![Pushpajit A. Khaire & Dr. N. V. Thakur
International Journal of Image Processing (IJIP), Volume (6): Issue (6), 2012 404
and image processing. In recent years, researchers have applied various soft computing
techniques for edge detection to improve segmentation results for various images and to enhance
edge detection technique. Canny [1] proposed a method which is able to detect both strong and
weak edges and look more promising to detect edges under noisy conditions. In [2] comparative
analysis of various edge detection techniques is given. It is shown that Canny, LOG, Sobel,
Prewitt, Roberts’s exhibit better performance, respectively.
1.2. Characteristics of Edge Detector.
1. To identify less number of false edges and detection of real edges should be maximum.
2. The marked pixels should be closer to the true edge.
3. Error of detecting more than one response to single edge (double edges) should be less.
4. To design one edge detector that performs well in several contexts (Satellite images, face
recognition, medical images, natural images etc.)
This paper is organized as follows: Section (2) emphasizes on work done on edge detection and
image segmentation using soft computing approaches with images and parameters used for
evaluation. Proposed approach is presented in Section (3). Experimental setup and results are
shown in Section (4) and conclusion and future scope are discussed in Section (5).
2. RELATED WORK
Several approaches have been proposed for edge detection, a few of them are discussed here.
Konishi and et al. [12] formulate edge detection as a statistical inference. They used pre-
segmented images to learn the probability distributions of filter responses conditioned on whether
they are evaluated on or off an edge. Ground truths of images are considered and performance is
measured on Receiver Operator Characteristic (ROC) curves basis. The main disadvantage of
this method is, it uses pre-segmented images for learning on one dataset of images and then it is
applied on other dataset. J Patel and et al. [7] proposed an algorithm based on fuzzy systems
and fuzzy rules, where Sobel and Lapalacian values are computed and applied to fuzzy system.
The proposed approach reduces false edge detection and detection of multiple responses to a
single edge is less when compared to Sobel and Laplacian methods. Ground truth evaluation was
not discussed here. An algorithm to detect continuous and smooth edges using particle swarm
optimization was proposed by Mahdi Setayesh and et al. [14]. The results showed that the
algorithm performs better and less sensitive to impulsive noise than Canny. The algorithm takes
much longer time to execute when compared to Canny method. An approach for edge detection
using independent component analysis is proposed by Mendhurwar and et al. [15], the proposed
approach works well under noisy conditions when compared with Canny’s method. The
performance is compared on PSNR and no ground truth evaluations of images are considered.
The method is robust to noise and detect better edges under noisy conditions. Abdallah A.
Alshennawy and Ayman A. Aly [8] proposed a fuzzy logic technique for edge detection without
determining the threshold value. The algorithm works well and gives line smoothness and straight
for the straight lines, corners get sharper and less detection of double edges when compared to
Sobel method, Ground truth evaluation was absent. Many of these approaches discussed here
evaluate edge detection without using ground truth of images, results in perplexity for quantitative
and qualitative performance evaluation of approaches.
3. PROPOSED APPROACH
In this paper, an approach for edge detection using fuzzy set theory is proposed. In Psychological
terms, when humans view a color object, we tend to describe it by its hue, saturation and intensity
(H, S, I). Keeping in mind these terms, first RGB color image is converted into HSI image. We, as
humans perceive image primarily due to dominant wavelength of light reflected by an object i.e.
Hue and amount of light reflected by that object i.e. Intensity. Using this fact, saturation
component is removed from HSI image and hue and intensity components are added to form a
new hue and intensity (HI) image. The pixel values in the range [0 to 1] are mapped to [0 to 255]
to make computations easier to understand. The obtained (hue and intensity) HI image looks like
a gray image with pixel values from 0 to 255.](https://image.slidesharecdn.com/ijip-695-160103192850/85/A-Fuzzy-Set-Approach-for-Edge-Detection-2-320.jpg)
![Pushpajit A. Khaire & Dr. N. V. Thakur
International Journal of Image Processing (IJIP), Volume (6): Issue (6), 2012 406
3.2 Edge Detection of Obtained Fuzzy Image
A 3×1 gradient operator in horizontal and vertical direction is shown in Figure (a). These masks
are convolved over partially segmented image obtained in step 3.1. Gx, Gy are used to detect
edges in horizontal and vertical directions respectively.
Gx Gy
1
0
-1
FIGURE (a): 3×1 Edge operator
The resultant magnitude of edge pixels are calculated using equation (3.1)
2 2
G= (Gx) +(Gy)
(3.1)
These 3×1 masks requires less computations to detect edges compared to other 3×3 masks used
(Like Prewitt, Sobel). It also reduces blurring effect while detecting edges. Generally, real image
comprises of both strong and weak edges. Here, two thresholds are set for edges, higher
threshold and lower threshold. Edges above higher threshold are strong edges and edges above
lower threshold are weak edges. Higher threshold value used is 0.3 for strong edges and for
weak edges lower threshold is 0.4 × high threshold. Figure (3) shows the original, partially
segmented, ground truth and obtained edged image in (a) (b) (c) and (d) respectively.
(a) (b) (c) (d)
FIGURE 3: (a) BSD image, (b) Partially Segmented, (c) Ground Truth, (d) Proposed Approach
4. EXPERIMENTAL SETUP AND RESULTS
The approach is simulated using MATLAB 7.11 (R2010b). BSD (Berkeley Segmentation Dataset)
images [5] and respective ground truths are used for experimentation. Performance parameters
used are PSNR and PR (Ratio of true to False Edges). Results shows that the proposed
approach detect real edges as shown in ground truth and gives higher PR. Performance Ratio
(PR) is the ratio of true to false edges. It is calculated as given in equation (4.1).
True Edges (Edge pixels identified as Edges)
PR= ×100 (4.1)
False Edges (Non edge pixels identified as edges) +
(Edge pixels identified as Non-Edge pixels)
The performance and comparative results are shown in Table 4.1. The proposed approach is
compared with Canny’s algorithm using Ground Truth of respective images. Results show that
the proposed approach gives higher PSNR and PR than Canny’s approach. After number of
1 0 -1](https://image.slidesharecdn.com/ijip-695-160103192850/85/A-Fuzzy-Set-Approach-for-Edge-Detection-4-320.jpg)
![Pushpajit A. Khaire & Dr. N. V. Thakur
International Journal of Image Processing (IJIP), Volume (6): Issue (6), 2012 411
As mentioned in [4] decades of research on edge detection has produced N edge detectors
without a solid basis to evaluate the performance. Many researchers compare edge detection
algorithms without using ground truth of images, results in perplexity to evaluate and compare
these algorithms. In this paper, an attempt is made to evaluate edge detection using ground truth
for quantitative and qualitative comparison. Experimentation is carried out using BSD (Berkeley
Segmentation Database) images [5] and respective Ground Truths. The performance evaluation
parameters used are PSNR and PR (Ratio of True to false Edges). Experimental Results shows
that the proposed approach gives higher PSNR and PR values compared to Canny’s approach. It
reduces false edge detection and identification of double edges are minimum, Also the marked
pixel is closer to the true edge. Here memberships of pixels are calculated based on their
constant gray (HI) value. In future, using spatial co-ordinates, different combinations color
components of different color models, fuzzy membership of pixels can be calculated.
6. REFERENCES
1. J. Canny, “A Computational Approach to Edge Detection”, IEEE Transactions on Pattern
Analysis and Machine Intelligence. 8 (6), pp- 679-687, 1986.
2. Raman Maini and Himanshu Aggarwal, “Study and Comparison of Various Image Edge
Detection Techniques”, International Journal of Image Processing (IJIP), Volume (3), 2010,
pp-1-12.
3. Adam Hoover and et al., “An Experimental Comparison of Range Image Segmentation
Algorithms”, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol.18 no.7,
July 1996.pp- 673-689.
4. M. Heath, S. Sarkar, T. Sanocki, and K.W. Bowyer, “Comparison of Edge Detectors: A
Methodology and Initial Study”, Computer Vision and Image Understanding, vol. 69, no. 1,
Jan. 1998, pp- 38-54.
5. D. Martin, C. Fowlkes, D. Tal and J. Malik, “A database of human segmented natural
images and its application to evaluating segmentation algorithms and measuring ecological
statistics”, in: Proceedings of the IEEE International Conference on Computer Vision (ICCV
’01), 2001, pp- 416–425.
6. K. Bowyer, C. Kranenburg, and S. Dougherty, “Edge Detector Evaluation Using Empirical
ROC Curves,” Computer Vision and Image Understanding, vol. 84, no. 1, Oct. 2001, pp- 77-
103
.
7. J. Patel, J. Patwardhan, K Sankhe and R Kumbhare, “Fuzzy Inference based Edge
Detection System using Sobel and Laplacian of Gaussian Operators”, ICWET’11, ACM
978-1-4503-0449-8, February 25–26, 2011, pp- 694-697.
8. Abdallah A. Alshennawy, and Ayman A. Aly, “Edge Detection in Digital Images Using Fuzzy
Logic Technique”, World Academy of Science, Engineering and Technology, 2009, pp-178-
186.
9. Song Wang, Feng Ge, and Tiecheng Liu, “Evaluating Edge Detection through Boundary
Detection”, EURASIP Journal on Applied Signal Processing, Article ID 76278, June 2006
Pages 1–15.
10. Aborisade, D.O, “Fuzzy Logic Based Digital Image Edge Detection”, Global Journal of
Computer Science and Technology, Volume 10, November 2010, pp- 78-83.
11. Raman Maini, J.S.Sohal, “Performance Evaluation of Prewitt Edge Detector for Noisy
Images”, GVIP Journal, Volume 6, December, 2006.](https://image.slidesharecdn.com/ijip-695-160103192850/85/A-Fuzzy-Set-Approach-for-Edge-Detection-9-320.jpg)
A Fuzzy Set Approach for Edge Detection
- 1. Pushpajit A. Khaire & Dr. N. V. Thakur International Journal of Image Processing (IJIP), Volume (6): Issue (6), 2012 403 A Fuzzy Set Approach for Edge Detection Pushpajit A. Khaire [email protected] Department of Computer Technology, Karmavir Dadasaheb Kannamwar College of Engineering, Nagpur-440009, India Dr. Nileshsingh V. Thakur [email protected] Department of Computer Science & Engineering, Prof Ram Meghe College of Engineering and Management, Badnera-Amravati-444701, India Abstract Image segmentation is one of the most studied problems in image analysis, computer vision, pattern recognition etc. Edge detection is a discontinuity based approach used for image segmentation. An edge detection using fuzzy set is proposed here, where an image is considered as a fuzzy set and pixels are taken as elements of fuzzy set. The proposed approach converts the color image to a partially segmented image; finally an edge detector is convolved over the partially segmented image to obtain an edged image. The approach is implemented using MATLAB 7.11. (R2010b). In this paper, an attempt is made to evaluate edge detection using ground truth for quantitative and qualitative comparison. 30 BSD (Berkeley Segmentation Database) images and respective ground truths are used for experimentation. Performance parameters used are PSNR (dB) and Performance ratio (PR) of true to false edges. Experimental results shows that the proposed approach gives higher PSNR and PR values when compared with Canny’s edge detection algorithm under almost all scenarios. The proposed approach reduces false edge detection and identification of double edges are minimum. Keywords: Edge Detection, Fuzzy Set, BSD (Berkeley Segmentation Database), Ground Truth, PSNR. 1. INTRODUCTION Image Segmentation is an important and difficult task in low level image processing, image analysis etc. Edge detection is one of the important techniques used for image segmentation. Earlier the segmentation algorithms were divided into two groups. 1) Discontinuity based approach (Edge detection) and 2) Similarity based approach (Thresholding, Region Growing). Each of these methods has their own advantages and disadvantages. At earlier stages of research on image segmentation, edge detection (Like Prewitt, Sobel) was gaining more attention compared to region growing. Image Segmentation process simplifies, further analysis of images by reducing the amount of data to be processed significantly, at the same time useful structural information of object boundaries are preserved. There are numerous applications of image segmentation like Remote Sensing, Analysis of Medical Images, Industrial Machine Vision for Product Assembly and Inspection, Automated Target Detection and Tracking, Fingerprint Recognition, Face Recognition, Astronomical Study etc. As a result it remains an active area of research. 1.1 Edge Detection An edge is a sudden change in the pixel intensity of the image. It contains the critical characteristics and important features of an image. An edge is a boundary between the object and its background, also the process of detecting boundaries between object and background in image is known as edge detection. It facilitates, further processing of image like feature selection etc. These all put together edge detection as one of the most important task in computer vision
- 2. Pushpajit A. Khaire & Dr. N. V. Thakur International Journal of Image Processing (IJIP), Volume (6): Issue (6), 2012 404 and image processing. In recent years, researchers have applied various soft computing techniques for edge detection to improve segmentation results for various images and to enhance edge detection technique. Canny [1] proposed a method which is able to detect both strong and weak edges and look more promising to detect edges under noisy conditions. In [2] comparative analysis of various edge detection techniques is given. It is shown that Canny, LOG, Sobel, Prewitt, Roberts’s exhibit better performance, respectively. 1.2. Characteristics of Edge Detector. 1. To identify less number of false edges and detection of real edges should be maximum. 2. The marked pixels should be closer to the true edge. 3. Error of detecting more than one response to single edge (double edges) should be less. 4. To design one edge detector that performs well in several contexts (Satellite images, face recognition, medical images, natural images etc.) This paper is organized as follows: Section (2) emphasizes on work done on edge detection and image segmentation using soft computing approaches with images and parameters used for evaluation. Proposed approach is presented in Section (3). Experimental setup and results are shown in Section (4) and conclusion and future scope are discussed in Section (5). 2. RELATED WORK Several approaches have been proposed for edge detection, a few of them are discussed here. Konishi and et al. [12] formulate edge detection as a statistical inference. They used pre- segmented images to learn the probability distributions of filter responses conditioned on whether they are evaluated on or off an edge. Ground truths of images are considered and performance is measured on Receiver Operator Characteristic (ROC) curves basis. The main disadvantage of this method is, it uses pre-segmented images for learning on one dataset of images and then it is applied on other dataset. J Patel and et al. [7] proposed an algorithm based on fuzzy systems and fuzzy rules, where Sobel and Lapalacian values are computed and applied to fuzzy system. The proposed approach reduces false edge detection and detection of multiple responses to a single edge is less when compared to Sobel and Laplacian methods. Ground truth evaluation was not discussed here. An algorithm to detect continuous and smooth edges using particle swarm optimization was proposed by Mahdi Setayesh and et al. [14]. The results showed that the algorithm performs better and less sensitive to impulsive noise than Canny. The algorithm takes much longer time to execute when compared to Canny method. An approach for edge detection using independent component analysis is proposed by Mendhurwar and et al. [15], the proposed approach works well under noisy conditions when compared with Canny’s method. The performance is compared on PSNR and no ground truth evaluations of images are considered. The method is robust to noise and detect better edges under noisy conditions. Abdallah A. Alshennawy and Ayman A. Aly [8] proposed a fuzzy logic technique for edge detection without determining the threshold value. The algorithm works well and gives line smoothness and straight for the straight lines, corners get sharper and less detection of double edges when compared to Sobel method, Ground truth evaluation was absent. Many of these approaches discussed here evaluate edge detection without using ground truth of images, results in perplexity for quantitative and qualitative performance evaluation of approaches. 3. PROPOSED APPROACH In this paper, an approach for edge detection using fuzzy set theory is proposed. In Psychological terms, when humans view a color object, we tend to describe it by its hue, saturation and intensity (H, S, I). Keeping in mind these terms, first RGB color image is converted into HSI image. We, as humans perceive image primarily due to dominant wavelength of light reflected by an object i.e. Hue and amount of light reflected by that object i.e. Intensity. Using this fact, saturation component is removed from HSI image and hue and intensity components are added to form a new hue and intensity (HI) image. The pixel values in the range [0 to 1] are mapped to [0 to 255] to make computations easier to understand. The obtained (hue and intensity) HI image looks like a gray image with pixel values from 0 to 255.
