Oh well, normally I manage to keep my mouth shut until a new algorithm is ready for release before I talk about it. But I couldn't resist this time. Since the cat is out of the bag now, anyway, I can just as well post a few more details, so here we go...
I've read multiple times now that Sony's convergence / panel alignment feature results in pink (or multi-colored) discoloration in specific test patterns. It was on my to do list for a long time to add panel alignment / convergence correction to madVR, and so I wondered *why* the discoloration would occur. And after some tests I found out. Here are some test images:original test pattern
---> correction in gamma corrected light
The corrected image is a result of my own panel alignment / convergence correction algorithm, by using gamma corrected light (which is what movies are usually encoded in) to "realign" the green channel. The red and blue channels are left untouched. In the corrected image you will see the typical pink discoloration. I think that this is probably what Sony users are seeing when using the Sony projector panel alignment feature? The cause of the discoloration is that the processing is not done in linear light, as it should be.
Of course madVR will use linear light processing. Since your computer monitors (which you will probably use to look at these images) are probably not properly calibrated, I've created a few different images now, with different gamma values:
linear light correction, with gamma value: - 1.8
One of them should be neutral without noticeable discolorations on your display.
FWIW, the main purpose of this feature will be panel alignment correction for 3-chip projectors (doesn't matter if it's DLP, LCD, LCoS or whatever). But the algorithm should also be able to correct chromatic aberations caused by non-perfect lenses. So it might also be helpful for some single chip projectors.
This feature may result in a small loss in sharpness, though. So it will be to everyone's personal judgement whether using the algorithm improves overall image quality or not. That might also depend on the exact correction necessary for any given projector. E.g. sharpness loss will be biggest if a channel needs to be corrected by 0.5 pixels. Correction by 0.1 or 0.9 pixels costs much less sharpness than correcting by 0.5 pixels. If convergence differs for the 4 edges of the screen, sharpness loss will also vary across the screen, depending on whether correction will be nearer to 0.0 or to 0.5 for each separate pixel.