labelimage-tools 0.1.3

Utilities for 2-D labeled tissue segmentation images

labelimage-tools

labelimage-tools is a small utility package for 2-D labeled tissue segmentation images. It provides loading, validation, preprocessing, adjacency and contact extraction, junction detection, graph coloring, contours, and matplotlib plotting helpers.

Conventions

• Label images are 2-D NumPy arrays.
• Image coordinates are represented as (y, x).
• The default background label is 0.
• Labels do not need to be consecutive.
• Integer label values are preserved unless a function explicitly documents a relabeling operation.

Installation

Light install

pip install labelimage-tools

The light install supports array-based preprocessing, adjacency/contact graph calculation, and native NPZ/JSON graph I/O.

TIFF image I/O

For label-image I/O, installing the tiff extra is recommended because most scientific label images are TIFF files:

pip install "labelimage-tools[tiff]"

This installs tifffile, the recommended backend for scientific TIFF label images.

Recommended scientific-image install

For most scientific label-image workflows, install TIFF support plus accelerated junction scanning:

pip install "labelimage-tools[recommended]"

Other optional features

pip install "labelimage-tools[pillow]"          # PNG/JPEG/general image I/O
pip install "labelimage-tools[plot]"            # plotting, contours, graph coloring
pip install "labelimage-tools[graph-standard]"  # GraphML/GEXF graph I/O
pip install "labelimage-tools[junctions-accelerated]"  # numba-accelerated junction scanning
pip install "labelimage-tools[all]"             # everything

Feature	Install
Core array processing, adjacency, NPZ/JSON graph I/O	labelimage-tools
TIFF image I/O	labelimage-tools[tiff]
TIFF image I/O + accelerated junction scanning	labelimage-tools[recommended]
PNG/JPEG/general image I/O	labelimage-tools[pillow]
Plotting, contours, graph coloring	labelimage-tools[plot]
GraphML/GEXF graph I/O	labelimage-tools[graph-standard]
Everything	labelimage-tools[all]

From a source checkout:

python -m pip install -e .

For tests:

python -m pip install -e '.[test]'
python -m pytest

Load and preprocess labels

Use these helpers directly from scripts or notebooks. The intended starting point is load_image_pipeline(...): with its defaults, it crops foreground, removes disconnected bits of repeated labels, and fills internal gaps. This prepares the image so labels are clean, self-connected, unique regions that are ready for adjacency, contour, and junction operations, with neighboring labels touching across filled internal gaps rather than being separated by stray background holes.

import labelimage_tools as lit

labels = lit.load_image_pipeline("segmentation.tif")

Adjacency and contact graph

Adjacency is computed by vectorized neighbor scanning. Contact values are neighboring pixel-pair counts, useful as weights but not exact geometric lengths. Original label IDs are preserved as graph node IDs.

neighbors, contacts, centroids, pixel_counts = lit.graph_from_labels(labels)

lit.save_label_graph(
    "label_graph.npz",
    neighbors,
    contacts=contacts,
    centroids=centroids,
    pixel_counts=pixel_counts,
    source_image="segmentation.tif",
)

neighbors, contacts, centroids, pixel_counts, metadata = lit.load_label_graph(
    "label_graph.npz"
)

# JSON is a readable alternative for smaller graphs or inspection.
lit.save_label_graph("label_graph.json", neighbors, contacts=contacts)

Junction detection

Junction pixels are pixels whose 3×3 neighborhood contains at least three distinct labels. Connected junction pixels are clustered into Junction objects with subpixel (y, x) centroids and the set of labels that meet there.

junction_label_image, junctions = lit.junctions_from_labels(
    labels,
    background=None,
    min_labels=3,
    connectivity=2,
)

for junction in junctions:
    print(junction.id, junction.yx, sorted(junction.labels))

Graph-colored plotting

Plotting helpers return matplotlib objects and never call plt.show(), so they compose cleanly in notebooks.

fig, ax = lit.plot_label_image(
    labels,
    use_graph_coloring=True,
    K=8,
    seed=1,
    title="Graph-colored labels",
)

fig, ax = lit.plot_junctions(labels, junctions=junctions, ax=ax)

You can also use the lower-level coloring helper:

image, lut, ax = lit.show_map_with_colors(labels, K=8, seed=1)

Examples and cookbook

The examples/ directory contains script-style examples. Edit the constants at the top of each script, run it, or copy sections into a notebook.

python examples/01_graph_coloring.py
python examples/02_preprocessing_gallery.py
python examples/03_junctions_and_contours.py
python examples/04_graph_io.py

The cookbook walks through the same workflows and embeds the generated images in the GitHub/local documentation.

• Examples: https://github.com/zen-laboratory/labelimage-tools/tree/main/examples
• Cookbook: https://github.com/zen-laboratory/labelimage-tools/blob/main/docs/cookbook.md
• Generated example plots: https://github.com/zen-laboratory/labelimage-tools/tree/main/examples/plots

Essential processing outputs

The GitHub README and cookbook show rendered images for the graph-colored label image, detected junctions, and ordered contours. For PyPI/TestPyPI, those local relative image embeds are replaced with repository links so the long description renders safely.

Graph-colored label image using the cyclic managua colormap:

labels = lit.load_image_pipeline("samples/test_cells2D.tif")

fig, ax = lit.plot_label_image(
    labels,
    use_graph_coloring=True,
    K=8,
    seed=4,
    cmap="managua",
    cyclic_cmap=True,
    title="Graph-colored label image",
)

Detected junctions:

labels = lit.load_image_pipeline("samples/test_cells2D.tif")
junction_label_image, junctions = lit.junctions_from_labels(
    labels,
    background=0,
    min_labels=3,
    connectivity=2,
)
fig, ax = lit.plot_label_image(labels, cmap="managua", cyclic_cmap=True)
lit.plot_junctions(junctions=junctions, junction_mask=junction_label_image > 0, ax=ax)

Ordered contours:

labels = lit.load_image_pipeline("samples/test_cells2D.tif")
contours = lit.ordered_contours_from_labels(labels, background=0)
fig, ax = lit.plot_label_image(labels, cmap="managua", cyclic_cmap=True)
lit.plot_contours(labels, ax=ax, background=0, color="black", linewidth=0.6)