figure-8 that identifies in which the cost of the Laplacian passes through zero edges in the
MRI image. Zero crossings constantly lie on closed contours, and so the output from the Zero crossing detectors is often a binary
image with unmatched pixel thickness lines at the zero crossing factors.

Figure 9: Edge detection of all methods

The edge detection of all the methods is represented in a single window shown in figure 9.

Table: Comparison of Edge detection algorithms.

Metrics Sobel Prewitt Canny Zero

cross

MSE 0.0607 0.0607 0.0728 0.0617

PSNR 60.3354 60.3361 59.5455 60.2586

MSE: Mean Square Error

PSNR: Peak Signal to Noise Ratio

Volume 8, Issue VIII, AUGUST/2018 Page No:1050

International Journal of Management, Technology And Engineering ISSN NO : 2249-7455

V. CONCLUSION AND FUTURE SCOPE

Zero crossing edge detection method is used to extract edges in an image. We implemented our algorithm on FPGA board and
our method yields better results compared with state of art scenario. We prove that zero crossing edge detection method yields
better compared to canny, Sobel, Prewitt. In the near future, we would like to implement this edge detection methods for other
medical images.

ACKNOWLEDGMENT
It is a great opportunity for me to work on the subject like “Zero crossing Image Edge Detection For MRI Scanning”. Thank
the anonymous referees for their useful suggestions.

REFERENCES

[1] John F. Canny, “A Computational Approach to Edge Detection”. IEEE Transactions on Pattern Analysis and Machine
Intelligence, Vol. PAMI - eight, No. 6, November 1986.
[2] John F. Canny, “Finding Edges and Lines in Images”. M.I.T. Artificial Intelligence Lab., Cambridge, Massachusetts,
Rep. Al-TR-720, 1983.
[3] R. W. Hamming, “Digital Filters”. Englewood Clis, NJ. Prentice Hall, 1983.
[4] Leslie Lam port, “LATEX: A Document Preparation System”. Addison Wesley, Massachusetts, Second Edition, 1994.
[5] Agaian, S. S., Baran, T. A., & Panetta, K. A. (2003). “Transform-based totally image compression via manner of noise
discount and spatial change the usage Of Boolean minimization”. IEEE Workshop on Statistical Signal Processing, 28
Sept.-1 Oct. Pp. 226 – 229.
[6] Baker, S., &Nayar, S. K. (1996). Pattern rejection. Proceedings of IEEE Conference Computer Vision and Pattern
Recognition, 544-549.
[7] D. Marr and E. C. Hildreth, “Theory of Edge Detection”, Proceedings of the Royal Society of London, vol. B
207, pp. 187-217, 1980.
[8] J. Canny, “A Computational Approach to Edge Detection”, IEEE Transactions on Pattern Analysis and
Machine Intelligence, vol. 8, no. 6, pp. 679-698, 1986.
[9] S. M. Smith and J. M. Brady, “SUSAN – A New Approach to Low-Level Image Processing”, International
Journal of Computer Vision, vol. 23, no. 1, pp. 45-78, 1997
[10] M. Basu, “Gaussian-Based Edge-Detection Methods – A Survey”, IEEE Transactions on Systems, Man, and
Cybernetics – Part C: Applications and Reviews, vol. 32, no. 3, pp.. 252-260, 2002.

Volume 8, Issue VIII, AUGUST/2018 Page No:1051