Hi all,
Reply #10 to this thread, buy Juan (as usual
) contains some of the clearest reasoning behind choosing L+R+G+B, versus L+Rbin+Gbin+Bbin, versus Lsyn+R+G+B (or, indeed, Lsyn+RoscGoscBosc), versus 'plain old' R+G+B (or, again, RoscGoscBosc).
Certainly, in my recent M1 attempt, I still have never yet managed to get the L data for the sequence (remember, I managed to save three hours of 5-min subs in JPEG format - rendering them useless for any further processing, and so have since deleted the entire evening's data !). But, even without the L data, I still have a very pleasing image (to my jaded eyes, anyway).
Sure, I do also have the Ha data (taken as 10-min subs, to compensate for the narrow bandwidth - but I have learned from that experiment as well and, next time, I will restrict even my Ha data to well below the limit where Differential Flexure starts causing pixel abberation of more than 0.5 pixel.
And, for Dave, what I did to incorporate the Ha data into my (near enough) fully processed R+G+B image was to channel-split the RGB, then take the Rd and Ha images (already precisely aligned, naturally, and with the Ha data already pre-processed as well) and then use PixelMath to MULTIPLY the two images together, rescaling back to the PI [0,1] range afterwards. Before I recombined the channels, I applied a Histo transform to the Ha.R image, such that the background and median values were near enough the same as the original Rd channel.
And then all I needed to do was 'cosmetic' tweaks to get the image looking the way 'I wanted' it to look - which is always personal choice, and usually has NOTHING to do with the actual scientific content of the images.
So, my advice (advice that I will follow myself, after at least one more attempt to TRY and get my M1 Lum data) is to concentrate on RGB alone and - AS A SEPARATE TASK ALTOGETHER to concentrate on L data, just to get to know the processes in just dealing with MONOCHROME data - which you can, of course, then consider applying to Ha, Rd, Gn or Bu - if you need to.
Up until now I would have said that there was NO ADVANTAGE to using an extracted, synthesised, L channel from either R+G+B data, or from RoscGoscBosc data. But now, based on Juan's message, I will have to rethink that statement, and I would certainly not be quite so quick to pass it on as 'gospel'.
So, Dave, it just goes to show that - as usual - there are NEVER any 'dumb questions' in this game
And, just to keep Harry laughing, here is my Nov-2009 "M1 Crab" (Ha.RGB, full-size - remember the DSI-II is only 748x577 native)
Cheers,
