SNR* Tool (ImageJ Macro)

Covariance-based SNR estimation macro for ImageJ.

Overview

SNR* is a covariance-based signal-to-noise ratio (SNR) estimator calculated from two observed images acquired under identical imaging conditions.

This method separates signal variance and noise variance using statistical relationships between two images.

Theory

Given two observed images:

I₁ = S + N₁
I₂ = S + N₂

where
S : true signal
N₁, N₂ : independent noise components

The variances are estimated as:

Signal variance: σ_s² = Cov(I₁, I₂)

Noise variance: σ_n² = Var(I₁ − I₂) / 2

SNR* is defined as:

SNR* (dB) = 10 log₁₀ (σ_s² / σ_n²)

Requirements

• ImageJ 1.53 or later
• Exactly two observed images
• Images must:
  • Have identical dimensions
  • Be acquired under identical imaging conditions
  • Represent independent noise realizations

Installation (Toolset Version)

1. Copy SNR_star_Tool.ijm
2. Place the file into:

ImageJ/macros/toolsets/

Example (Windows):

C:\ImageJ\macros\toolsets\

1. Restart ImageJ
2. Activate via:

More Tools >> SNR_star_Tool

The SNR* icon will appear in the ImageJ toolbar.

Usage

1. Prepare a folder containing exactly two observed images.
2. Click the SNR* tool icon.
3. Select the folder.
4. Draw an ROI on the first image.
5. The SNR* result is displayed in the Log window.

Output

The macro outputs:

• SNR* [dB]
• ROI size
• Signal variance (σ_s²)
• Noise variance (σ_n²)

Output format follows the internal macro implementation.

Notes

• The method assumes:
  • Additive noise
  • Zero-mean independent noise between the two images
• If covariance becomes negative, imaging conditions may not satisfy assumptions.
• Larger ROIs improve estimation stability.

References

• Tabuchi M, Kiguchi T, Ikenaga H. SNR estimation for image quality evaluation in X-ray CT. Jpn J Radiol Technol. 2022;78:464–72. https://doi.org/10.6009/jjrt.2022-1154.

Version

1.0 (2026)

Author

Motohiro TABUCHI

License

This project is licensed under the MIT License. See the LICENSE file for details.