Mask Design

This module provides utilities to create different types of masks (CodedAperture, PhaseContour, FresnelZoneAperture) and simulate the corresponding PSF.

Abstract Mask Class

class lensless.hardware.mask.Mask(resolution, distance_sensor=None, size=None, feature_size=None, psf_wavelength=[4.6e-07, 5.5e-07, 6.4e-07], use_torch=False, torch_device='cpu', centered=True, **kwargs)

Parent Mask class. Attributes common to each type of mask.

__init__(resolution, distance_sensor=None, size=None, feature_size=None, psf_wavelength=[4.6e-07, 5.5e-07, 6.4e-07], use_torch=False, torch_device='cpu', centered=True, **kwargs)

Constructor from parameters of the user’s choice.

Parameters

• resolution (array_like) – Resolution of the mask (px).

• distance_sensor (float) – Distance between the mask and the sensor (m). Needed to simulate PSF.

• size (array_like) – Size of the mask (m). Only one of size or feature_size needs to be specified.

• feature_size (float or array_like) – Size of the feature (m). Only one of size or feature_size needs to be specified.

• psf_wavelength (list, optional) – List of wavelengths to simulate PSF (m). Default is [460e-9, 550e-9, 640e-9] nm (blue, green, red).

• use_torch (bool, optional) – If True, the mask is created as a torch tensor. Default is False.

• torch_device (str, optional) – Device to use for torch tensor. Default is ‘cpu’.

• centered (bool, optional) – If True, the mask is centered. Default is True.

compute_psf(distance_sensor=None, wavelength=None, intensity=True)

Compute the intensity PSF with bandlimited angular spectrum (BLAS) for each wavelength. Common to all types of masks.

Parameters

• distance_sensor (float, optional) – Distance between mask and sensor (m). Default is the distance specified at initialization.

• wavelength (float or array_like, optional) – Wavelength(s) to compute the PSF (m). Default is the list of wavelengths specified at initialization.

abstract create_mask()

Abstract mask creation method that creates mask with subclass-specific function.

classmethod from_sensor(sensor_name, downsample=None, **kwargs)

Constructor from an existing virtual sensor that copies over the sensor parameters (sensor resolution, sensor size, feature size).

Parameters

• sensor_name (str) – Name of the sensor. See SensorOptions.

• downsample (float, optional) – Downsampling factor.

• **kwargs – Additional arguments for the mask constructor. See the abstract class Mask and the corresponding subclass for more details.

Example

mask = CodedAperture.from_sensor(sensor_name=SensorOptions.RPI_HQ, downsample=8, ...)

height_map_to_field(wavelength, return_phase=False)

Compute phase from height map.

Parameters

• height_map (ndarray) – Height map.

• wavelength (float) – Wavelength of the light (m).

• return_phase (bool, optional) – If True, return the phase instead of the field. Default is False.

plot(ax=None, **kwargs)

Plot the mask.

Parameters

• ax (Axes, optional) – Axes to plot the mask on. Default is None.

• **kwargs – Additional arguments for the plot function.

Coded Aperture (FlatCam)

class lensless.hardware.mask.CodedAperture(method='MLS', n_bits=8, **kwargs)

Coded aperture mask as in FlatCam.

__init__(method='MLS', n_bits=8, **kwargs)

Coded aperture mask contructor (FlatCam).

Parameters

• method (str) – Pattern generation method (MURA or MLS). Default is MLS.

• n_bits (int, optional) – Number of bits for pattern generation, must be prime for MURA. Results in 2^n - 1``x``2^n - 1 for MLS. Default is 8 (for a 255x255 MLS mask).

• **kwargs – The keyword arguments are passed to the parent class Mask.

create_mask(row=None, col=None, mask=None)

Creating coded aperture mask.

generate_mura(p)

Generate MURA square pattern.

Parameters

p (int) – Number of bits.

get_conv_matrices(img_shape)

Get theoretical left and right convolution matrices for the separable mask.

Such that measurement model is given P @ img @ Q.T.

