RuntimeError: filter weights array has incorrect shape.

[JuergenTerpe]

I'm a newbie in python programming, so maybe I'm doing something wrong. However, there are only limited parameters for these functions. I'm trying this:

self.imageData = demosaicing_CFA_Bayer_bilinear(self.imageData, self.fits_header['BAYERPAT'])

self.imageData is a numpy array from a fits image (CFA RGGB) and it is in linear state, the numpy array is of type float32. I tried all functions and always get this error. I don't see any filter weights array to be passed as parameter. What is my mistake?

Thanks in advance,
Juergen

[KelSolaar]

Hi Juergen,

Is it possible to print the value of self.fits_header['BAYERPAT'] and the shape of self.imageData?

Cheers,

Thomas

[JuergenTerpe]

Hi Thomas,

Sorry for the late answer. Sure it is possible:

RGGB
(4168, 6224, 1)

It's a normal not debayered image from my astro camera, which shoots using RGGB bayer pattern and has a resolution of 4168x6224 pixels. I think the third shape value is the number of color values and this would be the expected value 1, because it's not debayered, so it's still a grayscale image. Please let me know which other information you require. It would be really helpful to have a chance to debayer these images within my tool I want to make public for the astrophotography community.

Regards,
Jürgen

[KelSolaar]

Ah!

This is something that we could/should handle better, try that:

self.imageData = demosaicing_CFA_Bayer_bilinear(colour.utilities.tstack([self.imageData, self.imageData, self.imageData]), self.fits_header['BAYERPAT'])

The issue is that we are expecting a 3 channel image even though it is grayscale. I think that should fix your issue!

Cheers,

Thomas

[JuergenTerpe]

Thanks, Thomas.

Unfortunately, it still crashes with the same error.

File "/Users/jt/PycharmProjects/ImageSessionAnalysis/FitsImage.py", line 76, in createDisplayImage
self.imageData = demosaicing_CFA_Bayer_bilinear(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/jt/PycharmProjects/ImageSessionAnalysis/venv/lib/python3.11/site-packages/colour_demosaicing/bayer/demosaicing/bilinear.py", line 130, in demosaicing_CFA_Bayer_bilinear
R = convolve(CFA * R_m, H_RB)
^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/jt/PycharmProjects/ImageSessionAnalysis/venv/lib/python3.11/site-packages/scipy/ndimage/_filters.py", line 917, in convolve
return _correlate_or_convolve(input, weights, output, mode, cval,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/jt/PycharmProjects/ImageSessionAnalysis/venv/lib/python3.11/site-packages/scipy/ndimage/_filters.py", line 715, in _correlate_or_convolve
raise RuntimeError('filter weights array has incorrect shape.')
RuntimeError: filter weights array has incorrect shape.

Might be something unexpected deeper inside. I can upload an image if you are interested.

Regards,
Jürgen

[KelSolaar]

Sorry, it is me being stupid, we want the opposite, i.e. no extra dim! Try the following:

self.imageData = demosaicing_CFA_Bayer_bilinear(np.squeeze(self.imageData), self.fits_header['BAYERPAT'])

[JuergenTerpe]

Thank you, Thomas!

Now it works!!

[KelSolaar]

Awesome, I just updated the code in develop so that we squeeze systematically. Should avoid issues for future users!