>>3863789>the same resolution bayer image has three times less resolution in the color layers,
In practice it has roughly half the resolution.
Bayer interpolation *does* work in natural occurring examples.
It doesn’t work perfectly (=equivalent to 100% full colour res), but it’s not like it doesn’t work at all (=equivalent to 33% full colour res).
It’s in the middle, it works out to ~50% full colour res.
You can test that with a Foveon sensor against a Bayer.
The Foveon (the full Merrill one) will be roughly the same quality all round as a Bayer twice its output resolution.
If Bayer interpolation was perfect, you’d have 15MP Bayer = 15MP Foveon. That’s very clearly wrong, 15MP Foveon is much better.
If Bayer interpolation didn’t work at all (as you’re claiming), you’d have 15MP Foveon = 45MP Bayer. That’s also clearly wrong.
But check a 15MP Foveon vs a 30MP Bayer and you’ll see they’re virtually the same on average.
An easy proof of concept is checking a Merrill Foveon again a D800 and then against a 24MP APS-C bayer sensor.
The D800 (36MP) will edge out very slightly the Foveon, and the Foveon will edge out the 24MP Bayer, placing it right around 30MP “bayer-equivalent”.
I’m all about full colour sensors, but both sides are exaggerating their claims. Interpolation works, statistics is a powerful thing, just not as good as bayer manufacturers say though.
And Foveons work better than interpolation, just not as good as Sigma says though.
Long story short, cut your bayer res in half (and not to 1/3rd) and you’re right on the mark.