remove_background gives unexpected errors about kikuchipy HOT 6 CLOSED

Will look into this right away after making sure we can load HDF5 files with kp.load().

from kikuchipy.

• ii. Solution that seems to work is:

            if (omd.deadpixels_corrected and omd.deadpixels.any() and
                    omd.deadvalue.any()):

• iii. Also, a runtime error is encountered in the rescale:

RuntimeWarning: overflow encountered in byte_scalars
scale = float(omax / (pattern.max() - imin))

from kikuchipy.

i. Raising an error is the correct response since no background image is provided, however the error message could be clearer. I changed from except ValueError: to except (ValueError, AttributeError): so that the custom error message "No background image provided" is printed also when the original_metadata has no General.original_filepath parameter. Note that this error message will be raised when this is the case, regardless of which loading function (HyperSpy's or KikuchiPy's) is used.

ii. As you say, changing to omd.deadpixels.any() fixes the problem! There is no need to add it to a string though, i.e. omd.deadvalue.

iii. Pretty sure the problem is a too narrow data type (np.int8, [-128, 127]). If so, this warning is raised when the subtraction of a static background from a pattern yields some pixels with negative values below -128. This is prevented by setting both patterns data types to np.int16, [-32 768, 32 767], in kikuchipy/utils/expt_utils.py here:

def correct_background([...]):
[...]
    if static:
        # Change data types to avoid negative intensities in subtraction
        dtype = np.int16
        pattern = pattern.astype(dtype)
        bg = bg.astype(dtype)
[...]

Could you see if this fixes the problem? I cannot remember if correct_background() leaks much memory, I think not, but if so raising to 16-bit will make thinks worse... Will check this myself if this solves the problem.

from kikuchipy.

I see that the comment explaining the code (# Change data types...) is wrong, will remove this when pushing fixes to the problems.

from kikuchipy.

Have you checked my solution to problem iii?

from kikuchipy.

Solved in PR #28.

from kikuchipy.