Nd_spacing — README

Overview

This subfolder contains the Python port of the original MATLAB image analysis pipeline for measuring local Nd-rich spacing and orientations from micrographs. The main entrypoint is the ANALYSE_IMAGES.py script that:

• reads images,
• performs contrast enhancement & background-corrected binarisation,
• computes spacing (local mean distances) and orientations,
• computes distributions/histograms and exports results per image.

Pipeline (high level)

1. Image & parameters:

   • Main script: ANALYSE_IMAGES.py
   • Global settings: PARAMETRES.py

2. Binarisation:

   • CLAHE + background correction + thresholding (Otsu or adaptive).
   • Implementation: binarisation

3. Spacing & statistics:

   • Local distance map: distance
   • Normalized histogram / distribution: distribution

4. Orientation:

   • Orientation/angle calculation: calculangle

5. Post-processing / export:

   • Visualization and XLS export: dessinexport
   • Grain prolongation/cleanup helpers: prolong_nd

Files of interest

• ANALYSE_IMAGES.py — main orchestration script, loops samples/types/photos and calls the pipeline steps.
• PARAMETRES.py — dataset/sample parameters (ECHTS, TYPES, NBPHOTOS, etc.).
• Nd_spacing/Binarisation_methods/binarisation clahe+BGcorr+adaptive thresholding.py — detailed binarisation routine.
• distance.py — compute local mean spacing map.
• distribution.py — convert DM map into histogram classes and normalized distribution.
• calculangle.py — compute local crystal/grain orientations and class histograms.
• dessinexport.py — saves figures and XLSX results.
• prolong_nd.py — morphological helpers used for grain prolongation and cleaning.

Quick start

1. Install dependencies:

   pip install -r requirements.txt

2. Configure PARAMETRES.py for your dataset (sample names, photo counts, paths).
3. Run the analyzer from the Nd_spacing folder:

   python ANAYLSE_IMAGES.py

   (Script prints progress and writes per-image XLSX in the "Résultats" folder.)

Configuration notes

• Binarisation parameters (CLAHE clip_limit, tile_grid_size, blur kernel, threshold method, adaptive block_size/C) are configurable in the binarisation call made by ANALYSE_IMAGES.py.
• Distribution parameters such as dmax and pasdistri are imported from PARAMETRES.py and used by distribution.

Implementation quirks & pointers

• The pipeline mirrors the original MATLAB flow (see Code_matlab_OT_original for the MATLAB reference).
• The binarisation routine applies CLAHE then optional Gaussian background estimation before thresholding.
• distribution computes histogram class boundaries consistent with the MATLAB original.
• If you want to inspect intermediate outputs, ANALYSE_IMAGES.py writes masks and labeled images (TIFF) for debugging.

____________________________________________________

external folders

segment every grain

Segmenteverygrain on github

DDPM

Conditional Denoising Diffusion Probabilistic Model for the Cast-Forged AZ80 Microstructure

DDPM on github

This is the implementation of conditional denoising diffusion probabilistic model to generate SEM microstructure images of seen and unseen cast-forged AZ80 magnesium alloy components. The model is conditioned based on the cast geometry, casting cooling rate, soaking process, pre-forging heat treatment, forging temperature, location of extracted metallography sample, and magnification.

Please cite our paper as follows: E. Azqadan, H. Jahed, A. Arami, Predictive microstructure image generation using denoising diffusion probabilistic models, Acta Mater. 261 (2023), 119406, https://doi.org/10.1016/j.actamat.2023.119406.

The following repositories were used as inspiration for this implementation: https://github.com/dome272/Diffusion-Models-pytorch/tree/main https://github.com/lucidrains/denoising-diffusion-pytorch https://github.com/CompVis/latent-diffusion