Writing 16-bit PNG files

[benjaminplanche]

Hi there,

Thanks to @krasin (PR #76), the module supports the reading of 16bit PNG files.
As far as I can tell, it is however not true yet for the writing.

I am alas not familiar with the generic/png.c code, but it seems that passed images are automatically converted to 8bit format before writing (cf. l.208 and l.269).

[danieleghisi]

Hi, I also have this very same issue.
Does anyone knows how to save a tensor as 16-bit image?

Thanks,
Daniele

[b4zz4]

For my also this is a problem, lose the transparencies of the images

[b4zz4]

convert in.png -background '#000000' -alpha remove -alpha off PNG24:in.flat.png
convert in.png matte:- | convert - -negate PNG24:in.mask.png
(process)
convert out.flat.png out.mask.png -alpha Off -compose CopyOpacity -composite out.png