Combining contour-based and region-based in image segmentation

Issam Dagher; Elie Abboud

doi:10.12688/f1000research.140872.1

Home Browse Combining contour-based and region-based in image segmentation

ALL Metrics

Views

Downloads

Get PDF

Get XML

Export

▬

✚

Research Article

Combining contour-based and region-based in image segmentation

[version 1; peer review: 1 approved with reservations]

Issam Dagher ¹, Elie Abboud¹

PUBLISHED 11 Oct 2023

Author details Author details

¹ Computer Engineering Department, University of Balamand, Balamand, North Governorate, Lebanon

Issam Dagher
Roles: Methodology

Elie Abboud
Roles: Software

OPEN PEER REVIEW

REVIEWER STATUS

Abstract

Background: This paper presents an optimized clustering approach applied to image segmentation. Accurate image segmentation impacts many fields like medical, machine vision, object detection. Applications involve tumor detection, face detection and recognition, and video surveillance.
Methods: The developed approach is based on obtaining an optimum number of clusters and regions of an image. We combined Region-based and contour-based approaches. Initial rough regions are obtained using edge detection. We have used Gabor wavelets for texture classification and spatial resolutions. Color frequencies are also used to determine the number of clusters of the Fuzzy c-means (FCM) algorithm which gave an optimum number of clusters or regions.
Results: We have compared our approach with other similar wavelet and clustering techniques. Our algorithm gave better values for segmentation metrics like SNR, PSNR, and MCC.
Conclusions: Optimizing the number of clusters or regions has a significant effect on the performance of the image segmentation techniques. This will result in better detection and localization of the segmentation-based application.

Keywords

Image Segmentation. Clustering. Edge detection. Colour frequencies. Texture.

Corresponding author: Issam Dagher

Competing interests: No competing interests were disclosed.

Grant information: The author(s) declared that no grants were involved in supporting this work.

Copyright: © 2023 Dagher I and Abboud E. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

How to cite: Dagher I and Abboud E. Combining contour-based and region-based in image segmentation [version 1; peer review: 1 approved with reservations]. F1000Research 2023, 12:1312 (https://doi.org/10.12688/f1000research.140872.1) First published: 11 Oct 2023, 12:1312 (https://doi.org/10.12688/f1000research.140872.1) Latest published: 02 Apr 2024, 12:1312 (https://doi.org/10.12688/f1000research.140872.3)

Introduction

Image segmentation is a fundamental step in computer vision for object recognition and classification. Despite many techniques and algorithms have been proposed, image segmentation remains one of the most challenging research topics because none of them can provide a coherent framework for achieving quick and efficient segmentation of images.¹ Two explanations can be attributed to the complexity of image segmentation. The first is that image segmentation has many solutions for the problem i.e. for one image, there are many best results of segmentation. The second is because of noise, background, low signal-to-noise-ratio, and uninformed intensity.² For that, it is difficult to only suggest one image segmentation method. We can distinguish between two concepts in image segmentation: region-based and contour-based techniques.

Region-based approaches partition the image into different homogenous regions based on similarities in color, location, and texture.

Contour-based techniques start with edge detection technique followed by linking and forming the segments.

In this paper, we tried to combine both approaches. We start with the Canny edge detector. Then we form initial regions accordingly. Those regions are optimized and merged according to similarities in color, location, and texture.

Over recent years, several techniques have been developed to segment images. Wavelet-based segmentation can be found in Ref. 3. Unsupervised image segmentation⁴ is performed using k-means clustering. It clusters (segments) the image into different homogenous regions. In Ref. 5 Graph theory was employed using greedy decisions. Segmentation using Texture is shown in Sagiv et al.⁶ Shi et al.⁷ used smoothness and boundary continuity. Ren and Malik⁸ used contours and textures. In Refs. 9 and 10 the concept of superpixels was used where the redundancy of the image can be highly decreased Superpixel methods¹¹^,¹² have been researched intensively using NCut, mean shift, and graph-based methods. Genetic algorithm was also employed in Ref. 13. Edge detection techniques in image segmentation is shown in Ref. 14.

Image segmentation techniques

Image segmentation is the process of dividing an image into multiple partitions. It is typically used to locate objects and change the representation of the image into something more meaningful. It is also used in multiple domains such as medical imaging, object detection, face recognition, and machine vision.

Image segmentation consists of assigning a label for every pixel in an image. Moreover, different labels have different characteristics, and the same labels share the same characteristics at some point such as color, intensity, or texture. The result of image segmentation is a set of segments that collectively cover the entire image or a set of contours extracted from the image.

Different image segmentation techniques exist like threshold-based, region growth, edge detection, and clustering methods.¹

Threshold-based segmentation

Threshold segmentation¹⁵ is one of the most common segmentation techniques. It splits the picture into two or multiple regions using one or multiple thresholds. The most commonly used threshold segmentation algorithm is the Otsu method, which selects optimum threshold by optimizing deviation between groups. Its downside is that it is difficult to get correct results where there is no noticeable grayscale variation or overlap between the grayscale values in the image.² Since Thresholding recognizes only the gray information of the image without taking into consideration the spatial information of the image, it is vulnerable to noise and grayscale unevenness, for that it is frequently combined with other methods.

Region growth segmentation

The regional growth approach¹⁶ is a traditional serial segmentation algorithm, and its basic concept is to use identical pixel properties together to construct a region. An arbitrary seed pixel is chosen and compared with neighboring pixels. The region is grown from the seed pixel by adding neighboring pixels that are similar, increasing the size of the region. When the expansion of one region stops, another seed pixel that doesn’t yet belong to any region is chosen and therefore the flow is repeated.

