modifiedOtsu 0.1.2

Numba-accelerated local modified Otsu thresholding following Park et al.'s 2D+time formulation, with an optional 3D extension.

modifiedOtsu

modifiedOtsu implements the local modified Otsu thresholding step used by Park et al. in Segmentation-based tracking of macrophages in 2D+time microscopy movies inside a living animal.

The package default is window_size=(3, 3, 3) for volumetric use. For a 2D+time movie with shape (t, y, x), use window_size=(1, s, s) explicitly for the exact 2D local window used by Park et al. A larger first window dimension, such as (3, 7, 7), is a 3D/temporal extension, not the exact 2D local window from the paper.

The binary decision is:

binary = image > threshold_map

The contrast rejection rule is the relative class-mean rule:

abs(mu_foreground - mu_background) / abs(mu_background) > delta

If mu_background == 0 and the two means differ, the relative separation is considered infinite rather than rejecting the window.

Install

python -m pip install .

For TIFF I/O:

python -m pip install '.[tiff]'

Python API

import tifffile
from modifiedOtsu import getMask

movie = tifffile.imread("movie.tif")  # shape: (t, y, x)
thresholds, mask = getMask(movie, window_size=(1, 7, 7), delta=0.2)

Default volumetric call:

thresholds, mask = getMask(volume, delta=0.2)  # default window_size=(3, 3, 3)

Use delta=0 for ordinary local Otsu without the extra relative-contrast rejection.

For a 3D volume extension:

thresholds, mask = getMask(volume, window_size=(3, 7, 7), delta=0.2)

CLI

One TIFF/NumPy volume:

modified-otsu --input image.tif --window 3 3 3 --delta 0.2

Directory, recursively preserving subdirectories:

modified-otsu \
  --input dataset \
  --output-dir results \
  --recursive \
  --window 3 3 3 \
  --delta 0.2

With threshold maps:

modified-otsu --input image.tif --window 3 3 3 --save-thresholds

Masks are saved as TIFF files with values 0 and 255.