estimate_flat_distortion does not work for all image sizes

Trying the following code with different image sizes:

from nabu.estimation.distortion import estimate_flat_distortion
from scipy.ndimage import zoom
from scipy.misc import face
tmp1 = zoom(face()[:,:,0],4)[:2500,:2500]
v = estimate_flat_distortion(tmp1, tmp1, tile_size=100, padding_mode='constant')

This gives an error:

---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
Input In [2], in <cell line: 5>()
      3 from scipy.misc import face
      4 tmp1 = zoom(face()[:,:,0],4)[:2500,:2500]
----> 5 v = estimate_flat_distortion(tmp1, tmp1, tile_size=100, padding_mode='constant')

File /mnt/multipath-shares/scisoft/users/favre/dev/pynx-py38-power9-env/lib/python3.8/site-packages/nabu/estimation/distortion.py:51, in estimate_flat_distortion(flat, image, tile_size, interpolation_kind, padding_mode, correction_spike_threshold, logger)
     49 for ic, c in enumerate(starts_c):
     50     try:
---> 51         coeff_v, coeff_h, shifts_vh_per_img = shift_finder.find_shift(
     52             np.array([image[r : r + tile_size, c : c + tile_size], flat[r : r + tile_size, c : c + tile_size]]),
     53             np.array([0, 1]),
     54             return_shifts=True,
     55             low_pass=(1.0, 0.3),
     56             high_pass=(tile_size, tile_size * 0.3),
     57         )
     58         cor1[ir, ic], cor2[ir, ic] = shifts_vh_per_img[1]
     60     except ValueError as e:

File /mnt/multipath-shares/scisoft/users/favre/dev/pynx-py38-power9-env/lib/python3.8/site-packages/nabu/estimation/translation.py:183, in DetectorTranslationAlongBeam.find_shift(self, img_stack, img_pos, roi_yxhw, median_filt_shape, padding_mode, peak_fit_radius, high_pass, low_pass, return_shifts, use_adjacent_imgs)
    180     plt.show(block=False)
    182 if return_shifts:
--> 183     return coeffs_v[1], coeffs_h[1], shifts_vh
    184 else:
    185     return coeffs_v[1], coeffs_h[1]

IndexError: index 1 is out of bounds for axis 0 with size 1

More generally with an image size of (n, n), it works until 2400 but fails after (no idea why).

(I also had an issue with an asymmetric size but can't reproduce it now..)