Edge detection

Edge detection¹⁷ is used to find the boundaries of objects in an image. It detects discontinuities in brightness. The most common edge detection technique is Canny edge detector which can be described by the 5 following steps.

1. Gaussian filter is used to smooth the image.
2. Get the gradient magnitude and the gradient angle of the image.
3. Non-maximum suppression is applied.
4. Double thresholding is applied.
5. Suppress weak edges using hysteresis.

Finally, the image is segmented, and edges are drowned at the boundaries of each object.

Clustering

Clustering¹⁸ is the task of dividing the population or data points into several groups such that similar data points within the same groups are dissimilar to the data points in other groups. A common clustering algorithm is the Fuzzy C-means (FCM).

Fuzzy c-means (FCM) is a clustering method that permits one piece of data to be a member of two or more clusters. Based on the distance between the cluster center and the data point, this algorithm determines each data point’s membership in relation to each cluster center. The FCM algorithm can be described by the following steps:

1. Randomly select ‘c’ cluster centers.
2. Calculate the fuzzy membership ‘μ_ij’ using:
$μij = 1 / \sum_{k = 1}^{c} {(\frac{dij}{dik})}^{\frac{2}{m} - 1}$
3. Compute the fuzzy centers ‘v_j’ using:
$vj = \sum_{i = 1}^{n} ((μij)^m * xi) / \sum_{k = 1}^{n} ((μij)^m), \forall j = 1, 2, \dots c$
4. Repeat step 2 and 3 until ||U^(k+1) - U^(k)||< ԑ.

Connected component algorithm

The Connected component algorithm¹⁹ scans an image and groups the pixels into components dependent on pixel connectivity, i.e. all pixels in the connected component share identical pixel intensity values and are in some way connected. Until all classes have been determined, each pixel shall be labelled with a gray level or a color (color marking) according to the portion to which it has been allocated. Connected part labeling works by scanning an image, pixel-by-pixel (from top to bottom and from left to right) to identify connected pixel regions, i.e. neighboring pixel regions that share the same collection of intensity values as V. The following is the labeling for p:

• If all four neighbors are zero, give p a new label; otherwise
• If only one neighbor has V= 1, give its label to p; otherwise
• If more than one neighbor has V= 1, give one of the labels to p and note the equivalences.

After screening, the identical label pairs are sorted into equivalence groups and a unique label is assigned to each class. As a final stage, a second scan is performed through the image, during which each label is replaced by the label assigned to its equivalence class.

Texture filters: Gabor wavelets

The objective of Texture filters²⁰ is to separate the regions in an image based on their texture content. While smooth regions are characterized with a small range of values in the neighborhood around a pixel, rough texture regions are characterized by a large range of values. Gabor Wavelets are band pass filters which extract the image local important features. A convolution is done between the image and the filters in order to get texture frequency and orientation. We have used the outputs of Gabor filters with 8 orientations and 5 wavelengths.

Methods

The proposed approach is based on obtaining an optimum number of clusters and regions of an image obtained from the Berkeley segmentation dataset. This is done using the following three consecutive steps:

I. Obtaining a good initial set of centers:
- • Apply edge detection. This is done using the canny edge detector.
- • Apply the connected component algorithm on the binary image obtained.
- • Using the labeled image, find the properties of each region.
- • Join similar regions and keep the unique ones.
- • Finally, find the center of each region.
  Figure 1 illustrates the procedures of step I.
II. Reducing the number of centers
This is done using texture filters as follows:
- • Get the feature vectors of each center using Gabor filters.
- • Merge the centers according to their Euclidian distances and the results obtained from the Gabor filters using:
  The Euclidian distance between 2 centers is given by:
  $Distance = \sqrt{{(Xcenter 1 - Xcenter 2)}^{2} + {(Ycenter 1 - Ycenter 2)}^{2}}$
  Where Xcenter 1 and Ycenter 1 are the xy coordinates of the first center and Xcenter 2 and Ycenter 2 are the xy coordinates of the second center
  The features distance between 2 centers is given by:
  $Feature Distance = \sqrt{{(Feature center 1 - Feature center 2)}^{2}}$
- • If the 2 centers are close to each other and approximately belong to the same texture, then merge them.
  $Distance×Feature Distance < Threshold$
  The results are shown in Figure 2.
  Figure 2 shows that the number of centers was reduced from 246 to 97.
III. Apply the FCM clustering algorithm:
It should be noted that the FCM clustering requires the specification of the number of clusters. Noting that in color image segmentation the similarity used by the FCM is based on Euclidian distance between RGB pixels, getting the number of clusters is done by using Color frequencies. The color frequencies²¹ index is computed by three steps:
- 1. All the color frequencies of the image are computed and added to an array
- 2. Then, the duplications in the array are removed and unique frequencies are kept
- 3. Finally, only the main colors are kept for example if there are multiple shades of a color only the main color is kept, and the other ones are removed
The color frequencies index is equal to the size of the array and is given as an input to the FCM function. After this step is applied the number of RGB centroids is reduced from 97 centroids to only 13 (Figure 3). Then the RGB distance is computed between each pixel and the center to determine its corresponding label.
Our algorithm is summarized in Figure 4.

Figure 1. (a) Original image; (b) Grayscale image; (c) Edge image; (d) Initial set of centers.

