Wrong value of single pixels when reading some TIFF images

[mraschi]

Hello,

I found a problem reading LZW-compressed TIFF images (MetaMorph) with BioFormats in Matlab. Around 1% of my image set have one or more pixels with a wrong value (as checked against Matlab's imread() and another image reader).

You can find examples of images with 1, 2, and 3 wrong pixel values here:
https://www.dropbox.com/s/p8ombbmk215d8 ... 1.tif?dl=0
https://www.dropbox.com/s/g9m315cjau3y8 ... 1.tif?dl=0
https://www.dropbox.com/s/gv0yx0r30ea2j ... 1.tif?dl=0

The images are grayscale, 16-bit depth, 1392x1040, and the pixels with the wrong values are always on the right side of the image (x=1392).

I am using bio-formats 5.0.6 and java version "1.7.0_21", 64-Bit Server VM (build 23.21-b01, mixed mode)

Thanks,
Marcelo

[mraschi]

I think this bug is related to https://github.com/openmicroscopy/bioformats/pull/1200

In this case, it seems that the fix was added to 5.1.0-m1. I will check reading the images using that version.

Marcelo

[mlinkert]

Hi Marcelo,

Thank you for the bug report. https://github.com/openmicroscopy/bioformats/pull/1200 is related, but as it was also included in 5.0.3 and later, that does not fully solve the problem. We are reviewing one more set of changes which should fully solve this:

https://github.com/openmicroscopy/bioformats/pull/1497

I would expect these changes to be included in the 5.1.0 release due early next year.

-Melissa