spectral_unmixing.correct_intra_stack_z_drift

spectral_unmixing.correct_intra_stack_z_drift(stack, *, registration_channel=0, method='phase_cross_correlation', reference_mode='neighbor', neighbor_window_size=3, pre_median_filter=False, post_median_filter=False, median_kernel_size=3, verbose=True)[source]

Correct XY drift between Z slices within each time point of a TZCYX stack.

Parameters:

  • stack (array-like) – Input stack in canonical TZCYX order.

  • registration_channel (int, optional) – Channel used to estimate the slice-wise XY shifts. The computed shifts are then applied to all channels of the corresponding Z slice.

  • method ({“phase_cross_correlation”, “pystackreg”}, optional) – Backend used for shift estimation.

  • reference_mode ({“neighbor”, “full_projection”}, optional) – Strategy used to build the per-slice reference image. "neighbor" uses a local max projection around each slice. "full_projection" uses the max projection across the entire Z stack of the current time point.

  • neighbor_window_size (int, optional) – Odd number of slices used for reference_mode="neighbor". 3 means z-1, z, z+1; 5 means z-2 through z+2.

  • pre_median_filter (bool, optional) – If True, apply a slice-wise median filter to the registration channel before building the reference images. This affects only shift estimation.

  • post_median_filter (bool, optional) – If True, apply a 2D median filter to both the moving slice and the reference image just before shift estimation.

  • median_kernel_size (int, optional) – Median filter kernel size used by the optional pre/post filters.

  • verbose (bool, optional) – If True, print the estimated shifts line-wise for each t and z.

Return type:

ndarray

Returns:

np.ndarray – Z-drift-corrected stack with the same TZCYX shape as the input.

Notes

This function estimates XY shifts independently for each Z slice within each time point. The shifts are computed from a user-selected registration channel, but are applied to all channels of the affected slice.