[argyllcms] Printtarg dithering for TIFF output request

  • From: Nikolay Pokhilchenko <nikolay_po@xxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Sat, 07 Apr 2012 00:37:53 +0400

Hello Graeme!
I'm interesting in the dithering option for printtarg TIFF output. The option 
is needed for 8-bpp devices and proposed to obtaining even and tolerant 
measurements for printing workflows with dithering by default.
At a moment, my generic workflow for target creatiin consist of:

1) generating the target data with preliminary (from similar device) 
ICC-profile;
2) generating 16-bit TIFF target images;
3) changing the "color depth" of images from 16-bit to 8-bit in Adobe Photoshop 
Elements (the dithering is applying in Photoshop by default);
4) saving 8 bit per channel targets.

This workflow allows me to obtain the best target color rendering in darkest 
and lightest areas of device gamut. This is important for not linearized 
devices with native 24bpp RGB image data path (the devices with Windows GDI 
drivers). In some not very rare cases the device behavior in highlights or 
darkest areas is very steep. Without the dithering there will be noticeable 
posterizing or banding. So, the most of modern color-managed printing programs 
performs the dithering after the conversion by device ICC profile but before 
sending the job to the printer.
In case of steep device response and if I have preliminary device profile I can 
render the problematic color areas more evenly with dithering and without using 
"-Q8" in printtarg. Often I have no reliable information about printing 
workflow and know only that the device driver is 24-bpp RGB. So I should 
provide 8 bit per channel target images with dithering to measure practice 
device behavior smoothly.
I've tried to wrote a simple script using ImageMagic (binary for MS Windows) 
for dithering. But I've didn't fine how to apply error-diffusion dithering 
correctly. It's just didn't work. I find a way to apply ordered dither by 
ImageMagic. But in case of ordered dither there may be aliasing with regular 
target structure (I saw it on a targets with relative low resolution, about 
180dpi). I had to process target images "by hand" in Adobe Photoshop Elements 
because the last performs dithering but have very little automation 
capabilities.

P.S. Thank You for the answer my question about targen filtering option,
http://www.freelists.org/post/argyllcms/targen-samples-filter-out-bug-v135,1
I suppose the clarification "Note that some classes of the full spread patches 
don't support generating the total number of patches while being filtered - 
instead the points are generated then filtered" would be appropriate in targen 
-F option description.

Other related posts: