Images stored upside down

[letmaik]

I find it extremely unintuitive that imageio (through freeimage) expects image arrays to be upside down to be saved correctly. Virtually every other library does it the other way. There should be a big fat warning in the imageio documentation that I have to flip my image. Ideally though, imageio would do the flipping for me (when reading or writing using freeimage). Or did I miss something fundamental? Btw, I'm writing 16bit tiff files.

Quoting freeimage docs:

In FreeImage, FIBITMAP are based on a coordinate system that is upside down
relative to usual graphics conventions. Thus, the scanlines are stored upside down,
with the first scan in memory being the bottommost scan in the image.

EDIT: Just noticed that I actually have to flip both axes, so img[::-1,::-1], really confusing.

EDIT2: Seems to be the same issue as in scikit-image/scikit-image#1101

[almarklein]

I find it /extremely/ unintuitive that imageio (through freeimage)
expects image arrays to be upside down to be saved correctly. Virtually
every other library does it the other way. There should be a big fat
warning in the imageio documentation that I have to flip my image.
Ideally though, imageio would do the flipping for me (when reading or
writing using freeimage). Or did I miss something fundamental? Btw, I'm
writing 16bit tiff files.

AFAIK imageio saves the images in the same "orientation" as e.g.
matplotlib and Matlab do. What is your experience with these?

I was not aware that most other toolkits do it differently. If this is
the case, or if indeed its confusing, we should add this information to
the docs.

• Almar

[letmaik]

Please have a look at scikit-image/scikit-image#1101 I think this is the same issue and really a bug.

[almarklein]

Ah, I did not understand that this was specific to 16bits images. I backported Stefan's Fix. Can you try and see whether it works?

[letmaik]

Yep it's working.

[almarklein]

Great!