Preprocessing features for re-saving mcorr results in C order

[ethanbb]

This feature adds some new parameters that can be specified under the "preprocessing" top-level key of params (rather than "main") to do extra steps while re-saving the movie in C-order. (This is distinct from the "main.preprocess" namespace for the preprocessing done in CNMF.fit - I couldn't think of a better name, but we could potentially still change it to avoid confusion.

These new parameters are supported under "preprocessing":

• border_pixels: controls setting border pixels to 0 (as in the border_to_0 argument to save_memmap). Value can be:
  • An int, to set this many pixels on each side to 0
  • A dict with int entries left, right, top, and bottom, plus z_top and z_bottom for 3D movies, for unequal borders
  • None, to infer borders directly from motion correction results (only supported if input is an mcorr mesmerize item)
• add_to_movie: constant to add to the movie (default is 0.0001, matching save_memmap)
• highpass_cutoff and highpass_order: Set highpass_cutoff to a float > 0 to high-pass filter the data during transposition to C-order. This is implemented using a zero-phase Butterworth filter of order highpass_order (default 4).
  • Related to discussion Getting denoised and deconvolved traces that may be pathological flatironinstitute/CaImAn#1516, this may be a useful solution if there is a baseline drift (variable across space) that you want to remove, although I now believe that it may be better to try to account for the drift using more background components instead (especially if the source of drift is multiplicative rather than additive).