prysm.detector#

Detector-related simulations.

prysm.detector.apply_lut(img, lut)#

Apply a lookup table to img.

Parameters:

  • img (ndarray) – n dimensional array (2D and 3D are both OK) of an unsigned integer dtype

  • lut (ndarray) – 1 dimensional array whose indices are input values and values are output values

Returns:

ndarray of the same shape as img the output array must not be modified in place, or lut will be modified as well.

Return type:

ndarray

class prysm.detector.Detector(dark_current, read_noise, bias, fwc, conversion_gain, bits, exposure_time, prnu=None, dcnu=None, lut=None)#

Bases: object

Model of a detector.

Initialize a new camera model.

Parameters:

  • dark_current (float) – e-/sec, charge accumulated with no light reaching the sensor.

  • read_noise (float) – e-, random gaussian noise associated with readout

  • bias (float) – e-, uniform value added to readout to avoid negative numbers

  • fwc (float) – e-, maximum number of electrons that can be held by one pixel

  • conversion_gain (float) – e-/DN gain converting e- to DN,

  • bits (int) – number of bits for the ADC, multiples of 2 in 8..16 are contemporary

  • exposure_time (float) – exposure time, seconds

  • prnu (ndarray, optional) – relative pixel response nonuiformity, a fixed map that the input field is multiplied by. ones_like is perfectly uniform.

  • dcnu (ndarray, optional) – dark current nonuniformity, a fixed map that the dark current is multiplied by. ones_like is perfectly uniform.

  • lut (ndarray, optional) – look-up table of ideal output DN values to output DN values, representing the nonlinearity of the detector

expose(aerial_img, frames=1)#

Form an exposure of an aerial image.

Parameters:

  • aerial_img (ndarray) – aerial image, with units of e-/sec. Should include any QE as part of its Z scaling

  • frames (int) – number of images to expose, > 1 is functionally equivalent to calling with frames=1 in a loop, but the random values are all drawn at once which can much improve performance in GPU-based modeling.

Returns:

of shape (frames, aerial_img.shape), if frames=1 the first dim is squeezed, and output shape is same as input shape. dtype=uint8 if nbits <= 8, else uint16 for <= 16, etc not scaled to fill containers, i.e. a 12-bit image will have peak DN of 4095 in a container that can reach 65535.

has units of DN.

Return type:

ndarray

prysm.detector.olpf_ft(fx, fy, width_x, width_y)#

Analytic FT of an optical low-pass filter, two or four pole.

Parameters:

  • fx (ndarray) – x spatial frequency, in cycles per micron

  • fy (ndarray) – y spatial frequency, in cycles per micron

  • width_x (float) – x diameter of the pixel, in microns

  • width_y (float) – y diameter of the pixel, in microns

Returns:

FT of the OLPF

Return type:

ndarray

prysm.detector.pixel_ft(fx, fy, width_x, width_y)#

Analytic FT of a rectangular pixel aperture.

Parameters:

  • fx (ndarray) – x spatial frequency, in cycles per micron

  • fy (ndarray) – y spatial frequency, in cycles per micron

  • width_x (float) – x diameter of the pixel, in microns

  • width_y (float) – y diameter of the pixel, in microns

Returns:

FT of the pixel

Return type:

ndarray

prysm.detector.pixel(x, y, width_x, width_y)#

Spatial representation of a pixel.

Parameters:

  • x (ndarray) – x coordinates

  • y (ndarray) – y coordinates

  • width_x (float) – x diameter of the pixel, in microns

  • width_y (float) – y diameter of the pixel, in microns

Returns:

spatial representation of the pixel

Return type:

ndarray

prysm.detector.bindown(array, factor, mode='avg')#

Bin (resample) an array.

Parameters:

  • array (ndarray) – array of values

  • factor (int or sequence of int) – binning factor. If an integer, broadcast to each axis of array, else unique factors may be used for each axis.

  • mode (str, {'avg', 'sum'}) – sum or avg, how to adjust the output signal

Returns:

ndarray binned by given number of samples

Return type:

ndarray

Notes

For each axis of array, shape must be an integer multiple of factor.

array may be ND, a scalar factor will broadcast to all dimensions.

To bin an image cube e.g. of shape (3, m, n), use bindown(img, [1, factor, factor])

Raises:

ValueError – invalid mode

prysm.detector.tile(array, factor, scaling='sum')#

Tile (repeat) an array by factor.

Parameters:

  • array (ndarray) – array of values

  • factor (int or sequence of int) – binning factor. If an integer, broadcast to each axis of array, else unique factors may be used for each axis.

  • scaling (str, {'avg', 'sum'}) – sum or avg, how to adjust the output signal

Returns:

ndarray binned by given number of samples

Return type:

ndarray

Notes

This function is the adjoint operation for bindown.

It works with ND arrays, with the same rules as bindown.

In 2D, it is equivalent to array.repeat(factor, axis=0).repeat(factor, axis=1) (and is more generally equivalent for higher dimensionality, but it runs about ndim times faster (twice as fast in 2D, 3x in 3D, etc).

The return may be a view into the argument and is mutated after calling tile at the user’s risk