Figure 2. (a) Initial set of centers; (b) After the first reduction.

Figure 3. Reduction of RGB centers using FCM.

Figure 4. Flowchart of the proposed approach.

Illustration of the algorithm

Figure 5 shows 3 images and their edge images. Figure 6 shows the edge images and their corresponding initial set of centers. The optimum number of cluster centers is shown in Figure 7. The final image segmented images are shown in Figure 8

Figure 5. Original 3 images (Upper Row) and their Corresponding Edge images (Lower Row).

Figure 6. Edge images (Upper Row) and their corresponding initial centers (Lower Row).

Figure 7. Initial set of centers (Upper row) and their corresponding optimum number of centers (Lower row).

Figure 8. Segmented images (lower row) and their corresponding optimum number of centers (upper row).

Results

Dataset

To evaluate this work, the BSDS500 database²² is chosen. It is used for most segmentation techniques. It consists of 500 images of outdoor scenes, landscapes, buildings, animals, and humans. Figure 9 shows sample images from the database.

Figure 9. Samples from the BSD500 database.

Segmentation metrics

The following segmentation metrics²³ are used to show the effectiveness of our novel approach: accuracy, F-measure, precision, MCC, dice, Jaccard, specificity. Those metrics are computed by comparing the result segmented image with the ground truth of the original image.

Given that: TP is the true positive, TN is the true negative, FN is the false negative and FP is the false positive

Accuracy = \frac{TP + TN}{TP + FN + FP + TN} . Precision = \frac{TP}{TP + FP} . F measure = \frac{(β^{2} + 1) TP}{(β^{2} + 1) TP + β^{2} FN + FP} .

MCC = \frac{TP * TN - FP * FN}{\sqrt{(TP + FP) (TP + FN) (TN + FN) (TN + FP)}} . J accard = \frac{TP}{TP + FN + FP} . Dice = \frac{2 TP}{2 TP + FN + FP}

Specivity = \frac{TN}{TN + FN} .

Results of proposed approach

In this section, the results of the proposed approach are compared with different methods on the same database and using the same classification metrics. For the K-means and the SLIC we have experimented with different values of K and we have chosen the value of K which gave good segmentation results. We used K=10 for the K-means and K=100 For the SLIC.

Graphical Illustration

The following figures illustrate the segmentation results of the Kmeans, SLIC, and our algorithm. Figure 10 shows the results obtained by the K-means, the SLIC, and our algorithm. The Figure shows the superior performance of our approach.

Figure 10. Original 3 images (First column).

Results of the K-means, the SLIC, and the proposed approach in second, third, and fourth columns respectively.

Comparisons based on the Segmentation metrics

Table 1 shows the segmentation metrics results of our algorithm compared to the K-means, the SLIC and the CAS²⁴ algorithms. The images of the BSD500 are used and the average segmentation metrics are shown in the table. Table 4.1 shows the accurate segmentation results of our algorithm compared to the others. It should be noted that our algorithm does not require a priori to specify the number of centers.

Table 1. Performance of the various approaches on the BSD500.

Method	Avg. Accuracy	Avg. Precision	Avg. F-measure	Avg. MCC	Avg. Jaccard	Avg. Dice	Avg. Specificity
K-means K=10	0.8842	0.8941	0.8886	0.8672	0.8696	0.8886	0. 8977
SLIC K=100	0.9296	0.9316	0.9284	0.8918	0.9111	0.9284	0.9267
CAS K=400	0.9711	0.9668	0.9454	0.9221	0.9654	0.9756	0.9775
Proposed approach	0.9841	0.9776	0.9543	0.9231	0.9732	0.9811	0.9844

To show the effectiveness of the proposed method, we have followed the experiments done in Ref. 3 using 2 images: Lena and the Cameraman images (Figure 11). We have used the SNR and the PSNR as verification indices. Table 2 shows the results obtained. It clearly shows the outperformance of our approach.

Figure 11. Cameraman and Lena images.

Table 2. Performance of our approach for the 2 images compared to the results obtained in Ref. 3.

Methods	SNR	PSNR
Lena
MVSM	46.0638	3.5854
BSM	45.6782	3.1999
VTSM	46.1026	3.6242
WSM	48.1855	5.7071
Our	50.89	7.78
Cameraman
MVSM	45.6261	3.0267
BSM	47.4184	4.819
VTSM	45.6929	3.0935
WSM	48.1859	5.5865
Our	50.76	7.88

Bigger SNR and PSNR imply better segmentation results. Our algorithm gave for the Lena image an SNR 0f 50.89 and PSNR of 7.78 which are bigger than the other 4 algorithms.

Conclusion

Image segmentation has become an important topic in many fields like medical, machine vision, object detection. In this work, a new approach is proposed to improve the accuracy and performance of image segmentation. We combined Region-based and Contour-based segmentation both approaches. Edge detection, Color frequencies, and texture measures are used in developing the new algorithm. We started with Canny edge detector. Then we formed initial regions accordingly. Those regions are optimized and merged according to similarities in color, location and texture. We obtained optimum number of clusters and regions of an image. To show the effectiveness of this work, the BSDS500 database is chosen and different segmentation and clustering measures were used. The results show the improved performance of the proposed technique compared to other wavelet-based and other techniques.

Data availability

Source data

All images used in this article were sourced from The Berkeley Segmentation Dataset and Benchmark (BSDS300): https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/resources.html#algorithms ²²

Extended data

Zenodo. COMBINING CONTOUR-BASED AND REGION-BASED IN IMAGE SEGMENTATION. https://doi.org/10.5281/zenodo.8319898.²⁵

This project contains the following extended data:

• Code.docx (analysis code)

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).

