>>4386529Demosaicing with intent to interpolate color results in blending samples.
By skipping this step and just dealing with the filtered but otherwise intact luminance values you get closer to pure luminance. Not as good as a B&W/Panchromatic sensor with no filter on it, but better than bayer+demosaicing later desaturated. Things like fine color patterns won't be aliased by demosaicing then carry over their moire into the B&W conversion the moire simply won't be present, at least not to the extent that it is.
The issue stems from the fact that while demosaicing is assumed it is not a lossless operation. It always takes its toll on the image and if you're shooting B&W you don't need to demosaic so you can avoid any degradation that's caused by it.
When demosaicing to RGB you'll also usually be applying a color profile which is an attempt to convert the raw averaged samples into a more accurately presented gamut into a color space but this too can be lossy.
>>4386655>But it is??It isn't, it's luminance + metadata.
All digital color is created. There's no real "right" way to do it, bayer and the normal methods are inherently lossy and copium tier. People regularly lie and think they know what they're talking about but they know nothing. They get "schooled" or "educated" and just regurgitate lies.
You think it's color, but it ain't.
It's greyscale with some undisclosed wavelength filtering applied over the photosites and that's what (((color science))) is all about. How the luminance is converted to RGB. It's never standard.
>can you post an example where that works?Sure.
1/3
Top is demosaiced(AMaZE) then desaturted (luminance) at 50% scale
Bottom skipped demosaicing and was just exported as greyscale then scaled to 50%.