- 3. Pushpajit A. Khaire & Dr. N. V. Thakur International Journal of Image Processing (IJIP), Volume (6): Issue (6), 2012 405 3.1 Fuzzy Membership of Pixels This HI image is considered as a fuzzy set and the pixels are taken as elements of a fuzzy set. Fuzzy membership of pixel elements is defined based on their constant gray (HI) value. Maximum number of pixels having a constant gray value has the highest degree of membership i.e. 1. Similarly, second maximum set of pixels having constant gray value (pixel value) has the next membership i.e. less than 1. Each pixel in an image holds their membership value depending upon number of pixels having same pixel (gray) value. Now a pixel in this Fuzzy image (Set) has three features: 1. Spatial co-ordinates i.e. (x, y) co-ordinates. 2. Pixel Value (gray value). 3. Fuzzy membership (membership value). The fuzzy Set F of image is defined as follows: F= {(x, µF(x), x ∈ X}, where µF(x) denotes the membership value of (pixel) element x in (Image) Fuzzy Set F. The next step is to employ fuzzy rule on all set of pixels, which results in a partially segmented image. Let A be the set of pixels in fuzzy set F with constant gray value g1 and membership value m1. Similarly, let B be the set of pixels in the same fuzzy set F with constant gray value g2 and membership value m2. Let C (g3, m3) be the union of the two sets A and B holds true if it satisfies following conditions: 1) If difference between membership values of A and B is less than or equal to 0.2 (|m1-m2| <=0.2). 2) Difference between gray values of A and B is less than or equal to 32 (|g1-g2|<=32). If the pixel sets satisfies above two conditions then a new set C(m3,g3) is created using set A and B i.e. C=(AUB), where m3=max(m1,m2) and g3=respective gray value of max(m1,m2).The two pixel sets A and B are replaced by pixel set C in image. This procedure is repeated for all set of pixels, results in partially segmented image. Histograms of HI image and partially segmented image are shown in Figure (1) and Figure (2) respectively. FIGURE 1: Histogram of HI image FIGURE 2: Histogram of Partially Segmented Image.
