An Edge Detection Method for Hexagonal Images

Kazi Mostafa & Innchyn Her
International Journal of Image Processing (IJIP), Volume (10) : Issue (4) : 2016 161
An Edge Detection Method for Hexagonal Images
Kazi Mostafa kazimostafa@gmail.com
Mechanical Engineering
Shandong University of Technology
Zibo, China
Innchyn Her her@mail.nsysu.edu.tw
Mechanical & Electro-mechanical Engineering
National Sun Yat-sen University
Kaohsiung, 804, Taiwan
Abstract
This paper presents a morphological image processing operation for hexagonally sampled
images and proposes a new edge detection method for these images by using a grayscale
morphology. This is achieved by applying morphological gradient operators and multiscale top-
hat transformations (white and black top-hat transformations) to hexagonal images. The proposed
study includes a method for converting hexagonally sampled images as well as the processing
and subsequent display of images on a hexagonal grid. Performance evaluation were performed
to assess the proposed method. The proposed study shows that a method of edge enhancement
by applying three by three hexagonal structuring element achieves results superior to those of a
rectangular images. The results indicated that the proposed edge detection algorithms improved
substantially after implementation of the edge enhancement method.
Keywords: Hexagonal Grid, Mathematical Morphology, Top-hat Transform, Edge Detection.
1. INTRODUCTION
Edge detection and enhancement [1, 2] are essential procedures in the fields of image
processing, computer vision, robotics, and automation. Several techniques are used for edge
detection. Amongst these techniques, morphological image processing is a conventional
approach. However, the success of edge detection relies heavily on the shape of the object in an
image. Morphological gradient often performs poorly when applied to rectangular image curved
objects. To overcome this limitation, high-resolution grids are frequently used; however, using
high-resolution images is expensive and reduces computational performance. Sampling on a
hexagonal lattice is an alternative solution for overcoming this difficulty [3, 4]. Hexagonal image
processing provides decreased storage and computation time, increased coding efficiency, less
quantisation error, the property of equidistance, and consistent connectivity [5]. In addition, the
sampling efficiency of a hexagonal lattice exceeds that of a rectangular lattice [6], and hexagonal
images have been determined to be more aesthetically pleasing to the human eye. Recently,
many companies have introduced image acquisition and display devices which use increased
numbers of rectangular pixels per inch. These devices require additional computational power
and storage capacity. Some applications, such as real-time monitoring programmes, require the
use of low-resolution imaging to save computational power and storage. Hexagonal grid systems
are an effective alternative for solving these problems. The main obstacle limiting the use of
hexagonal images has been the lack of hardware for capturing and displaying hexagonal-grid-
based images [5]. To overcome this obstacle, a resampling method [6-9] has been developed to
convert rectangular grid images to hexagonal counterparts. This method is based on image
resampling to generate a hexagonally sampled image from a traditional rectangular grid image.
A detailed overview of morphological operations such as dilation, erosion, opening, and closing is

