Estimator

Description

class corgihowfsc.sensing.Estimator.Estimator(name=None)[source]

Bases: object

Abstract estimator base class

estimate_efield(intensities, phases, imager, dmlist, lam_idx, crop, min_good_probes=2, eestclip=inf, eestcondlim=0)[source]

Estimates the electric field. Base arguments support Pair-Wise Probing method and perfect estimator method.

Parameters:

  • intensities (np.ndarray) – A 3D array of real-valued image data. The first element [0,:,:] is the initial intensity, and subsequent arrays are the positive and negative phase perturbations; ie 2N+1 arrays for N phase perturbations.

  • phases (np.ndarray) – a 3d array of phase data corresponding to the propagated perturbations through the optical system, in radians

  • min_good_probes (int, optional) – An integer >= 2 giving the number of probe intensity measurements that must be good at a pixel in order for the estimation to be done at that pixel. Given the number of unknowns in the matrix solve, 2 good measurements is mathematical minimum that can be accepted. Note that ‘good’ in this context means not in badpixels as True AND does not produce an estimate with negative intensity. No other data quality metric is applied. Defaults to 2.

  • eestclip (float, optional) – If the incoherent is less than -coherent*eestclip for an e-field array element, that element will be marked as bad. This permits incoherent estimates to be negative (due to e.g. read noise) but caps how negative the incoherent can be and how large the coherent can be. In particular, this mitigates spikes like the one seen in PFR 218555. Scalar value >= 0. Setting to infinity disables this check. Defaults to np.inf.

  • eestcondlim (float, optional) – The minimum condition number (ratio of smallest singular value to largest in the least-square solve) which can be accepted as a valid estimate. As the matrix is N x 2 for some N >= 2, there will only ever be two singular values. A poorly-conditioned solution is close to rank 1, which suggests the data is degenerate and not able to estimate independent real and imaginary parts of the data. >= 0. Setting to 0 disables this check. Defaults to 0.

  • imager (object) – Simulation object (e.g., GitlImage) capable of generating electric fields or images based on the optical model. Required for Perfect Estimator.

  • dmlist (list) – List of current DM command arrays [DM1, DM2] used to set the state of the optical model. Required for Perfect Estimator.

  • lam_idx (int) – Index of the current wavelength channel being processed. Required for Perfect Estimator.

  • crop (tuple) – 4-tuple of (lower row, lower col, number of rows, number of columns), indicating where in a clean frame a PSF is taken. All are integers; the first two must be >= 0 and the second two must be > 0. Required for Perfect Estimator.

Returns:

efield – A 2D array of complex electric field, with bad estimates as NaNs.

Return type:

np.ndarray

Raises:

NotImplementedError – This method is abstract and must be implemented by a subclass.