Acknowledgements

Acknowledgements go to the FOE research lab.

References

1. Saxena S, Jain S, Tripathi S, et al.: Comparative Analysis of Image Segmentation Techniques.Hura GS, Singh AK, Siong Hoe L, editors. Advances in Communication and Computational Technology. ICACCT 2019. Lecture Notes in Electrical Engineering. Vol. 668. . Singapore: Springer; 2021.
2. Fan L, Zhang F, Fan H, et al.: Brief review of image denoising techniques. Vis. Comput. Ind. Biomed. Art. 2019; 2: 7. PubMed Abstract | Publisher Full Text | Free Full Text
3. Gao J, Wang B, Wang Z, et al.: A wavelet transform-based image segmentation method. Optik. 2020; 208: 164123. Publisher Full Text
4. Yang AY, Wright J, Ma Y, et al.: Unsupervised segmentation of natural images via lossy data compression. Comput. Vis. Image Underst. 2008; 110(2): 212–225. Publisher Full Text
5. Felzenszwalb PF, Huttenlocher DP: Efficient graph-based image segmentation. Int. J. Comput. Vis. 2004; 59(2): 167–181. [7] X. Liu, Q. Xu, J. Ma, H. J. Publisher Full Text
6. Sagiv C, Sochen NA, Zeevi YY: Integrated active contours for texture segmentation. IEEE Trans. Image Process. 2006; 15(6): 1633–1646. PubMed Abstract | Publisher Full Text
7. Shi J, Belongie S, Leung T, et al.: Image and video segmentation: the normalized cut framework. Proceedings 1998 International Conference on Image Processing. ICIP98. 1998; vol. 1: pp. 943–947.
8. Malik J, Belongie S, Leung T, et al.: Contour and texture analysis for image segmentation. Int. J. Comput. Vis. 2001; 43(1): 7–27. Publisher Full Text
9. Wang C, Liu Z, Chan S: Superpixel-based hand gesture recognition with Kinect depth camera. IEEE Trans. Multimed. 2015; 17(1): 29–39. Publisher Full Text
10. Yang F, Lu H, Yang M: Robust superpixel tracking. IEEE Trans. Image Process. 2014; 23(4): 1639–1651. PubMed Abstract | Publisher Full Text
11. Gong Y, Zhou Y: Differential evolutionary superpixel segmentation. IEEE Trans. Image Process. 2018; 27(3): 1390–1404. PubMed Abstract | Publisher Full Text
12. Achanta R, Shaji A, Smith K, et al.: SLIC superpixels compared to state-of-the-art superpixel methods. IEEE Trans. Pattern Anal. Mach. Intell. 2012; 34(11): 2274–2282. PubMed Abstract | Publisher Full Text
13. Angelina S, Suresh LP, Veni SHK: Image segmentation based on genetic algorithm for region growth and region merging. 2012 International Conference on Computing, Electronics and Electrical Technologies (ICCEET), Kumaracoil, 2012. 2012; pp. 970–974. Publisher Full Text
14. Senthilkumaran N, Rajesh R: Edge detection techniques for image segmentation - A survey of soft computing approaches. Int. J. Recent Trends Eng. 2009; 1(2): 250.
15. Wang W, Duan L, Wang Y: Fast Image Segmentation Using Two-Dimensional Otsu Based on Estimation of Distribution Algorithm. J. Electr. Comput. Eng. 2017; 2017: 1–12. Article ID 1735176. Publisher Full Text
16. Zanaty EA, El-Zoghdy SF: A novel approach for color image segmentation based on region growing. Int. J. Comput. Appl. 2017; 39(3): 123–139. 12 pages, 2017. Publisher Full Text
17. Canny J: A Computational Approach to Edge Detection. IEEE Trans. Pattern Anal. Mach. Intell. Nov. 1986; PAMI-8(6): 679–698. Publisher Full Text
18. Huang H, Meng F, Zhou S, et al.: Brain Image Segmentation Based on FCM Clustering Algorithm and Rough Set. IEEE Access. 2019; 7: 12386–12396. Publisher Full Text
19. Abderaouf Z, Nadjia B, Saliha O: License plate character segmentation based on horizontal projection and connected component analysis. 2014 World Symposium on Computer Applications & Research (WSCAR), Sousse. 2014; pp. 1–5. Publisher Full Text
20. Park B-Y, Kim H-H, Hong B-W: A Multilabel Texture Segmentation Based on Local Entropy Signature. Math. Probl. Eng. 2013; 6. 2013. ArticleID 651581.
21. Garcia-Lamont F, Cervantes J, López A, et al.: Segmentation of images by color features: A survey. Neurocomputing. 2018; 292: 1–27. Publisher Full Text
22. Martin D, Fowlkes C, Tal D, et al.: A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics. Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001, Vancouver, BC, Canada. 2001; vol. 2: pp. 416–423. Publisher Full Text
23. Balafar MA: segmentation evaluatation, MATLAB Central File Exchange.2021. Retrieved February 5, 2021. Reference Source
24. Xiao X, Zhou Y, Gong Y: Content-Adaptive Superpixel Segmentation. IEEE Trans. Image Process. June 2018; 27(6): 2883–2896. PubMed Abstract | Publisher Full Text
25. Dagher I: COMBINING CONTOUR-BASED AND REGION-BASED IN IMAGE SEGMENTATION. [Code]. Zenodo. 2023. Publisher Full Text

Comments on this article Comments (0)

Version 3

VERSION 3 PUBLISHED 11 Oct 2023

Author details Author details

¹ Computer Engineering Department, University of Balamand, Balamand, North Governorate, Lebanon

Issam Dagher
Roles: Methodology

Elie Abboud
Roles: Software

Competing interests

No competing interests were disclosed.