- 4. Pushpajit A. Khaire & Dr. N. V. Thakur International Journal of Image Processing (IJIP), Volume (6): Issue (6), 2012 406 3.2 Edge Detection of Obtained Fuzzy Image A 3×1 gradient operator in horizontal and vertical direction is shown in Figure (a). These masks are convolved over partially segmented image obtained in step 3.1. Gx, Gy are used to detect edges in horizontal and vertical directions respectively. Gx Gy 1 0 -1 FIGURE (a): 3×1 Edge operator The resultant magnitude of edge pixels are calculated using equation (3.1) 2 2 G= (Gx) +(Gy) (3.1) These 3×1 masks requires less computations to detect edges compared to other 3×3 masks used (Like Prewitt, Sobel). It also reduces blurring effect while detecting edges. Generally, real image comprises of both strong and weak edges. Here, two thresholds are set for edges, higher threshold and lower threshold. Edges above higher threshold are strong edges and edges above lower threshold are weak edges. Higher threshold value used is 0.3 for strong edges and for weak edges lower threshold is 0.4 × high threshold. Figure (3) shows the original, partially segmented, ground truth and obtained edged image in (a) (b) (c) and (d) respectively. (a) (b) (c) (d) FIGURE 3: (a) BSD image, (b) Partially Segmented, (c) Ground Truth, (d) Proposed Approach 4. EXPERIMENTAL SETUP AND RESULTS The approach is simulated using MATLAB 7.11 (R2010b). BSD (Berkeley Segmentation Dataset) images [5] and respective ground truths are used for experimentation. Performance parameters used are PSNR and PR (Ratio of true to False Edges). Results shows that the proposed approach detect real edges as shown in ground truth and gives higher PR. Performance Ratio (PR) is the ratio of true to false edges. It is calculated as given in equation (4.1). True Edges (Edge pixels identified as Edges) PR= ×100 (4.1) False Edges (Non edge pixels identified as edges) + (Edge pixels identified as Non-Edge pixels) The performance and comparative results are shown in Table 4.1. The proposed approach is compared with Canny’s algorithm using Ground Truth of respective images. Results show that the proposed approach gives higher PSNR and PR than Canny’s approach. After number of 1 0 -1
- 5. Pushpajit A. Khaire & Dr. N. V. Thakur International Journal of Image Processing (IJIP), Volume (6): Issue (6), 2012 407 experiments it is found that the default sigma value available in Matlab 7.11 i.e. 1 and threshold=0.3 for Canny approach offer better result than other sigma and threshold values. Here threshold value =0.3 and default sigma=1 for Canny is used for comparison. In proposed approach higher threshold value used is 0.3 for strong edges and for weak edges lower threshold is 0.4 × high threshold. As thickness of edge determines whether an edge is strong or weak edge, to distinguish between strong and weak edges thinning operation is not performed on the resultant edged image. Resultant edged image and respective ground truth of images are shown in Figure (4) through Figure (6). BSD Image Proposed (T=0.3) Canny (T=0.3,σ=1) No. PSNR(dB) PR PSNR(dB) PR 135069 23.9931 28.0419 23.9735 14.1513 176039 23.487 10.113 23.4721 6.9261 15088 23.3953 17.7663 23.3726 8.8939 12074 23.246 10.8485 23.2335 7.2393 210088 23.0247 13.2036 23.0153 8.7167 28075 22.981 8.1661 22.9767 8.2033 108073 21.5224 10.3232 21.5102 5.8069 3096 21.4337 9.5958 21.4189 5.5835 134052 21.2345 11.0943 21.2264 6.4736 189080 20.9949 12.4702 20.9795 8.993 189011 20.6976 10.5861 20.6882 6.3055 253036 20.4997 21.8664 20.4877 12.6724 8068 20.1987 4.951 20.1962 4.9881 310007 20.1268 15.2524 20.122 11.3261 3063 20.0554 4.3045 20.0496 4.0618 118035 19.9305 17.5689 19.9168 10.3822 41004 19.8741 15.4356 19.8647 9.2601 23025 19.7884 11.2603 19.7855 10.5085 176035 19.758 11.4857 19.7472 6.8046 113044 19.442 16.7319 19.4237 11.3814 197017 19.2163 14.7429 19.2133 11.1349 181018 19.121 10.4562 19.1163 7.4872 35070 18.9703 23.8817 18.9636 16.6493 163014 18.7961 16.2093 18.7877 10.9464 101087 18.423 11.1953 18.417 9.962 157055 18.174 11.6058 18.1672 9.8996 242078 18.1425 16.5446 18.1324 11.006 42049 17.9908 27.2644 17.9737 14.8333 245051 17.3586 26.2115 17.3408 13.9813 35010 17.2309 21.85 17.214 12.5393 TABLE 4.1: Comparison of Approaches.
