fpm-py

Fourier ptychography is a computational image reconstruction technique that allows one high-resolution microscopic image to be acquired by combining multiple lower-resolution images captured at differing optical illumination angles.

More information about Fourier ptychography can be found here.

This library implements the Fourier ptychography technique in python to allow continued research and development. Existing implementations are undocumented, unreadable, and in MATLAB 😦.

Documentation is available at: https://rspcunningham.github.io/fpm-py-docs/

Structure

FPM-PY/
├── .github/
│   └── workflows/
│       └── publish.yml
├── .pytest_cache/
├── datasets/
│   └── example.pt
├── fpm_py/
│   ├── __pycache__/
│   ├── __init__.py
│   ├── algorithm.py
│   ├── data.py
│   ├── iteration_terminators.py
│   ├── optimizers.py
│   └── utils.py
├── tests/
│   ├── __pycache__/
│   ├── __init__.py
│   └── test_utils.py
├── .gitignore
├── example.py
├── LICENSE
├── poetry.lock
├── pyproject.toml
└── README.md

Example usage can be found in example.py and below under the 'Usage' heading.

The main module source is located under /fpm_py:

• algorithm.py: contains the core fpm algorithm function.
• data.py: contains dataclasses for ImageCapture and ImageSeries, the core datastructures used by fpm.
  • ImageCapture: a single image and its associated k-space vector [k_x, k_y].
  • ImageSeries: a complete set of ImageCaptures (the stack), and capture series metadata needed for reconstruction. When creating a series, only (optical_magnification and sensor pixel_size) or effective_magnification are needed. Device is automatically assigned by pytorch following the heirarchy: cuda > mps > cpu.
  • Functions to load a dataset (ie an ImageSeries) or save one to the disk.
• iteration_terminators.py: functions that return True when iteration should be stopped.
• optimizers.py: functions that update the object and pupil using the updated wavefunction.
• utils.py: some helpful functions and abstractions.

The dataset in /datasets/example.pt is for a USAF bar pattern. It is saved as a pickled dictionary.

Test cases for the module are under /tests --> these need to be worked on, lol.

Installation

To use in your project: pip install fpm-py

Usage

See example usage in example.py, and below:

import fpm_py as fpm
import matplotlib.pyplot as plt

# load example dataset
dataset = fpm.ImageSeries.from_dict("datasets/example.pt")

# reconstruct the object
output = fpm.reconstruct(dataset)

# plot the output
plt.imshow(output.abs().cpu().numpy(), cmap="gray")
plt.show()

Upcoming

1. Performance metric(s), to assess how well the resulting reconstruction worked.
2. Public 10k+ image dataset