A coherent propagation and diffraction simulation tool, all written in Python.
In this package, a set of utilities are provided to simulate coherent signal propagation. It is particularly made for free-space optical propagation, diffraction, and holography. However, the tools are compatible with 1D and 2D data structures and can be potentially used for any sort of spatially-coherent signals.
A set of modules are provided to solve the scalar diffraction problem. The codes are GPU-friendly, compatible with well-known computation libraries (Numpy, TensorFlow) and they support batch processing. Using the TensorFlow backend, computations become autograd-compatible and could be mounted on machine learning models. Angular spectrum algorithm is the primary work horse for field propagation though other custom solvers could be used. Aside the built-in Numpy and TensorFlow backends, any computational means could be employed to process tensor operations.
It also includes:
- a simple yet useful data pipeline to load and standardize data. For now, it only supports images.
- Gerchberg-Saxton multi-distance phase recovery algorithm. It can be easily tweaked to support other variations of signal e.g. by wavelength.
To install the package, run:
python -m pip install fringe
Fringe requires numpy, tensorflow 2.x, and scikit_image.
The example files are not included in the package and should be downloaded separately. Also they require matplotlib to show plots.
- Import or create data
For images:
import numpy as np
from fringe.utils.io import import_image
from fringe.utils.modifiers import ImageToArray, Normalize, MakeComplexImages need to be standardized, normalized, and casted to complex data type. Modifiers are tools made for this purpose which apply these operations on import.
p1 = ImageToArray(bit_depth=16, channel='gray', crop_window=None, dtype='float32')
p2 = Normalize(background=np.ones((512, 512)))
p3 = MakeComplex(set_as='amplitude', phase=0)
obj = import_image("images/squares.png", preprocessor=[p1, p2, p3])- Propagate
Solvers contain propagation algorithms and can be called by solver.solve. In particular, angular Spectrum algorithm convolves the input field with a free-space propagtor function which depends on wavelength λ (or wavenumber k=2π/λ) and distance z.
from numpy import pi, abs, angle
from fringe.solvers.AngularSpectrum import AngularSpectrumSolver as AsSolver
solver = AsSolver(shape=obj.shape, dr=1, is_batched=False, padding="same", pad_fill_value=0, backend="Numpy")
rec = solver.solve(hologram, k=2*pi/500e-3, z=-1000)
amp, phase = abs(rec), angle(rec)
ax = pyplot.sublots(2, 1, 1)
ax[0].imshow(abs(obj))
ax[1].imshow(amp)
pyplot.show()
Example notebooks provide further details with 1D and 2D diffraction, GPU acceleration, batch processing, and phase recovery.
Fringe is released under the MIT license. See LICENSE for details.