Alignment

The alignment module includes support for aligning projection data. The provided tools are:

Pre-alignment routines for tomographic data
Image stack shift finding

Pre-alignment

The class alignment.shifts.DetectorShiftsPRE offers support for both finding the vertical and horizontal shifts of tomographic projection data. They are exposed through the methods fit_v and fit_u.

class DetectorShiftsPRE(DetectorShiftsBase):
    """Compute the pre-alignment detector shifts for a given dataset."""

    def fit_v(
        self,
        use_derivative: bool = True,
        use_rfft: bool = True,
        normalize_fourier: bool = True,
    ) -> NDArrayFloat:
        ...

    def fit_u(
        self,
        fit_l1: bool = False,
        background: Union[float, NDArray, None] = None,
        method: str = "com",
    ) -> tuple[NDArrayFloat, float]:
        ...

The fit_v method computes the vertical shifts of the stack with 1D cross-correlations. The cross-correlation function is computed per angle on the intensity profile resulting from computing the integral of the projections along the U axis, and their derivative along the V axis.

The fit_u method computes the horizontal shifts of the stack, by computing the sinusoid that interpolates the chosen value of interest across all the rotation angles. The value of interest can include the center-of-mass (CoM) or the position of the highest intensity peak of the projections.

We present here an example of how to use the fit_u method to compute the horizontal shifts.

import corrct as cct

align_pre = cct.alignment.DetectorShiftsPRE(data_vwu, angles_rad)

diffs_u_pre, cor = align_pre.fit_u()

where the projection data is passed to the DetectorShiftsPRE class with the following axes order: [V], W, U, which means that V is the slowest varying axis, but also optional (in case of 2D data).

These shifts can be used to create a ProjectionGeometry object, which can be used to correct the projection data, when passed to projection operators as follows:

prj_geom = cct.models.get_prj_geom_parallel(geom_type="2d")
prj_geom.set_detector_shifts_vu(diffs_u_pre, cor)
vol_geom = cct.models.get_vol_geom_from_data(data_vwu)

solver = cct.solvers.SIRT()
with cct.projectors.ProjectorUncorrected(vol_geom, angles_rad, prj_geom=prj_geom) as A:
    rec_pre, _ = solver(A, data_test, iterations=100)

Image stack alignment

The alignment.shifts.DetectorShiftsXC.fit_vu_accum_drifts function calculates the shifts in the vertical and possibly horizontal directions of each image in a stack relative to a reference image or images. It ensures that the number of reference images matches the number of data images, and returns an array containing these shifts.