Grant information

The author(s) declared that no grants were involved in supporting this work.

Article Versions (3)

version 3

Revised

Published: 02 Apr 2024, 12:1312

https://doi.org/10.12688/f1000research.140872.3

version 2

Revised

Published: 16 Nov 2023, 12:1312

https://doi.org/10.12688/f1000research.140872.2

version 1

Published: 11 Oct 2023, 12:1312

https://doi.org/10.12688/f1000research.140872.1

© 2023 Dagher I and Abboud E. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Download

Export To

metrics

	Views	Downloads
F1000Research	-	-
PubMed Central Data from PMC are received and updated monthly.	-	-

Citations

SEE MORE DETAILS

CITE

how to cite this article

Dagher I and Abboud E. Combining contour-based and region-based in image segmentation [version 1; peer review: 1 approved with reservations]. F1000Research 2023, 12:1312 (https://doi.org/10.12688/f1000research.140872.1)

NOTE: If applicable, it is important to ensure the information in square brackets after the title is included in all citations of this article.

track

receive updates on this article

Track an article to receive email alerts on any updates to this article.

Open Peer Review

Version 1

VERSION 1

PUBLISHED 11 Oct 2023

Views

Reviewer Report 03 Nov 2023

Sandra Jardim, Instituto Politecnico de Tomar, Tomar, Santarém District, Portugal

Approved with Reservations

https://doi.org/10.5256/f1000research.154274.r215122

In the manuscript “Combining contour-based and region-based in image segmentation” is presented an image segmentation framework that combines region and contour-based approaches. The work presented by the authors addresses an interesting and current subject, but it presents some flaws that ... Continue reading

The statement "...because none of them can provide a coherent framework for achieving quick and efficient segmentation of images" has as reference a work published in 2019. Since then, many image segmentation approaches have been proposed. Indeed, in my opinion, some of them with very good results.
The authors should consider adding a "Related work" section, where they should present and discuss the most relevant and recent image segmentation approaches.
There are other and recent approaches like the one presented by the authors. Although the comparison made with the chosen methods is interesting, authors must compare the proposed method with similar approaches, to demonstrate the innovation of their proposal when compared to the most recent published works.
From the middle of the first page to the beginning of the 5th, the authors only describe basic and well-known concepts/methods. I think that it can/should be resumed.
Figure 4 - The image is not a flowchart in a standard form. The authors should consider presenting a conventional flowchart.
Figure 8 - From the images presented it's difficult to evaluate the results. Indeed, it seems that the plane is not segmented.
The methods identified by acronyms in Table 2 must be identified in its full description. Additionally, a reference must be given for each one.
The conclusions are too succinct and should be rewritten in greater detail. The conclusions are too succinct and should be rewritten in greater detail. Additionally, the limitations of the proposed approach must be mentioned.

Is the work clearly and accurately presented and does it cite the current literature?

Partly
Is the study design appropriate and is the work technically sound?

Partly
Are sufficient details of methods and analysis provided to allow replication by others?

Yes
If applicable, is the statistical analysis and its interpretation appropriate?

Yes
Are all the source data underlying the results available to ensure full reproducibility?

Yes
Are the conclusions drawn adequately supported by the results?

Partly

Competing Interests: No competing interests were disclosed.

Reviewer Expertise: Machine/Deep Learning; Computational Vision; Image Processing

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.

CITE

Report a concern

Author Response 16 Nov 2023

Issam Dagher, Computer Engineering Department, University of Balamand, Balamand, Lebanon

16 Nov 2023

Author Response
1. As you suggested I removed the statement "...because none of them can provide a coherent framework for achieving quick and efficient segmentation of images"

2. I added a ... Continue reading
1. As you suggested I removed the statement "...because none of them can provide a coherent framework for achieving quick and efficient segmentation of images"

2. I added a Related work paragraph:

3. I compared my approach with similar approaches:

Content-Adaptive Superpixel Segmentation.

Linear Iterative Clustering.

k-means

Wavelet.

MVSM

BSM

VTSM

4. The basic and well known methods are summarized.

5. I changed the flowchart in Figure 4.

6. I changed Figure 8.

7. I explained all the methods in the Related work section.

8. I included more details in the Conclusion section.

I thank the reviewer for putting his time in improving the paper.
1. As you suggested I removed the statement "...because none of them can provide a coherent framework for achieving quick and efficient segmentation of images"

2. I added a Related work paragraph:

3. I compared my approach with similar approaches:

Content-Adaptive Superpixel Segmentation.

Linear Iterative Clustering.

k-means

Wavelet.

MVSM

BSM

VTSM

4. The basic and well known methods are summarized.

5. I changed the flowchart in Figure 4.

6. I changed Figure 8.

7. I explained all the methods in the Related work section.

8. I included more details in the Conclusion section.

I thank the reviewer for putting his time in improving the paper.
Competing Interests: No Competing interests. Close
Report a concern
Respond or Comment

COMMENTS ON THIS REPORT

Author Response 16 Nov 2023

Issam Dagher, Computer Engineering Department, University of Balamand, Balamand, Lebanon

16 Nov 2023

Author Response
1. As you suggested I removed the statement "...because none of them can provide a coherent framework for achieving quick and efficient segmentation of images"

2. I added a ... Continue reading
1. As you suggested I removed the statement "...because none of them can provide a coherent framework for achieving quick and efficient segmentation of images"

2. I added a Related work paragraph:

3. I compared my approach with similar approaches:

Content-Adaptive Superpixel Segmentation.

