TIF to INR seems wrong!

[jlegrand62]

There is a problem with TIF reader or INR writer: the resulting image obtained by the MWE is shifted along the Y axis...

• Differences snapshot: https://mycore.core-cloud.net/public.php?service=files&t=7ebd330fb28b7710a3ec7e24574ba6d5
• Example tif image: https://mycore.core-cloud.net/public.php?service=files&t=2fc0a1372513fb051eeca4e699bad9b5
• Obtained inr image: https://mycore.core-cloud.net/public.php?service=files&t=4bb3760afbcb3db669e35d6fa0192f39
• MWE: test_tif_convert_to_inr.py python script
  Use as follow: python test_tif_convert_to_inr.py -i test.tif

[jlegrand62]

I forgot to mention that the differences have been observed using FIJI software...
And to add the link for the MWE script: https://mycore.core-cloud.net/public.php?service=files&t=7a67b2e67d214dd863296e27b244070a

[sribes]

Hi,
Could you tell me:

• where is the script "test_tif_convert_to_inr.py" ?
  I did not find this script in the folder openalea/components...
  The given links do not work, maybe a mistake ?
  Cheers,

[jlegrand62]

Indeed it seems that clicking on the links do not work, but copy/paste them is working for me...
Sorry I did not check the links where working properly!
Cheers.

[sribes]

To be sure, you can:
1/ Open the images using Fiji and compare the main metadata (bits, pixeltype, size, type, voxelsize) by clicking on image/showInfo
2/ Load the images in memory and launch the numpy function: assert_almost_equal (see https://docs.scipy.org/doc/numpy-1.10.0/reference/generated/numpy.testing.assert_almost_equal.html)
Since the script is not in openalea/components, maybe it is not the right place to post comments (moderator ?)...
Cheers,

[jlegrand62]

I know how to use Fiji (even though it can be misleading sometimes) and compare images.
I posted it here because the imread and imsave functions used to read and save images are from this package: from openalea.image.serial.basics import imread, imsave.

All the MWE do is to read an image (tif), and immediately write it (inr.gz) with no modification.
And this is the result in Fiji (left: original - right read/write image):

So yeah, this is wrong...
Jo

[sribes]

The tiff reader uses libtiff whereas the .inr writer uses the stream. So if we can determine if errors come from read or write (or both) process, it can be useful. To do this, you should compare metadata and compare objects in memory.
Cheers,

[sribes]

Also, if you can try to write an .inr instead of a .inr.gz