[Paper Review] Hybrid Image Segmentation using Discerner Cluster in FCM and Histogram Thresholding

Firas A. Jassim|arXiv (Cornell University)|Feb 6, 2013

Medical Image Segmentation Techniques6 references11 citations

TL;DR

This paper proposes a hybrid image segmentation method that combines fuzzy c-means (FCM) clustering with histogram thresholding to automatically determine an optimal threshold using a 'discerner cluster'—the cluster with the highest peak in the smoothed histogram. By applying FCM to histogram frequencies rather than pixel intensities, the method identifies the dominant cluster as the object region, achieving robust segmentation on standard test images with improved threshold selection accuracy.

ABSTRACT

Image thresholding has played an important role in image segmentation. This paper presents a hybrid approach for image segmentation based on the thresholding by fuzzy c-means (THFCM) algorithm for image segmentation. The goal of the proposed approach is to find a discerner cluster able to find an automatic threshold. The algorithm is formulated by applying the standard FCM clustering algorithm to the frequencies (y-values) on the smoothed histogram. Hence, the frequencies of an image can be used instead of the conventional whole data of image. The cluster that has the highest peak which represents the maximum frequency in the image histogram will play as an excellent role in determining a discerner cluster to the grey level image. Then, the pixels belong to the discerner cluster represent an object in the gray level histogram while the other clusters represent a background. Experimental results with standard test images have been obtained through the proposed approach (THFCM).

Motivation & Objective

• To address the challenge of automatic threshold selection in image segmentation for gray-level images.
• To improve the robustness and accuracy of thresholding by leveraging cluster analysis on image histogram frequencies.
• To identify a 'discerner cluster'—the cluster with the highest peak in the smoothed histogram—as a reliable indicator for object segmentation.
• To reduce dependency on manual threshold setting by automating the process through FCM clustering of histogram data.
• To enhance segmentation performance by combining FCM clustering with histogram-based thresholding techniques.

Proposed method

• The method applies the standard FCM clustering algorithm not to pixel intensities, but to the frequency values (y-values) of the smoothed image histogram.
• The cluster with the highest peak in the histogram, representing the most frequent intensity value, is designated as the 'discerner cluster'.
• Pixels belonging to the discerner cluster are classified as the object, while other clusters represent the background.
• The algorithm uses the centroid of the discerner cluster to determine the optimal threshold for segmentation.
• The approach transforms the image segmentation problem into a clustering task on the histogram's frequency distribution.
• The method is evaluated using standard test images to validate its effectiveness in automatic threshold selection.

Experimental results

Research questions

• RQ1 Can FCM clustering of histogram frequencies improve automatic threshold selection in image segmentation?
• RQ2 Does identifying the cluster with the highest peak (discerner cluster) lead to more accurate segmentation than conventional thresholding?
• RQ3 How does the proposed hybrid method compare to existing thresholding techniques in terms of segmentation accuracy and robustness?
• RQ4 Can the discerner cluster effectively distinguish object regions from background in gray-level images?
• RQ5 What is the impact of smoothing the histogram on the performance of the FCM-based thresholding approach?

Key findings

• The discerner cluster, defined as the FCM cluster with the highest peak in the smoothed histogram, consistently identified the object region in test images.
• The method achieved accurate segmentation without requiring manual threshold input, demonstrating automatic threshold selection capability.
• The use of histogram frequencies instead of raw pixel data reduced computational complexity while preserving segmentation accuracy.
• Experimental results on standard test images confirmed the method's effectiveness in identifying object boundaries.
• The approach outperformed basic thresholding methods by leveraging cluster structure in the histogram.
• The hybrid FCM-histogram thresholding method provided a robust solution for automatic image segmentation in gray-level images.