Linear Iterative Clustering.

k-means

Wavelet.

MVSM

BSM

VTSM

4. The basic and well known methods are summarized.

5. I changed the flowchart in Figure 4.

6. I changed Figure 8.

7. I explained all the methods in the Related work section.

8. I included more details in the Conclusion section.

I thank the reviewer for putting his time in improving the paper.
1. As you suggested I removed the statement "...because none of them can provide a coherent framework for achieving quick and efficient segmentation of images"

2. I added a Related work paragraph:

3. I compared my approach with similar approaches:

Content-Adaptive Superpixel Segmentation.

Linear Iterative Clustering.

k-means

Wavelet.

MVSM

BSM

VTSM

4. The basic and well known methods are summarized.

5. I changed the flowchart in Figure 4.

6. I changed Figure 8.

7. I explained all the methods in the Related work section.

8. I included more details in the Conclusion section.

I thank the reviewer for putting his time in improving the paper.
Competing Interests: No Competing interests. Close
Report a concern

Comments on this article Comments (0)

Version 3

VERSION 3 PUBLISHED 11 Oct 2023

Open Peer Review

Reviewer Status

Reviewer Reports

	Invited Reviewers
	1	2	3
Version 3 (revision) 02 Apr 24	read		read
Version 2 (revision) 16 Nov 23		read
Version 1 11 Oct 23	read

Sandra Jardim, Instituto Politecnico de Tomar, Tomar, Portugal
Farid Garcia Lamont, Universidad Autónoma del Estado de México, Texcoco, Mexico
Nishtha Parashar, Amity University (Ringgold ID: 77282), Gwalior,, India

Comments on this article

All Comments(0)

Add a comment

Browse by related subjects

Back to all reports

Reviewer Report

4 Views

14 Aug 2024 | for Version 3

Nishtha Parashar, Department of Computer Science & Engineering, Amity University (Ringgold ID: 77282), Gwalior,, Madhya Pradesh, India

4 Views Cite this report Responses(0)

Approved With Reservations

1. The proposed strategy needs further optimization to enhance its effectiveness.
2. A detailed comparison of the proposed approach with existing methodologies should be added in the paper.
3. The flow chart should be redesigned.
4. The grammar should be rechecked to ensure that sentences follow the passive voice, improving the effectiveness of the research paper.

Is the work clearly and accurately presented and does it cite the current literature?

Partly
Is the study design appropriate and is the work technically sound?

Partly
Are sufficient details of methods and analysis provided to allow replication by others?

Yes
If applicable, is the statistical analysis and its interpretation appropriate?

Yes
Are all the source data underlying the results available to ensure full reproducibility?

Yes
Are the conclusions drawn adequately supported by the results?

Partly

References

1. X, Wang S, Li K, Kallidromitis Y, et al.: Hierarchical open-vocabulary universal image segmentation.https://www.researchgate.net/publication/372073999_Hierarchical_Open-vocabulary_Universal_Image_Segmentation/. 2023.
2. Yu H, Jiang L, Fan J, Xie S, et al.: A feature-weighted suppressed possibilistic fuzzy c-means clustering algorithm and its application on color image segmentation. Expert Systems with Applications. 2024; 241. Publisher Full Text

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

Image Mining, Computer Vision

Respond to this report

Responses (0)

Back to all reports

Reviewer Report

3 Views

10 Apr 2024 | for Version 3

Sandra Jardim, Instituto Politecnico de Tomar, Tomar, Santarém District, Portugal

3 Views Cite this report Responses(0)

Approved

All comments and suggestions were taken into consideration by the authors when reviewing the manuscript.

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

Machine/Deep Learning; Computational Vision; Image Processing

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.

Respond to this report

Responses (0)

Back to all reports

Reviewer Report

16 Views

07 Mar 2024 | for Version 2

Farid Garcia Lamont, Universidad Autónoma del Estado de México, Texcoco, Mexico

16 Views Cite this report Responses(1)

Approved With Reservations

The paper is well written and easy to follow. However, the main weakness of the paper is that the authors do not compare their results regarding related works, since they do not employ the metrics that usually are used to evaluate quantitatively the segmented images.

The authors employ the Berkeley Segmentation Database to test their methodology. This dataset is widely employed as bechmark by several related works to test algorithms of image segmentation by color features; the database contains ground truth (images hand-segmented) of each image. So, the resulting images are compared regarding their respective ground truth using the probabilistic random index, variation information and global consistency error metrics, to evaluate quantitatively. I suggest the authors to employ these metrics; thereby, it would be easier to compare the performance of their model with respect to previous works that employ this database.

Is the work clearly and accurately presented and does it cite the current literature?

Yes
Is the study design appropriate and is the work technically sound?

Partly
Are sufficient details of methods and analysis provided to allow replication by others?

Partly
If applicable, is the statistical analysis and its interpretation appropriate?

Partly
Are all the source data underlying the results available to ensure full reproducibility?

Yes
Are the conclusions drawn adequately supported by the results?

Partly

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

Image processing by color features.