- 6. Pushpajit A. Khaire & Dr. N. V. Thakur International Journal of Image Processing (IJIP), Volume (6): Issue (6), 2012 408 135069 176039 15088 12074 210088 28075 108073 3096 134052 189080 189011 (a) (b) (c) (d) (e) FIGURE 4: Column (a) Image No., Column (b) BSD image, Column (c) Ground Truth, Column (d) Canny’s approach, Column (e) Proposed approach
- 7. Pushpajit A. Khaire & Dr. N. V. Thakur International Journal of Image Processing (IJIP), Volume (6): Issue (6), 2012 409 253036 8068 310007 3063 118035 41004 23025 176035 113044 197017 181018 (a) (b) (c) (d) (e) FIGURE 5: Column (a) Image No., Column (b) BSD image,Column (c) Ground Truth, Column (d) Canny’s approach, Column (e) Proposed approach
- 8. Pushpajit A. Khaire & Dr. N. V. Thakur International Journal of Image Processing (IJIP), Volume (6): Issue (6), 2012 410 35070 163014 101087 157055 242078 42049 245051 35010 (a) (b) (c) (d) (e) FIGURE 6: Column (a) Image No., Column (b) BSD image,Column (c) Ground Truth, Column (d) Canny’s approach, Column (e) Proposed approach 5. CONCLUSION AND FUTURE SCOPE Edge detection is one of the important techniques used for image segmentation. Image segmentation remains a puzzled problem even after four decades of research. In this paper, a soft computing approach based on Fuzzy Set is proposed for edge detection, where an image is considered as a Fuzzy Set and pixels are taken as elements of Fuzzy Set. The fuzzy approach converts the color image to a partially segmented image, finally an edge detector is convolved over the partially segmented image to obtain edged image. As, proposed edge operator does not perform blurring on image, double edges are less identified. Generally real images comprises of both strong and weak edges. The proposed approach gives both strong and weak edges having different edge strength using higher and lower thresholds.
- 9. Pushpajit A. Khaire & Dr. N. V. Thakur International Journal of Image Processing (IJIP), Volume (6): Issue (6), 2012 411 As mentioned in [4] decades of research on edge detection has produced N edge detectors without a solid basis to evaluate the performance. Many researchers compare edge detection algorithms without using ground truth of images, results in perplexity to evaluate and compare these algorithms. In this paper, an attempt is made to evaluate edge detection using ground truth for quantitative and qualitative comparison. Experimentation is carried out using BSD (Berkeley Segmentation Database) images [5] and respective Ground Truths. The performance evaluation parameters used are PSNR and PR (Ratio of True to false Edges). Experimental Results shows that the proposed approach gives higher PSNR and PR values compared to Canny’s approach. It reduces false edge detection and identification of double edges are minimum, Also the marked pixel is closer to the true edge. Here memberships of pixels are calculated based on their constant gray (HI) value. In future, using spatial co-ordinates, different combinations color components of different color models, fuzzy membership of pixels can be calculated. 6. REFERENCES 1. J. Canny, “A Computational Approach to Edge Detection”, IEEE Transactions on Pattern Analysis and Machine Intelligence. 8 (6), pp- 679-687, 1986. 2. Raman Maini and Himanshu Aggarwal, “Study and Comparison of Various Image Edge Detection Techniques”, International Journal of Image Processing (IJIP), Volume (3), 2010, pp-1-12. 3. Adam Hoover and et al., “An Experimental Comparison of Range Image Segmentation Algorithms”, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol.18 no.7, July 1996.pp- 673-689. 4. M. Heath, S. Sarkar, T. Sanocki, and K.W. Bowyer, “Comparison of Edge Detectors: A Methodology and Initial Study”, Computer Vision and Image Understanding, vol. 69, no. 1, Jan. 1998, pp- 38-54. 5. D. Martin, C. Fowlkes, D. Tal and J. Malik, “A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics”, in: Proceedings of the IEEE International Conference on Computer Vision (ICCV ’01), 2001, pp- 416–425. 6. K. Bowyer, C. Kranenburg, and S. Dougherty, “Edge Detector Evaluation Using Empirical ROC Curves,” Computer Vision and Image Understanding, vol. 84, no. 1, Oct. 2001, pp- 77- 103 . 7. J. Patel, J. Patwardhan, K Sankhe and R Kumbhare, “Fuzzy Inference based Edge Detection System using Sobel and Laplacian of Gaussian Operators”, ICWET’11, ACM 978-1-4503-0449-8, February 25–26, 2011, pp- 694-697. 8. Abdallah A. Alshennawy, and Ayman A. Aly, “Edge Detection in Digital Images Using Fuzzy Logic Technique”, World Academy of Science, Engineering and Technology, 2009, pp-178- 186. 9. Song Wang, Feng Ge, and Tiecheng Liu, “Evaluating Edge Detection through Boundary Detection”, EURASIP Journal on Applied Signal Processing, Article ID 76278, June 2006 Pages 1–15. 10. Aborisade, D.O, “Fuzzy Logic Based Digital Image Edge Detection”, Global Journal of Computer Science and Technology, Volume 10, November 2010, pp- 78-83. 11. Raman Maini, J.S.Sohal, “Performance Evaluation of Prewitt Edge Detector for Noisy Images”, GVIP Journal, Volume 6, December, 2006.
