Source-linked AI summary
Image Segmentation by Using Threshold Techniques
Salem Saleh Al-amri, N. V. Kalyankar, Khamitkar S. D.
TL;DR
Selecting an effective threshold is difficult even though thresholding can simplify image segmentation. This paper compares five threshold techniques on three satellite images and reports that HDT and EMT perform best.
Problem
Selecting an adequate threshold value T for converting grayscale images into useful binary foreground-background segments is not always easy.
Method
The paper compares Mean, P-Tile, HDT, EMT, and visual threshold techniques on three satellite images.
Results
HDT and EMT are reported as the best-performing threshold techniques for the three satellite images.
Takeaways & Limitations
The comparison identifies HDT and EMT as the preferred techniques within the evaluated satellite-image cases.
Abstract
from arXiv · showhide
This paper attempts to undertake the study of segmentation image techniques by using five threshold methods as Mean method, P-tile method, Histogram Dependent Technique (HDT), Edge Maximization Technique (EMT) and visual Technique and they are compared with one another so as to choose the best technique for threshold segmentation techniques image. These techniques applied on three satellite images to choose base guesses for threshold segmentation image.
1 INTRODUCTION
Image segmentation methods are organized around intensity discontinuity or similarity, while prior aviation-image comparisons report that statistic iterative thresholding improves anti-noise capability and performs well on difficult images.
- Segmentation principles: Threshold segmentation is framed around two intensity properties: discontinuity, which partitions images at abrupt changes such as edges, and similarity, which groups regions according to predefined criteria.
- Statistical thresholding: The statistic iterative approach derives thresholding from the image gray histogram and Gaussian statistical distribution.
- Prior thresholding results: Compared with histogram and adaptive methods, statistic iterative thresholding improved anti-noise capability and performed well on aviation images that were difficult to segment using full-value thresholding.
2 THRESHOLDS
Thresholding separates foreground from background by converting a grayscale image into a binary image using a selected value T. Choosing T is not always easy, motivating automatic schemes that use image characteristics.
- Thresholding: Thresholding converts grayscale images into binary images that preserve the position and shape of foreground objects while reducing data complexity.This simplification supports subsequent object-recognition processing.
- Threshold selection: Selecting an adequate threshold value T is difficult in practice.
- Automatic thresholding: Automatic threshold schemes select a threshold without human intervention using object intensities, sizes, image fractions, and the number of object types.
4 THRESHOLD TECHNIQUES
The paper defines threshold-based image segmentation and applies five techniques—Mean, P-Tile, HDT, EMT, and Visual—to satellite images. Each method targets different assumptions about pixel distributions, object area, homogeneity, or illumination.
- 4.1 Mean Technique: The Mean technique uses the mean pixel value as the threshold and suits images with roughly equal object and background areas, a condition that rarely occurs.
- 4.2 P-Tile Technique: The P-Tile method selects a gray-level threshold from the expected object area percentage, assuming brighter objects occupy a fixed proportion of the image.That proportion is denoted P%.
- 4.3 Histogram Dependent Technique: HDT estimates a threshold separating two homogeneous, well-separated regions by minimizing within-group variance across pixels below and above the threshold.It is suitable when object and background regions are largely homogeneous, apart from their boundary.
- 4.4 Edge Maximization Technique: EMT is intended for images containing multiple homogeneous regions or illumination changes, where object and background portions may otherwise be merged or confused.The paper notes that automatic threshold selection performs better when regions are large, homogeneous, and well separated.
- 5.1 Testing Procedure: The techniques were implemented in MATLAB R2007a (7.4a) and tested on three satellite-image types, with figures showing Visual thresholds of 127, 167, and 43.
- 4 THRESHOLD TECHNIQUES: Five threshold techniques are applied: Mean, P-Tile, Histogram Dependent Technique (HDT), Edge Maximization Technique (EMT), and Visual.The techniques are evaluated on three types of satellite images.
6 CONCLUSIONS
The paper compares five threshold-segmentation techniques across three satellite images. It concludes that Histogram Dependent Technique (HDT) and Edge Maximization Technique (EMT) are the best-performing methods.
- 6 CONCLUSIONS: HDT and EMT are identified as the best techniques for threshold-based image segmentation.The conclusion reports this outcome from comparative experiments across the evaluated satellite images.
- 6 CONCLUSIONS: The study conducts comparative experiments using five threshold-segmentation methods on three satellite images.The evaluated methods are Mean, P-Tile, Histogram Dependent Technique (HDT), Edge Maximization Technique (EMT), and visual Technique.
- 6 CONCLUSIONS: The comparison includes Mean, P-Tile, HDT, EMT, and visual threshold techniques.
AUTHORS PROFILE
The authors profile presents Salem Saleh Al-amri, N.V. Kalyankar, and S.D. Khamitkar through their academic qualifications, institutional roles, memberships, and publication or supervision records.
- Salem Saleh Al-amri: Salem Saleh Al-amri holds engineering and computer science qualifications, is pursuing a Ph.D. in computer science, and lectures at Aden University, Yemen.He received a B.E. in Mechanical Engineering, an M.Sc. in Computer Science (IT), and reports international and national conference participation.
- Dr. N.V. Kalyankar: Dr. N.V. Kalyankar has degrees in science, nuclear physics, higher education, and physics, and serves as principal of Yeshwant Mahavidyalaya College.His profile also lists academic memberships and one book, 23 journal papers, two seminar papers, and three conference papers.