Legacy ImageMagick Discussions Archive

Hi all,

The premise is that I'm trying to batch resize a bunch of 32bit png-s with transparency. The 3rd party tool that I'm using to load the result images complains for some of them that it doesn't support non-8-bit Greyscale.

Quite clearly, neither the original, nor the resulting images are greyscale. The command I'm using is:
The result that identify yields is this (excerpt):

calendar_day_future_dark.png PNG 119x107 119x107+0+0 8-bit DirectClass 352B 0.000u 0:00.000 (sRGB / PaletteAlpha / 8-bit)
calendar_day_future_light.png[1] PNG 119x107 119x107+0+0 8-bit DirectClass 352B 0.000u 0:00.000 (sRGB / PaletteAlpha / 8-bit)
calendar_day_past_dark.png[2] PNG 119x107 119x107+0+0 8-bit PseudoClass 2c 339B 0.000u 0:00.000 (sRGB / Palette / 8-bit)
calendar_day_past_light.png[3] PNG 119x107 119x107+0+0 8-bit PseudoClass 2c 339B 0.000u 0:00.000 (sRGB / Palette / 8/4-bit)
calendar_day_present.png[4] PNG 119x107 119x107+0+0 8-bit PseudoClass 2c 327B 0.000u 0:00.000 (Gray / Bilevel / 8/1-bit)

Now, the problem is that out of these five images the first four load fine, and it is the last one that fails (the originals all load fine). "calendar_day_present.png" is actually a solid white rectangle. (Windows 7 Explorer's Bit Depth column reports a bit depth of 1 on all of these files; the originals had 32 - I'm not sure how accurate this feature is, but most of my 24/32 bit originals went to 1/4/8 bits according to it.)

How can I make sure that each image has the same Colourspace as its original when using mogrify, i.e. they're not converted into greyscale / palettised images and the alpha channel isn't lost?

Thank you for your help in advance.

Use the PNG32 subformat to force all PNGs to be written as 8-bit-per-sample RGBA

e.g.,

And when I've made it as far as spotting that it's a colorspace issue...

http://www.imagemagick.org/www/command- ... colorspace

@glennrp: thank you, I'll try that!

I revised my response to demonstrate "mogrify" instead of "convert", see above.

You could also try or if it's only the bit depth or colortype that is problematic. That will probably
result in much smaller files than forcing png32 RGBA output.

@glennrp: thank you so much for your help -- that did the trick (the -colorspace switch still doesn't seem to work).