Respond to this report

Responses (1)

Author Response

13 Apr 2024

Issam Dagher, Computer Engineering Department, University of Balamand, Balamand, Lebanon

"The authors employ the Berkeley Segmentation Database to test their methodology. This dataset is widely employed as bechmark by several related works to test algorithms of image segmentation by color features; the database contains ground truth (images hand-segmented) of each image. So, the resulting images are compared regarding their respective ground truth using the probabilistic random index, variation information and global consistency error metrics, to evaluate quantitatively. I suggest the authors to employ these metrics; thereby, it would be easier to compare the performance of their model with respect to previous works that employ this database."

I did more simulations using the PRI validation index.
And I compared our algorithm with other algorithms.
The results are shown in Table 3.

View more View less

Competing Interests

No Competing interests

Back to all reports

Reviewer Report

40 Views

03 Nov 2023 | for Version 1

Sandra Jardim, Instituto Politecnico de Tomar, Tomar, Santarém District, Portugal

40 Views Cite this report Responses(1)

Approved With Reservations

The statement "...because none of them can provide a coherent framework for achieving quick and efficient segmentation of images" has as reference a work published in 2019. Since then, many image segmentation approaches have been proposed. Indeed, in my opinion, some of them with very good results.
The authors should consider adding a "Related work" section, where they should present and discuss the most relevant and recent image segmentation approaches.
There are other and recent approaches like the one presented by the authors. Although the comparison made with the chosen methods is interesting, authors must compare the proposed method with similar approaches, to demonstrate the innovation of their proposal when compared to the most recent published works.
From the middle of the first page to the beginning of the 5th, the authors only describe basic and well-known concepts/methods. I think that it can/should be resumed.
Figure 4 - The image is not a flowchart in a standard form. The authors should consider presenting a conventional flowchart.
Figure 8 - From the images presented it's difficult to evaluate the results. Indeed, it seems that the plane is not segmented.
The methods identified by acronyms in Table 2 must be identified in its full description. Additionally, a reference must be given for each one.
The conclusions are too succinct and should be rewritten in greater detail. The conclusions are too succinct and should be rewritten in greater detail. Additionally, the limitations of the proposed approach must be mentioned.

Is the work clearly and accurately presented and does it cite the current literature?

Partly
Is the study design appropriate and is the work technically sound?

Partly
Are sufficient details of methods and analysis provided to allow replication by others?

Yes
If applicable, is the statistical analysis and its interpretation appropriate?

Yes
Are all the source data underlying the results available to ensure full reproducibility?

Yes
Are the conclusions drawn adequately supported by the results?

Partly

Competing Interests

No competing interests were disclosed.

Reviewer Expertise

Machine/Deep Learning; Computational Vision; Image Processing

Respond to this report

Responses (1)

Author Response

16 Nov 2023

Issam Dagher, Computer Engineering Department, University of Balamand, Balamand, Lebanon

1. As you suggested I removed the statement "...because none of them can provide a coherent framework for achieving quick and efficient segmentation of images"

2. I added a Related work paragraph:

3. I compared my approach with similar approaches:

Content-Adaptive Superpixel Segmentation.
Linear Iterative Clustering.
k-means
Wavelet.
MVSM
BSM
VTSM

4. The basic and well known methods are summarized.

5. I changed the flowchart in Figure 4.

6. I changed Figure 8.

7. I explained all the methods in the Related work section.

8. I included more details in the Conclusion section.

I thank the reviewer for putting his time in improving the paper.

View more View less

Competing Interests

No Competing interests.

Alongside their report, reviewers assign a status to the article:

Approved - the paper is scientifically sound in its current form and only minor, if any, improvements are suggested

Approved with reservations - A number of small changes, sometimes more significant revisions are required to address specific details and improve the papers academic merit.

Not approved - fundamental flaws in the paper seriously undermine the findings and conclusions

[1] 1. Saxena S, Jain S, Tripathi S, et al.: Comparative Analysis of Image Segmentation Techniques.Hura GS, Singh AK, Siong Hoe L, editors. Advances in Communication and Computational Technology. ICACCT 2019. Lecture Notes in Electrical Engineering. Vol. 668. . Singapore: Springer; 2021.

[2] 2. Fan L, Zhang F, Fan H, et al.: Brief review of image denoising techniques. Vis. Comput. Ind. Biomed. Art. 2019; 2: 7. PubMed Abstract | Publisher Full Text | Free Full Text

[3] 3. Gao J, Wang B, Wang Z, et al.: A wavelet transform-based image segmentation method. Optik. 2020; 208: 164123. Publisher Full Text

[4] 4. Yang AY, Wright J, Ma Y, et al.: Unsupervised segmentation of natural images via lossy data compression. Comput. Vis. Image Underst. 2008; 110(2): 212–225. Publisher Full Text

[5] 5. Felzenszwalb PF, Huttenlocher DP: Efficient graph-based image segmentation. Int. J. Comput. Vis. 2004; 59(2): 167–181. [7] X. Liu, Q. Xu, J. Ma, H. J. Publisher Full Text

[6] 6. Sagiv C, Sochen NA, Zeevi YY: Integrated active contours for texture segmentation. IEEE Trans. Image Process. 2006; 15(6): 1633–1646. PubMed Abstract | Publisher Full Text

[7] 7. Shi J, Belongie S, Leung T, et al.: Image and video segmentation: the normalized cut framework. Proceedings 1998 International Conference on Image Processing. ICIP98. 1998; vol. 1: pp. 943–947.