provided in [10]. In addition, a morphological edge detection algorithm was proposed therein to
detect the edges of lung computed tomography images with salt-and-pepper noise [11]. The
method is based on a combination of dilation, erosion, opening, and closing; dilation and erosion
are employed to detect edges, and opening-closing operations are used as preprocessing
procedures to eliminate noise. In recent studies, morphological edge detection has been
proposed and compared with Prewitt, Laplacian of Gaussian, and Canny edge operators [12, 13].
Moreover, multiscale top-hat transformations [14, 15] based on structuring elements (SEs) have
been constructed for various applications.
Hexagonal image processing is as old as image processing itself but sufficient research has been
lacking because the availability of hexagonal pixel display devices is limited. In this paper, an
edge detection method that involves applying a grayscale morphology to hexagonally sampled
images is proposed to alleviate this problem. The proposed edge detection and enhancement
method entails converting rectangular grid images into hexagonal grid image and is suitable and
convenient for traditional display devices.
This guideline is used for all journals. These are the manuscript preparation guidelines used as a
standard template for all journal submissions. Author must follow these instructions while
preparing/modifying these guidelines. This guideline is used for all journals. This guideline is used
for all journals. These are the manuscript preparation guidelines used as a standard template for
all journal submissions. Author must follow these instructions while preparing/modifying these
guidelines. This guideline is used for all journals.
2. RESEARCH OBJECTIVES
This paper proposes a design framework to overcome the limitations of edge detection by using
rectangular grid imaging. In our framework, a morphological gradient operator and multiscale top-
hat transformations to hexagonally sampled images are applied. To exploit the advantages of
hexagonal images for broader applications, this study developed an efficient method for detecting
edges by using a grayscale morphological operator. Our objectives are summarized as follows:
 To provide a comprehensive guideline of hexagonally sampled images converted from
rectangular images, and the processing and display of images on hexagonal grids. Image
processing involves applying grayscale morphological operations to hexagonal
counterparts to detect edges; specifically, morphological gradient operators are to detect
edges and multiscale top-hat transformations are employed to enhance edge detection
performance.
 To develop multiscale SEs with various sizes for hexagonally sampled images and then
select the better structuring element for use in the proposed method.
 To evaluate the performance of the proposed algorithm by using qualitative comparison
and the linear index of fuzziness.
3. BACKGROUND
3.1 Morphological Image Processing
Mathematical morphology is an approach for extracting geometrical features from signals based
on set theory [10]. Morphological edge detection [16] has generally been performed using SEs
[17, 18]. The most basic morphological operations are dilation and erosion. Grayscale erosion
and dilation produce results identical to those of nonlinear minimum and maximum filters. The
general effects of performing dilation on a grayscale image are that the output image tends to be
brighter than the input, and dark details are reduced. The dilation and erosion of grayscale image
A by SE B is shown as follows:
},D),(;)(),(|),(),(max{),)(( BA yxDytxsyxBytxsftsBA 
(1)
},D),(;)(),(|),(),(min{),)(( BA yxDytxsyxBytxsftsBA 

3.2 Hexagonal Image Processing
A rectangular grid of sampling points and pixels is shown in Figure 1(a). Several types of
sampling scheme were listed by Whitehouse 19. Figures 1(b, c) show hexagonally sampled
image, which are characterized by the hexagonal tiling. Hexagonal grids have numerous
advantages in many applications, e.g., robot vision [20], deep-space imaging [21], image
processing. For a morphological operator, a 3 by 3 SE is the closest fit for a circular SE in a
rectangular grid image. However, the use of such an element to erode curved objects can
introduce unwanted discontinuities. This is because the elements within the SE are at different
distances from the center. The consistent connectivity of pixels in hexagonal images is an
attractive feature which improves circular SE definition, as shown in Figure 2. Moreover,
computation of a neighborhood is simple in a hexagonal lattice, as shown in Figure 3.
FIGURE 1: Image tiling: (a) square grid; (b) hexagonal grid with hexagonal tiles; (c) hexagonal grid with
rectangular tiles.
FIGURE 2: Distance in a square grid & hexagonal grid.
FIGURE 3: Neighborhood relationships.
4. METHODOLOGY
The current study was divided into three phases as depicted in Figure 4. We first used an image
resampling method to convert the grid of an image from rectangular to hexagonal one and then
performed edge detection by using the morphological gradient method. Edges were enhanced by
applying multiscale top-hat transformations. Finally, performance evaluation algorithms were
employed to assess the results of the proposed methods. We converted rectangular grid image to
hexagonal grid by using simulated hexagonal grid method. Simulated hexagonal grid can be
created through clusters of rectangular sub-pixels. Many rectangular grids can be combined
together to create clusters of sub-pixels (as shown in Figure 5). For example, d = 7; is composed
of 120 rectangular grids to create one hexagonal pixel. A cluster of sub-pixels will closely
resemble the shape of hexagon. D7 simulated hexagonal conversion method performed better
than others. The geometric properties of D7 hexagonal pixel are similar to hexagon shape and it
showed equivalent inter-pixel distance and smooth curved edges. Moreover, rectangular grid may
perform well only for horizontal and vertical direction. Proposed D7 hexagonal grid performed not

