Related Experiment Video
Updated: Dec 25, 2025
12:08
From Voxels to Knowledge: A Practical Guide to the Segmentation of Complex Electron Microscopy 3D-Data
Published on: August 13, 2014
24.9K
Generative and discriminative model-based approaches to microscopic image restoration and segmentation
Shin Ishii1,2,3, Sehyung Lee1,3, Hidetoshi Urakubo1
1Graduate School of Informatics, Kyoto University, Kyoto 606-8501, Japan.
Microscopy (Oxford, England)
|March 28, 2020
Summary
Machine learning (ML) advances image processing, particularly for noisy, large microscopic images. This review explores ML applications like super-resolution, image restoration using convolutional neural networks (CNNs), and segmentation for electron microscopy (EM).
Area of Science:
- Computer Vision
- Machine Learning
- Microscopy Imaging
Background:
- Machine learning (ML) and convolutional neural networks (CNNs) are powerful tools for image processing.
- Microscopic images present challenges due to their 3D/4D nature, large size, and optical noise.
- Existing ML applications in microscopy are limited by these inherent complexities.
Purpose of the Study:
- To review recent ML advancements for feature reconstruction in microscopic images.
- To highlight the utility of ML techniques in overcoming limitations of microscopic image analysis.
- To focus on electron microscopy (EM) image processing applications.
Main Methods:
- Discusses multi-frame super-resolution using statistical generative models and Bayesian inference.
- Explains data-driven image restoration employing supervised discriminative ML techniques, with CNNs showing superior performance.
- Details image segmentation using data-driven CNNs, particularly for EM datasets.
Main Results:
- CNNs demonstrate effective performance in image restoration tasks for microscopic images.
- ML techniques, including generative and discriminative models, offer solutions for super-resolution and noise reduction.
- Data-driven CNNs are highly effective for image segmentation in EM.
Conclusions:
- ML technologies significantly enhance feature reconstruction for challenging microscopic images.
- CNNs are pivotal for achieving high performance in image restoration and segmentation.
- This review underscores the transformative potential of ML in advancing EM image analysis.
