[Cin] User Defined Blend Algebra

[Georgy Salnikov]

On Fri, 14 Feb 2025, Andrea paz wrote:

> 1- The colors of the 2 tracks that will be blended. We often consider
> only these channels to guess the final result.
> 2- The Alpha channel of the 2 tracks that can interact with each other
> in ways that are intuitively (but not mathematically) unexpected.
> 3- The choice of Source track and destination track (top or bottom).
> 4- The presence of a third black background track (or even the color
> of the canvas, which in CinGG is black by default but whose color can
> be varied manually), which can interfere with viewing the blend in the
> Compositor and show unexpected results.

And even more factors.

Overlays effect generally depends on the colorspace of the project (in YUV
can differ from RGB). In Blend Algebra this dependency is addressed: the
necessary colorspace conversion can be done automatically.

Overlays can depend on implicit clipping. In RGBA-FLOAT no clipping is done,
in RGBA8888 clipping to 8 bit unsigned cannot be avoided per definition.
Blend Algebra addresses this dependency also, clipping for float colorspace
can be switched on explicitly, and then the results for float and 8888
should be identical.

Unclipped overlays can be displayed differently depending on the graphic
driver (bare X11, X11-Xv, OpenGL, etc.). To my mind, OpenGL driver could be
more robust in these cases, it runs perhaps under control of some CGG shader
and seems to clip the bounds correctly.
_______________________________________________________________________________

Georgy Salnikov
NMR Group
Novosibirsk Institute of Organic Chemistry
Lavrentjeva, 9, 630090 Novosibirsk, Russia
Phone   +7-383-3307864
Email   sge at nmr.nioch.nsc.ru
_______________________________________________________________________________