Usage

This section documents the practical, script-based usage model of spectral-unmixing.

The package is designed to be driven from project-specific Python scripts that can be executed cell by cell in the VS Code interactive window or a similar notebook-like environment. The reusable package functions live in spectral_unmixing/, while the repository’s user_scripts/ folder shows how these functions are combined for real workflows.

• Example datasets
  • What is included
  • Dataset-to-tutorial mapping
  • Dataset descriptions and provenance
  • Citation
• Functionality overview
  • Spectral unmixing workflows
  • Filtering and projection helpers
  • Registration helpers
  • Visualization helpers
• Basic unmixing example
  • How to use this tutorial
  • What this tutorial covers
  • Imports
  • Define input and output paths
  • Spectral unmixing with a manually set alpha
  • mean_ratio method
  • linear_fit method
  • corr_min method
  • mi_min method
  • What to change for your own data
• Full 3D+t unmixing example
  • How to use this tutorial
  • What this tutorial covers
  • Imports
  • Define input and output paths
  • Manually set alpha
  • mean_ratio on a full TZCYX stack
  • linear_fit on a full TZCYX stack
  • corr_min on a full TZCYX stack
  • mi_min on a full TZCYX stack
• Bidirectional unmixing example
  • How to use this tutorial
  • Core idea
  • Imports
  • Define input and output paths
  • Bidirectional unmixing with fixed coefficients
  • Bidirectional mean_ratio
  • Bidirectional linear_fit
  • Bidirectional corr_min
  • Bidirectional mi_min
  • What to tune most often
• PICASSO 2-color example
  • How to use this tutorial
  • What this tutorial covers
  • Imports
  • Define input and output paths
  • Inspect the measured channels
  • matlab_n blind unmixing
  • source_sink_n blind unmixing
• PICASSO 3-color example
  • How to use this tutorial
  • Imports
  • Define input and output paths
  • Inspect the measured channels
  • matlab_n blind unmixing
  • matlab_3c blind unmixing
  • source_sink_n blind unmixing
• PICASSO 5-color example
  • How to use this tutorial
  • What this tutorial is good for
  • Imports
  • Define input and output paths
  • Inspect the measured channels
  • matlab_n blind unmixing on five channels
  • source_sink_n blind unmixing on five channels
• Filtering, registration, and projection helper example
  • How to use this tutorial
  • What this tutorial covers
  • Imports
  • Define input and output paths
  • Load the stack
  • Correct intra-stack z-drift
  • Register the stack across time
  • Match histograms across time
  • Filter the registered stack
  • Project along z
  • Filter the projected stack again
  • Save the processed stack
  • Fine-tuning

How to read the tutorials

Each tutorial page is based on one interactive script from the repository’s user_scripts/ folder. The documentation follows the script cell by cell and explains:

• what each cell does,

• which parameters matter most,

• what a user can modify for a new dataset,

• and how those changes affect the unmixing result.

Recommended starting point

If you are new to the package, start in this order:

1. Basic unmixing example

2. Full 3D+t unmixing example

3. Filtering, registration, and projection helper example

4. Bidirectional unmixing example

5. the PICASSO-family blind-unmixing tutorials

The standard two-channel tutorial introduces the core package logic with the least complexity. The TZCYX tutorial then shows the same logic on full time-lapse z-stacks. The helper tutorial then shows how filtering, registration, histogram matching, and projection can be chained after unmixing. The bidirectional tutorial adds a more advanced alpha-estimation strategy for cases where the cross-talk is not strictly unidirectional. The PICASSO tutorials build on the same ideas but add blind coefficient estimation and multi-channel (more than two) configuration.