Parameters

img_shape (tuple) – Shape of the image to being convolved.

Returns

• P (ndarray) – Left convolution matrix.

• Q (ndarray) – Right convolution matrix.

is_prime(n)

Assess whether a number is prime or not.

Parameters

n (int) – The number we want to check.

simulate(obj, snr_db=20)

Simulate the mask measurement of an image. Apply left and right convolution matrices, add noise and return the measurement.

Parameters

• obj (ndarray) – Image to simulate.

• snr_db (float, optional) – Signal-to-noise ratio (dB) of the simulated measurement. Default is 20 dB.

Phase Contour (PhlatCam)

class lensless.hardware.mask.PhaseContour(noise_period=(16, 16), refractive_index=1.2, n_iter=10, design_wv=5.32e-07, **kwargs)

Phase contour mask as in PhlatCam.

__init__(noise_period=(16, 16), refractive_index=1.2, n_iter=10, design_wv=5.32e-07, **kwargs)

Phase contour mask contructor (PhlatCam).

Parameters

• noise_period (array_like, optional) – Noise period of the Perlin noise (px). Default is (8, 8).

• refractive_index (float, optional) – Refractive index of the mask substrate. Default is 1.2.

• n_iter (int, optional) – Number of iterations for the phase retrieval algorithm. Default is 10.

• design_wv (float, optional) – Wavelength used to design the mask (m). Default is 532e-9, as in the PhlatCam paper.

• **kwargs – The keyword arguments are passed to the parent class Mask.

create_mask()

Creating phase contour from edges of Perlin noise.

Fresnel Zone Aperture

class lensless.hardware.mask.FresnelZoneAperture(radius=0.00056, **kwargs)

Fresnel Zone Aperture (FZA) mask as in this work, namely binarized cosine function.

__init__(radius=0.00056, **kwargs)

Fresnel Zone Aperture mask contructor.

Parameters

• radius (float) – Radius of the FZA (m). Default value is 0.56e-3 (largest in the paper, others are 0.32e-3 and 0.25e-3).

• **kwargs – The keyword arguments are passed to the parent class Mask.

create_mask()

Creating binary Fresnel Zone Aperture mask.

Trainable Mask

class lensless.hardware.trainable_mask.TrainableMask(optimizer='Adam', lr=0.001, **kwargs)

Abstract class for defining trainable masks.

The following abstract methods need to be defined:

• get_psf: returning the PSF of the mask.

• project: projecting the mask parameters to a valid space (should be a subspace of [0,1]).

__init__(optimizer='Adam', lr=0.001, **kwargs)

Base constructor. Derived constructor may define new state variables

Parameters

• optimizer (str, optional) – Optimizer to use for updating the mask parameters, by default “Adam”

• lr (float, optional) – Learning rate for the mask parameters, by default 1e-3

abstract get_psf()

Abstract method for getting the PSF of the mask. Should be fully compatible with pytorch autograd.

Returns

The PSF of the mask.

Return type

Tensor

abstract project()

Abstract method for projecting the mask parameters to a valid space (should be a subspace of [0,1]).

update_mask()

Update the mask parameters. According to externaly updated gradiants.

class lensless.hardware.trainable_mask.TrainablePSF(initial_psf, grayscale=False, **kwargs)

Class for defining an object that directly optimizes the PSF, without any constraints on what can be realized physically.

Parameters

grayscale (bool, optional) – Whether mask should be returned as grayscale when calling get_psf. Otherwise PSF will be returned as RGB. By default False.

__init__(initial_psf, grayscale=False, **kwargs)

Base constructor. Derived constructor may define new state variables

Parameters

• optimizer (str, optional) – Optimizer to use for updating the mask parameters, by default “Adam”

• lr (float, optional) – Learning rate for the mask parameters, by default 1e-3

get_psf()

Abstract method for getting the PSF of the mask. Should be fully compatible with pytorch autograd.

Returns

The PSF of the mask.

Return type

Tensor

project()

Abstract method for projecting the mask parameters to a valid space (should be a subspace of [0,1]).