[8] 8. Malik J, Belongie S, Leung T, et al.: Contour and texture analysis for image segmentation. Int. J. Comput. Vis. 2001; 43(1): 7–27. Publisher Full Text

[9] 9. Wang C, Liu Z, Chan S: Superpixel-based hand gesture recognition with Kinect depth camera. IEEE Trans. Multimed. 2015; 17(1): 29–39. Publisher Full Text

[10] 10. Yang F, Lu H, Yang M: Robust superpixel tracking. IEEE Trans. Image Process. 2014; 23(4): 1639–1651. PubMed Abstract | Publisher Full Text

[11] 11. Gong Y, Zhou Y: Differential evolutionary superpixel segmentation. IEEE Trans. Image Process. 2018; 27(3): 1390–1404. PubMed Abstract | Publisher Full Text

[12] 12. Achanta R, Shaji A, Smith K, et al.: SLIC superpixels compared to state-of-the-art superpixel methods. IEEE Trans. Pattern Anal. Mach. Intell. 2012; 34(11): 2274–2282. PubMed Abstract | Publisher Full Text

[13] 13. Angelina S, Suresh LP, Veni SHK: Image segmentation based on genetic algorithm for region growth and region merging. 2012 International Conference on Computing, Electronics and Electrical Technologies (ICCEET), Kumaracoil, 2012. 2012; pp. 970–974. Publisher Full Text

[14] 14. Senthilkumaran N, Rajesh R: Edge detection techniques for image segmentation - A survey of soft computing approaches. Int. J. Recent Trends Eng. 2009; 1(2): 250.

[15] 15. Wang W, Duan L, Wang Y: Fast Image Segmentation Using Two-Dimensional Otsu Based on Estimation of Distribution Algorithm. J. Electr. Comput. Eng. 2017; 2017: 1–12. Article ID 1735176. Publisher Full Text

[16] 16. Zanaty EA, El-Zoghdy SF: A novel approach for color image segmentation based on region growing. Int. J. Comput. Appl. 2017; 39(3): 123–139. 12 pages, 2017. Publisher Full Text

[17] 17. Canny J: A Computational Approach to Edge Detection. IEEE Trans. Pattern Anal. Mach. Intell. Nov. 1986; PAMI-8(6): 679–698. Publisher Full Text

[18] 18. Huang H, Meng F, Zhou S, et al.: Brain Image Segmentation Based on FCM Clustering Algorithm and Rough Set. IEEE Access. 2019; 7: 12386–12396. Publisher Full Text

[19] 19. Abderaouf Z, Nadjia B, Saliha O: License plate character segmentation based on horizontal projection and connected component analysis. 2014 World Symposium on Computer Applications & Research (WSCAR), Sousse. 2014; pp. 1–5. Publisher Full Text

[20] 20. Park B-Y, Kim H-H, Hong B-W: A Multilabel Texture Segmentation Based on Local Entropy Signature. Math. Probl. Eng. 2013; 6. 2013. ArticleID 651581.

[21] 21. Garcia-Lamont F, Cervantes J, López A, et al.: Segmentation of images by color features: A survey. Neurocomputing. 2018; 292: 1–27. Publisher Full Text

[22] 22. Martin D, Fowlkes C, Tal D, et al.: A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics. Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001, Vancouver, BC, Canada. 2001; vol. 2: pp. 416–423. Publisher Full Text

[23] 23. Balafar MA: segmentation evaluatation, MATLAB Central File Exchange.2021. Retrieved February 5, 2021. Reference Source

[24] 24. Xiao X, Zhou Y, Gong Y: Content-Adaptive Superpixel Segmentation. IEEE Trans. Image Process. June 2018; 27(6): 2883–2896. PubMed Abstract | Publisher Full Text

[25] 25. Dagher I: COMBINING CONTOUR-BASED AND REGION-BASED IN IMAGE SEGMENTATION. [Code]. Zenodo. 2023. Publisher Full Text

Combining contour-based and region-based in image segmentation

Abstract

Keywords

Introduction

Image segmentation techniques

Threshold-based segmentation

Region growth segmentation

Edge detection

Clustering

Connected component algorithm

Texture filters: Gabor wavelets

Methods

Figure 1. (a) Original image; (b) Grayscale image; (c) Edge image; (d) Initial set of centers.

Figure 2. (a) Initial set of centers; (b) After the first reduction.

Figure 3. Reduction of RGB centers using FCM.

Figure 4. Flowchart of the proposed approach.

Illustration of the algorithm

Figure 5. Original 3 images (Upper Row) and their Corresponding Edge images (Lower Row).

Figure 6. Edge images (Upper Row) and their corresponding initial centers (Lower Row).

Figure 7. Initial set of centers (Upper row) and their corresponding optimum number of centers (Lower row).

Figure 8. Segmented images (lower row) and their corresponding optimum number of centers (upper row).

Results

Dataset

Figure 9. Samples from the BSD500 database.

Segmentation metrics

Results of proposed approach

Figure 10. Original 3 images (First column).

Table 1. Performance of the various approaches on the BSD500.

Figure 11. Cameraman and Lena images.

Table 2. Performance of our approach for the 2 images compared to the results obtained in Ref. 3.

Conclusion

Data availability

Source data

Extended data

Acknowledgements

References

Comments on this article Comments (0)

Open Peer Review

Comments on this article Comments (0)

Open Peer Review

Reviewer Status

Reviewer Reports

Comments on this article

Browse by related subjects

Competing Interests Policy

Stay Updated