[Bf-vfx] Improving the 2D stabilization tool :: Next Steps

[Ichthyostega]

Am 08.06.2014 15:25, schrieb Sergey Sharybin:
> internally all the stabilization values are to be stored in normalized
> [0..1] space in order to be re-applied for the proxied movie.


...another remark:
In the code used up to now, they where *not* stored in normalized coords.
This is just one of the examples hinting at a slightly confusing code structure;
I tried to re-order the situation somewhat for that reason.

The existing code *calculated* in normalized coords, but then scaled up to
canvas size. The compensation for non-square pixels was scattered over several
places and not done completely.

Then these values where passed up through two forwarding functions right into
the matrix calculation.


In the version you got for review, I deliberately didn't change the interface
and thus the translation[2] is still in pixel space. But ideally this should
be changed on the interface to be in normalised coords. Please note that
these "normalized" coordinates are still unlimited in value range, since
in a traveling shot basically the translation can go as far as we want.

	-- Hermann