ImageMagick x64 , Q8 for windows

[picscout108]

Hello ,
I couldn’t find a version for : ImageMagick-6.6.0-8-Q8-windows-x64-static.exe x64 static at 8 bits-per-pixel
Is there a wizard configuration for this ? where can I get relevant binaries?

Thanks,
Simon.

[magick]

We do not provide a Q8 version of ImageMagick for Windows, only a Q16. You of course can build ImageMagick yourself from source at Q8 for Windows 64.

[picscout108]

Thank you for your reply.

[OMA]

Sorry to bring up an old post (came from a Google search) but, why aren't you offering a x64 version of the Q8 version for Windows? Q8 is faster and suitable for most general uses. Why no 64 bit version? I suppose the 64 bit version has better performance in a 64 bit OS, doesn't it?

BTW, from the available options, which would be faster? 32-bit Q8 or 64-bit Q16?

[magick]

We already providing 7 different versions of ImageMagick Windows binaries which typically takes upwards of 4 hours to produce. We are not willing to support additional configurations. For out-of-band configurations, you can download the ImageMagick source and build with Visual C++.

[OMA]

Thanks for answering. What about the other question? ("from the available options, which would be faster? 32-bit Q8 or 64-bit Q16?")
Regards.

[magick]

32-bit Q8 would most likely be faster simply because there is a smaller memory requirement. However, we have not formally benchmarked recently to confirm our conjecture.

[OMA]

32-bit Q8 would most likely be faster simply because there is a smaller memory requirement. However, we have not formally benchmarked recently to confirm our conjecture.

Ok, I'm using this for rather low quality images, so I suppose then I'm better off with the 32-bit Q8 version even in a 64 bit system.
Thanks for answering!

[anthony]

Just a word of warning.

Q8 versions of ImageMagick have a much smaller memory footprint per image, but that is at a cost in terms of the number of significant bits during processing.

In Q8 resizing image will not be a exact (but it isn't too bad), and cropping, padding, drawing, composition (overlaying), masking, and simple recoloring and tinting, etc. etc., all work reasonably well.

However the more heavily processed the image is, the worse the errors from this lack of significant bits becomes. heavy coloring multiple times, resizing multiple times, distorting images, Background removal with anti-aliasing, Removing logos from images, panorama processing, and so on may start showing problems and artefacts. This is why Q16 is standard.

Even Q16 breakdown with significant errors for extreme image processing such as Fourier Transforms. Also for light/dark shade composition handling (HDRI), is an area where Q16 can have some problems (though still works rather well). As such the next level is Q32 or better still a HDRI (floating point) version of ImageMagick.