- 10. Pushpajit A. Khaire & Dr. N. V. Thakur International Journal of Image Processing (IJIP), Volume (6): Issue (6), 2012 412 12. S. Konishi, A. Yuille, J. Coughlan and S.C. Zhu, “Statistical Edge Detection: Learning and Evaluating Edge Cues”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Jan, 2003. 13. Bhoyar and Kakde, “Color Image Segmentation using Fast Fuzzy C-Means Algorithm”, Electronic letters on (CVIA) 2010, pp-18-31. 14. Mahdi Setayesh, Mengjie Zhang, and Mark Johnston, “Detection of Continuous, Smooth and Thin Edges in Noisy Images Using Constrained Particle Swarm Optimization”, ACM 978-1-4503-0557-0, GECCO’11, Dublin, Ireland, July 12–16, 2011, pp- 45-52. 15. Kaustubha Mendhurwar, Shivaji Patil, Harsh Sundani, Priyanka Aggarwal, and Vijay Devabhaktuni, “Edge-Detection in Noisy Images Using Independent Component Analysis”, ISRN Signal Processing, 9 pages, February 2011. 16. N. Pal and S. Pal, “A Review on Image Segmentation Techniques”, Pattern Recognition, Vol. 26, no. 9, 1993, pp- 1,277-1, 294. 17. Evans, A. N. and Lin, X. U., “A Morphological Gradient Approach to Color Edge Detection”, IEEE Transactions on Image Processing, 15 (6), pp. 1454-1463, 2006. 18. Chandra Sekhar Panda and Srikanta Patnaik, “Filtering Corrupted Image and Edge Detection in Restored Grayscale Image Using Derivative Filters”, International Journal of Image Processing (IJIP), Volume 3, 2010, pp- 105-119. 19. O. J. Tobias and R. Seara, “Image segmentation by histogram thresholding using fuzzy sets”, IEEE Transaction on Image Processing., Vol. 11, 2002, pp-1457-1465. 20. R. Unnikrishnan, C. Pantofaru, and M. Hebert, “Towards objective evaluation of image segmentation algorithms”, IEEE Transactions on Pattern Analysis and Machine Intelligence 29 (6), (2007), pp-929–944. 21. R. C. Gonzalez and R. E. Woods. “Digital Image Processing”, Addison Wesley, 2nd edition, 1992. 22. George J.Klir and Bo Yuan, “Fuzzy sets and Fuzzy logic: Theory and Applications”, Prentice Hall, 1995. 23. Francisco J. Estrada and Allan D. Jepson, “Benchmarking Image Segmentation Algorithms”, International Journal of Computer Vision. Vol. 85, no. 2, Nov 2009, pp. 167- 181. 24. Wenshuo Gao and et al., “An Improved Sobel Edge Detection”, 978-1-4244-5540-9 IEEE, ICICT 2010.