[Paper Review] Image Segmentation Algorithms Overview

Yuheng Song, Hao Yan|arXiv (Cornell University)|Jul 7, 2017

Image and Object Detection Techniques2 references128 citations

TL;DR

This paper analyzes and summarizes major image segmentation algorithms (region-based, edge-based, clustering, weakly-supervised CNN) and discusses their advantages, disadvantages, and future trends.

ABSTRACT

The technology of image segmentation is widely used in medical image processing, face recognition pedestrian detection, etc. The current image segmentation techniques include region-based segmentation, edge detection segmentation, segmentation based on clustering, segmentation based on weakly-supervised learning in CNN, etc. This paper analyzes and summarizes these algorithms of image segmentation, and compares the advantages and disadvantages of different algorithms. Finally, we make a prediction of the development trend of image segmentation with the combination of these algorithms.

Motivation & Objective

• Motivate the importance of image segmentation in fields like medical imaging, face recognition, and pedestrian detection.
• Systematically review current segmentation techniques and categorize them by underlying principles.
• Compare strengths and weaknesses of different algorithm families.
• Forecast potential development trends by combining insights from various approaches.

Proposed method

• Survey and categorization of existing image segmentation algorithms.
• Qualitative comparison of advantages and disadvantages across categories.
• Discussion of integration opportunities among algorithm families to improve performance.

Experimental results

Research questions

• RQ1 What are the main categories of image segmentation algorithms?
• RQ2 What are the comparative advantages and limitations of region-based, edge-based, clustering-based, and weakly-supervised CNN approaches?
• RQ3 What development trends can be predicted by combining different segmentation techniques?

Key findings

• Various segmentation approaches have distinct strengths and trade-offs relevant to applications like medical imaging and recognition tasks.
• Region-based, edge-based, clustering-based, and weakly-supervised CNN methods each have unique drawbacks that influence applicability.
• Combining elements from multiple algorithms is suggested as a path toward improved segmentation performance.
• The paper provides a structured summary to guide future research in image segmentation.