only for horizontal and vertical lines but also oblique edges. The research methodology can be
summarized as follows:
Step 01: Image resampling is conducted to convert rectangular grid input images into hexagonally
sampled images.
Step 02: Several SEs of various sizes are used in dilation and erosion.
Step 03: Edges are detected by calculating the differences between the dilated and eroded
images of all SEs and then choosing the better SE.
Step 04: Detected edges are enhanced by applying multiscale top-hat transformations.
Step 05: Performance evaluation is conducted using QC and the index of fuzziness.
FIGURE 4: System diagram for performing image processing on a hexagonally sampled grid.
FIGURE 5: Simulated hexagonal grid.
4.1 Resampling (Square to Hexagonal)
In this study, we employed a simulated hexagonal grid to convert the sampling of images from
rectangular to hexagonal. A simulated hexagonal grid can be created using clusters of
rectangular subpixels5, which are created by combining several rectangular grids, as shown in
Figure 5. For example, the D7 method entails combining 120 rectangular grids to create one
hexagonal pixel. In this resampling method, a cluster of subpixels closely resembles a hexagon.
Each simulated hexagonal pixel is formed by 120 sub-pixels arranged. The light intensity of each

constructed hexagonal pixel can be easily computed as the average of the intensities of the 120
sub-pixels forming the hexagonal pixel. Hence, the number of hexagonal pixels is less than the
number of square pixels to cover the same image. From the observation result obtained, it is
claimed that fewer sampling points (or pixels) are required with a hexagonal structure to maintain
equal amount of image information (or the same image resolution) with the traditional square
structure. The hexagonal pixels constructed in the way above will not lose image resolution.
Following the resampling procedure, of which the result is shown in Figure 6(b), the dimensions
were 512 × 512. A small portion of the enlarged view was added at the right side of the standard
test image Lena to highlight the rectangular and hexagonal grids. To increase the computation
speed, we converted these dimensions to 256 × 256. We applied sub pixel-size and D7 method
to generate hexagonal pixels to obtain the corresponding intensity values of a simulated
hexagonal grid.
FIGURE 6: Image resampling (a) rectangular grid image (b) hexagonal grid image (512 by 512).
FIGURE 7: Symmetrical hexagonal coordinate frame.
However, hexagonal pixels cannot be labelled in a normal column-row order as in a rectangular
grid. To properly address and store the hexagonal image data, we applied symmetrical
coordinate systems 7. A three-coordinate system was developed to represent hexagonal data. As
shown in Figure 7, the system uses a three-tuple to represent the displacement from each of the
major axes of the hexagon. The three coordinates are x = 0, y = 0, and z = 0.
4.2 Hexagonal Structuring Elements
After image conversion, the application of morphological operations to hexagonal images to
detect edges was conducted. The performance of the morphological operation depends heavily
on the size and shape of the SE. When certain SEs are used, erosion and dilation can be
achieved more rapidly and with more favourable results. In this paper, four (3 × 3, 5 × 5, 7 × 7
and 9 × 9) new hexagonal grid SEs are proposed to detect and enhance edges in hexagonally
sampled images, as shown in Figure 8.

FIGURE 8: Proposed structuring elements.
4.3 Hexagonal Grayscale Morphological Operator
The most basic morphological image processing operations are dilation and erosion. Grayscale
erosion and dilation produce results identical to those of nonlinear minimum and maximum filters.
This study presents grayscale dilation and erosion on hexagonally sampled images, as shown in
Figure 9. Grayscale dilation and erosion were achieved by applying four new hexagonal SEs (3 ×
3, 5 × 5, 7 × 7, 9 × 9) to dilate and erode standard test images (Lena, Barbara, Pepper). Figures
10 and 11 show 3 × 3 and 5 × 5 hexagonal SE operations applied to the Lena test image. After
dilation and erosion, closing and opening, which are sequential combinations of dilation and
erosion, were performed by using aforementioned hexagonal SEs.
FIGURE 9: Dilation and Erosion on hexagonal image.
FIGURE 10: Dilated image (a) 3x3 SE, (b) 5x5 SE.

FIGURE 11: Eroded Image (a) 3x3 SE (b) 5x5 SE.
Conventional mathematical morphology entails implementing single-scale analysis. A fixed SE is
used based on prior knowledge. However, prior knowledge is not always available for some
applications. Thus, multiscale morphological analysis [22] seems to be more favourable than
single-scale analysis. Multiscale morphology can be implemented by using hexagonal SEs with
various scales. Let n (n= 1, 2, …. nmax) be the scale and B be the unit SE. Thus, the SE used at
scale n can be defined as:
,
1
   
