ImagePipes

ImagePipes: MicroCT Image Processing Pipelines

ImagePipes is a modular and extensible Python-based toolkit designed for processing, analyzing, and segmenting high-resolution 3D micro-computed tomography (µCT) images. Developed with a focus on porous media and geoscientific applications, this suite of scripts enables streamlined, reproducible workflows for transforming raw µCT image data into meaningful quantitative insights.

These pipelines were created and tested in real-world research projects at ETH Zürich, Eawag, and Empa, and are now made publicly available to support the broader scientific and engineering community.

Core Functionalities

Image Registration: Rigid and Affine Registration with Elastix: Align misaligned or distorted image stacks using robust multi-resolution, affine transformations via the Elastix library.

Phase Cross-Correlation Registration: Quickly compute and apply translation-only corrections using Fourier-based phase cross-correlation, optimized for small shifts in µCT stacks.

Noise Reduction: 2D and 3D Non-Local Means (NLM) Denoising: Apply structure-preserving NLM filters on a slice-by-slice or volumetric basis, tuned to match commercial denoising settings. Handles various data types and supports zero-padding.

Segmentation & Clustering: K-Means Clustering of Masked Tomograms: Segment tomograms into clusters based on voxel intensity using scikit-learn’s KMeans. Customizable parameters allow flexible unsupervised classification.

Phase Segmentation: Post-process clustered tomograms to assign phase labels (e.g., solid, liquid, air). Includes hole-filling using neighborhood filtering and interface validation against raw intensity values.

Edge Enhancement: 3D Binary Erosion-Based Edge Refinement: Improve segmentation quality near phase interfaces by applying morphological erosion in 3D, with special handling for boundary slices.

Conversion Tools: CT Number to Concentration Conversion: Linearly map CT grayscale intensities to physical concentration values using experimentally calibrated slope/intercept values. Includes thresholding and metadata preservation.

3D Mesh Generation (Marching Cubes): Generate watertight surface meshes (.PLY format) from binary image stacks using marching cubes, with control over voxel spacing for real-world scale reconstruction.

Resampling & Masking: 3D Isotropic Resampling: Rescale volumetric image data to desired voxel sizes using nearest-neighbor or other interpolation methods, preserving binary or grayscale fidelity.

Tomogram Masking: Apply binary masks to extract regions of interest across large tomogram stacks. Supports folder hierarchies and data type consistency.

Use Cases: ImagePipes was built for high-resolution µCT analysis in:

Reactive transport modeling

Pore-scale flow visualization

Multiphase fluid characterization

CO₂ trapping and leakage risk analysis

Fracture-matrix interface studies

Sediment structure and grain morphology assessment

Its modular structure also makes it useful in materials science, medical imaging, and machine learning pre-processing.

Built With:

NumPy – numerical computation

SciPy – advanced image processing and resampling

scikit-image – denoising, registration, and morphology

scikit-learn – clustering

SimpleITK – image registration and format conversion

OpenCV – image I/O and processing

trimesh – mesh generation and export

tifffile – TIFF read/write

rasterio – geospatial TIFF transformation

License: This project is licensed under the MIT License – see the LICENSE file for details. You are free to use, modify, distribute, and build upon this work with attribution.
Copyright (c) 2025 EarthSciTech

Author: Amirsaman Rezaeyan

---

Cite as:
Amirsaman Rezaeyan, 2025, ImagePipes: multi-modular, open-source pipelines for the analysis of X-ray micro-computed tomography images of porous media and physical processes in porous media. https://github.com/EarthSciTech/ImagePipes