Electron Microscopy Denoising

Work in Progress

This example is currently being written. Check back soon for the complete guide.

Overview

Electron microscopy images often contain both random noise and structured artifacts like scan lines. This example demonstrates how autoStructN2V handles both types of noise in a single pipeline.

The Challenge

EM images typically have:

• Shot noise - Random, pixel-independent noise from electron detection
• Scan lines - Horizontal stripes from the scanning process
• Periodic artifacts - Regular patterns from electronics or beam instability

Standard Noise2Void can remove the shot noise but leaves the scan lines intact.

The Solution

AutoStructN2V's two-stage approach:

1. Stage 1 removes the random shot noise
2. Stage 2 automatically detects and removes the scan line pattern

Example Code

from autoStructN2V.pipeline import run_pipeline

config = {
    'input_dir': './em_images/',
    'output_dir': './denoised/',
    'run_stage1': True,
    'run_stage2': True,
    # EM-specific settings
    'patch_size': 64,
    'mode': '2d'
}

results = run_pipeline(config)

Expected Results

Coming soon: Before/after comparisons and quality metrics.

Adapting to Your Data

• Adjust patch_size based on your image resolution
• Use mode: '2.5d' for tomographic stacks
• See Configuration for all options