timesn
BBBBBnB


(2)
where B is the SE and n is the number of operations.
Dilation and erosion can then be obtained:
Multiscale Dilation = A⊕nB, and Multiscale Erosion = AΘnB,
where A is the input image, B is the SE, and n is the number of operations.
Correspondingly, opening and closing are considered extensions of basic dilation and erosion
operations. Opening consists of erosion followed by dilation. Closing is dilation followed by
erosion. Multiscale opening and closing can be denoted by:
Multiscale Opening = AonB= (AΘnB)⊕nB, (3)
Multiscale Closing = A●nB= (A⊕nB)ΘnB,
4.4 Edge Detection (Morphological Gradient)
A morphological gradient is the difference between the dilation and the erosion of a given image.
The most basic morphological gradient operator is shown as follows:
Edge = (A⊕B) – (AΘB), (4)
A few more operators based on the most basic operator are available; for example,
Edge = (A●B) – (AoB), (5)
Multiscale morphological gradient operators at scale n can be defined as the differences between
dilation and erosion or between closing and opening:
Multiscale Gradient = (A⊕nB) – (AΘnB), (6)
Multiscale Gradient = (A●nB) – (AonB),
After dilation, erosion, opening, and closing, the proposed method involves detecting edges by
applying a morphological gradient operator to a hexagonal image. The Lena test image was
dilated and eroded by applying 3 × 3, 5 × 5, 7 × 7, and 9 × 9 hexagonal SEs. The eroded image
was then subtracted from the dilated image. The transformed Lena, Barbara, and Pepper images
exhibited numerous edges, as shown in Figures 12, 13, 14, and 15.

FIGURE 12: Edge image by using hexagonal morphological gradient operator (Lena) (a) 3x3; (b) 5x5.
FIGURE 13: Edge image by using hexagonal morphological gradient operator (Pepper) (a) 3x3; (b) 5x5.
FIGURE 14: Edge image by using hexagonal morphological gradient operator (Barbara) (a) 3x3(b) 5x5.
FIGURE 15: Multiscale hexagonal morphological gradient 3x3 SE (Lena) (a) n=1(b) n=2(c) n=3(d) n=4.
4.5 Edge Enhancement (Top-Hat Transformation)
To enhance the edges detected from single- and multiscale morphological gradient operations, a
multiscale top-hat transformation algorithm is used. Recently, multiscale theory has become an
attractive option for image processing [23, 24]. Based on opening and closing, the classical top-
hat transformation defined [10] by White (WTH) entails computing the morphological opening of

an image and then subtracting the result from the original image. The WTH transformation returns
elements that are smaller than the SEs and brighter than their surroundings. The black top-hat
(BTH) transformation involves subtracting the original image from morphological closing results.
Thus, top-hat transformations using multiscale hexagonal SEs [14, 15] yield a better output. The
WTH and BTH of gray image by applying SE are defined as follows:
(7)
,
where ⊕ are erosion and dilation operations, respectively, and B is a multiscale hexagonal
SE with various sizes (3 × 3, 5 × 5, 7 × 7 and 9 × 9). According to the formula presented
previously, light and dark image regions yield the highest gray values. Therefore, they can be
defined as follows:
MW = max (WTH1, WTH2, … , WTHm), (8)
MB = max (BTH1, BTH2, … , BTHm),
where m is the size of hexagonal SEs (e.g., m1 is a 3 × 3 SE, m2 is a 5 × 5 SE, m3 is a 7 × 7
SE, and m4 is a 9 × 9 SE). Thus, edge can be enhanced as follows:
Enhanced Edges = (Original Image x W1) + (MW x W2) – (M B x W3), (9)
where W1, W2, W3 are the weights for adjusting the results for various applications. Using a low
W1 and high W2 and W3 brightens the light regions. Thus, we set W1 = 1 and W2 and W3 = 3.
The edge enhancement formula is applied to edges detected from a single-scale morphological
gradient, as shown in Figures 12, 13, and 14. Enhanced edges are shown in Figures 16 and 17.
FIGURE 16: Enhanced edges after applying top-hat transform; (a) 3x3(b) 5x5.
FIGURE 17: Enhanced edges after applying top-hat transform (3x3 SE); (a) Pepper (b) Barbara.

