[reportlab-users] Transparent png

Tim Roberts timr at probo.com
Tue May 23 15:46:32 EDT 2017

Glenn Linderman wrote:
> However, the concept seems extremely straightforward for printing:
> transparent means "don't print there".  This seems to be exactly the
> same concept as stripping images for the 4-color printing process:
> color images are reduced to "print density of each of CYMK" for each
> color for each pixel, and the 0 density is (effectively) transparent.

Does it?  Does it mean "deposit no ink", or does it mean "deposit white

The problem is not dissimilar to doing projection rendering in a 3D
graphics application.  To determine the color of a pixel, if there is no
transparency or alpha, all I have to look at is the top-most object at
this location.  If there is transparency or alpha, I have to look at all
of the objects to blend them.

> Could you elaborate on the issues that professional print shops have
> with transparency? It doesn't seem to be that they can't produce a
> density of 0 for swath of certain (even all) colors in areas of the
> printed page: except for prints consisting totally of color pictures,
> nearly all pages in all publications have areas where no ink of any
> color is laid on the paper.  Is it because the concept of transparency
> in image files (particularly, those embedded in Postscript or PDF)
> hasn't been sufficiently standardized? Or what?

You're assuming all commercial printers use a 4-color process.  4-color
is only used in lower-end print jobs, and when you move to spot colors
it gets way more complicated.  Transparency confuses overprinting and
knockouts.  If they have a sophisticated rasterizer, it should be able
to come up with something close, but in high-end printing, close is not
good enough.  The production printing company I work with will not
accept PDF files that include images with transparency.  Their preflight
software simply rejects it.

Also, transparency means more than just "no ink here".  You move into
generalized alpha blending, and blending print inks produces way
different results than blending pixels.

Tim Roberts, timr at probo.com
Providenza & Boekelheide, Inc.

More information about the reportlab-users mailing list