AxisError when processing BMP grayscale images with shape (H, W)

[arsalanmstn]

Hi, I ran into an issue when processing grayscale (monochrome) BMP images of the second deployment of Spartacus in Trondheim fjord.

In ImagePrep class of the pyopia.instrument.silcam module, the "im_minimum" is computed using:
data['im_minimum'] = np.min(image, axis=2)
but this assumes that all images have a channel dimension (axis=2), which is true for RGB images with shape (H, W, 1 or 3). However, this raises the following error for grayscale images with shape (H, W):
"AxisError: axis 2 is out of bounds for array of dimension 2"

data['im_minimum'] is computed unconditionally in ImagePrep and cannot be disabled via user preferences, so processing monochrome images breaks at this step if they don't have channel dimension. The following correction can easily solve the problem:
data['im_minimum'] = image if image.ndim==2 else np.min(image, axis=2)

** Another approach is to comment out the imageprep step in the config file, but the user must set segment_source= "im_corrected" or segment_source="imraw". This can be confusing for users.......

[emlynjdavies]

I think doing something with this type of logic that you suggest is sensible:
data['im_minimum'] = image if image.ndim==2 else np.min(image, axis=2)
However, i would recommend basing this logic on explicit image types like SilCamLoad.load_image() here does, as bayer and mono images will may the same dimensions, for example.