5. PERFORMANCE EVALUATION
Two performance evaluation measures, namely the linear index of fuzziness [25] was applied to
assess the results of the proposed method. The index of fuzziness was used to quantitatively
assess the performance of the proposed algorithm. The linear index of fuzziness is based on
spatial domain analysis and is defined as:
; (10)
where fxy and fmax represent the gray value of the pixel (x, y) and the maximum gray value of an
image with size M × N, respectively. The fuzzy property is denoted by . A lower value of γ
indicates a more desirable performance.
SE Lena Pepper Barbara
Index of fuzziness Index of fuzziness Index of fuzziness
Hexagonal
Grid
3x3 0.064 0.062 0.116
5x5 0.137 0.137 0.203
7x7 0.201 0.205 0.283
9x9 0.255 0.265 0.353
Rectangula
r
Grid
3x3 0.098 0.091 0.201
5x5 0.247 0.207 0.302
7x7 0.301 0.306 0.386
9x9 0.359 0.369 0.464
TABLE 1: Comparison of different HSE edge detection using morphological gradient.
The linear index of fuzziness values for the morphological gradient edge detection operators of
various SEs are listed in Table 1. Table 2 shows comparisons of the linear index of fuzziness
values amongst the test images when the proposed edge enhancement method was used. A
lower index of fuzziness value represents a more desirable result. Table 1 shows that 3 × 3
hexagonal SEs (HSEs) lower index of fuzziness. Thus, the 3 × 3 HSE was better than rectangular
counterpart. We applied the proposed edge enhancement method to the four edge images listed
in Table 1 showed in Table 2. For 3 × 3 HSE edges, the proposed method produced a lower
index of fuzziness compared with rectangular grid image. Moreover, comparing Tables 1 and 2
revealed that edge results improved after the proposed edge enhancement method was applied.
Conclusions were formulated based on the figures showing the enhanced edge images as well as
the linear index of fuzziness values. Comparing the hexagonal SE (HSE) edge detection and
enhanced edge method images clearly indicated that the proposed algorithm outperformed in
edge detection and edge enhancement. The linear index of fuzziness demonstrate that the
proposed 3 × 3 HSE method achieved the lowest index of fuzziness. A substantial difference was
observed between the hexagonal and rectangular grids. As shown in Table 1 and 2, the index of
fuzziness also indicated that the proposed method showed quantitatively favorable results. A
comparison of the two grids revealed that, overall, the edge of the proposed hexagonal method
was more continuous, well-defined and clearer.

SE Lena Pepper Barbara
Index of fuzziness Index of fuzziness Index of fuzziness
Hexagonal
Grid
3x3 0.054 0.055 0.101
5x5 0.126 0.105 0.155
7x7 0.153 0.133 0.189
9x9 0.181 0.175 0.229
Rectangular
Grid
3x3 0.089 0.090 0.177
5x5 0.168 0.110 0.232
7x7 0.219 0.219 0.286
9x9 0.279 0.257 0.374
TABLE 2: Comparison of different HSE enhanced edge detection
6. CONCLUSION
Our aim was to explore hexagonal grid and compared with traditional rectangular grid. Moreover,
to see the hexagonal grid would improve the performance of edge detection method for low
resolution images. This paper presents a new edge detection and enhancement method based
on hexagonally sampled images and a grayscale morphology. Edges were first detected by
performing hexagonal morphological gradient operations. The edge enhancement method was
subsequently applied to achieve better results. In both edge detection and enhancement, 3 × 3
hexagonal SE performed the most favorably. As shown in Tables 1 and 2, a three by three
hexagonal structuring element achieves results superior to those of a rectangular image. Thus,
hexagonal rather than rectangular images can be employed in edge detection by applying
morphological operators to produce better results. The present study was designed to determine
the effectiveness of a grayscale morphology in improving hexagonal images. The proposed
method entails detecting edges by using a gray morphology and enhancing edges by using
multiscale top-hat transformations. The results indicated that the edge detection performance
improved significantly after the proposed edge enhancement method was implemented. In
conclusion, from the performance evaluation and visual comparison, we determined that
consistent and accurate gradient operators can be obtained by applying hexagonal lattices. The
results clearly showed that the hexagonal grid performed adequately for curved objects as well as
for horizontal and vertical lines.
7. REFERENCES
[1] Y. Liu and S. Zhou, A self-adaptive edge matching method based on mean shift and its
application in video tracking. The Imaging Science Journal, 2014, 62(4), 206-216.
[2] Y. Xiong, J. Li, X. Zuo, and Z. Chen, Research on an Edge Detection Algorithm of Remote
Sensing Image Based on Wavelet Enhancement and Morphology. Journal of Computers,
2014, 9(5), 1247-1252.
[3] S. Coleman, B. Scotney, and B. Gardiner: 'A Biologically Inspired Approach for Fast Image
Processing', 2013, IAPR.
[4] S. A. Coleman, B. W. Scotney, and B. Gardiner: 'Design of Feature Extraction Operators for
Use on Biologically Motivated Hexagonal Image Structures', MVA, 2009, 178-181.
[5] L. Middleton and J. Sivaswamy: 'Hexagonal image processing: A practical approach'; 2006,
Springer.
[6] R. C. Staunton, The processing of hexagonally sampled images. Adv. Imaging Electron
Phys, 2001, 119, 191-265.

[7] I. Her, A symmetrical coordinate frame on the hexagonal grid for computer graphics and
vision. Journal of Mechanical Design, 1993, 115(3), 447-449.
[8] I. Her, Geometric transformations on the hexagonal grid. Image Processing, IEEE
Transactions on, 1995, 4(9), 1213-1222.
[9] I. Her and C.-T. Yuan, Resampling on a pseudohexagonal grid. CVGIP: Graphical Models
and Image Processing, 1994, 56(4), 336-347.
[10] J. Serra, Introduction to mathematical morphology. Computer vision, graphics, and image
processing, 1986, 35(3), 283-305.
[11] Z. Yu-qian, G. Wei-hua, C. Zhen-cheng, T. Jing-tian, and L. Ling-Yun: 'Medical images edge
detection based on mathematical morphology', Engineering in Medicine and Biology Society,
2005. IEEE-EMBS 2005. 27th Annual International Conference of the, 2006, IEEE, 6492-
6495.
[12] B. Kaur and A. Garg: 'Mathematical morphological edge detection for remote sensing
images', Electronics Computer Technology (ICECT), 2011 3rd International Conference on,
2011, IEEE, 324-327.
[13] T.-C. Su, M.-D. Yang, T.-C. Wu, and J.-Y. Lin, Morphological segmentation based on edge
detection for sewer pipe defects on CCTV images. Expert Systems with Applications, 2011,
38(10), 13094-13114.
[14] X. Bai and F. Zhou, Analysis of different modified top-hat transformations based on
structuring element construction. Signal Processing, 2010, 90(11), 2999-3003.
[15] X. Bai, F. Zhou, and B. Xue, Infrared image enhancement through contrast enhancement by
using multiscale new top-hat transform. Infrared Physics & Technology, 2011, 54(2), 61-69.
[16] T. A. Mahmoud and S. Marshall, Edge-detected guided morphological filter for image
sharpening. EURASIP Journal on Image and Video Processing, 2008, 2008.
[17] B. Chanda, M. K. Kundu, and Y. Vani Padmaja, A multi-scale morphologic edge detector.
Pattern Recognition, 1998, 31(10), 1469-1478.
[18] J. GUO, S. PAN, and X.-j. HU, Edge Detection in Tobacco Leaf Image Based on Grayscale
Morphology [J]. Computer Engineering, 2007, 21, 060.
[19] D. Whitehouse and M. Phillips, Sampling in a two-dimensional plane. Journal of Physics A:
Mathematical and General, 1985, 18(13), 2465.
[20] R. C. Staunton: 'Hexagonal image sampling: A practical proposition', 1988 Robotics
Conferences, 1989, International Society for Optics and Photonics, 23-27.
[21] J. Allen: 'Perfect reconstruction filter banks for the hexagon grid', Information,
Communications and Signal Processing, 2005 Fifth International Conference on, 2005,
IEEE, 73-76.
[22] P. Maragos, Pattern spectrum and multiscale shape representation. Pattern Analysis and
Machine Intelligence, IEEE Transactions on, 1989, 11(7), 701-716.
[23] I. De, B. Chanda, and B. Chattopadhyay, Enhancing effective depth-of-field by image fusion
using mathematical morphology. Image and Vision Computing, 2006, 24(12), 1278-1287.
[24] M. Oliveira and N. J. Leite, A multiscale directional operator and morphological tools for
reconnecting broken ridges in fingerprint images. Pattern Recognition, 2008, 41(1), 367-377.

[25] H. R. Tizhoosh, Fuzzy image processing. Publisher: Springer-Verlag. Kartoniert (TB),
Deutsch, 1997.

An Edge Detection Method for Hexagonal Images

Recommended

More Related Content

What's hot (19)

Similar to An Edge Detection Method for Hexagonal Images (20)

Recently uploaded (20)

An Edge Detection Method for Hexagonal Images