Saturday, February 4, 2023

Some believe "RGB only! We don't need Lum" ... but, data disputes that.

As a professional research scientist (career), I am against arbitrarliy throwing away data (in this case, photons). There are some trying to convince others that, "You don't need Lum...you can simply use RGB alone" ... Well, I will let the spectral plot speak for itself. There are some who want to use "AI" and "noise reduction" to essentially INVENT missing data that they decided wasn't important enough to collect (aka, they want to "skip Lum data").
Why ignore photons? I have no idea. As a scientist conducting basic research, in my mind ignoring data and attempting to re-create it later is just an exercise in futility.

If one wants the full visible spectrum of data...then one needs to collect the full visible spectrum of data - else, DATA WILL BE MISSING. And no amount of "AI" or so-called "noise reduction" will ever bring it back. Period.
Image BELOW:
Imagine peeling out only the Red, Green, and Blue part of the spectrum you see the prism splitting (on right) from white light/Luminance (on left). Imagine all that you are leaving out. That's what you are doing when you only use Red, Green, and Blue filter data - leaving out the rest of the detail that was actually there.
In the past, I experimented with 1x1 binning with BOTH Lum and RGB. Same night, same camera, same telescope, and exact same data collection period/window. Even when RGB had a total amount of photon collection GREATER than Lum, it still was missing data (as expected). I want to emphasize...this data collection was nearly a decade ago (and collected for a different purpose - but it still hammers the point home).

RAW FITS DATA ONLY ... ZERO processing... ZERO.

Again... I'll let the data (image - uncompressed TIFF) speak for itself.

LUM (70mins total) on Left ---- RGB (75mins total) on Right

No comments: