Combining 1x1 lum and 2x2 rgb: part2

[papaf]

Hi all,
I decided to start a new thread as the previous one was a little bit different.
My doubt arises from the fact that I have a smallish ccd, namely an Atik 314L. It's a bit more than 1mpix, 1391x1039.
I'm worried that taking 2x2 binned rgb frames, the star field will not match during LRGB reconstruction, and therefore I'll end up with colored star halos, much like if I didn't use an apo objective.
I used to try this out, but as Sander, which I thank for his time, wanted more info on my images, I discovered I only tried this a couple of times and I only had Ha as luminance, so star size difference was even bigger. And the result presented, as I said, big star halos.
Could this be because of the extreme Ha difference with RGB?

Another input: I tried simulating a 2x2binned rgb by simply resizing a normally taken 1x1 rgb set. When I combined those, after aligning them, I noticed that stars were not white but had some colored fringe. Could this be from the resize, and therefore making my test invalid?

Thanks to all!

Fabio

[lucchett]

Hello Fabio,
I can summarize my points of view(only base don my personal experience) like:

-if you make an LRGB the star halo, providing that the LUm and RGB file are well matched, shouldn't be a problem because the small stars in the lum frame don't fully support the RGB stars: that means that the outer halo doesn't have the support of the luminance. If you are using the LRGB tool the halo could come from a wrong setting of the chrominance noise setting.
-if you use the ha as a new red in a RGB, yes you can get the halo (usualy a gree halo outside of the bright core because of the lack of red

- for the third point, I think it depends on the original sampling of teh 1x1 image. but you can solve this using some third party raw RGB files (One source could be JIM Misti website)

Ciao,
Andrea

[sreilly]

I think something missing here is image scale. As an example, when imaging with my 12.5" RC, my image scale at 1x1 with my ST10XME is .48 arc seconds per pixel. So binning 2x2 for the RGB is giving me an image scale of .96". On the other hand, if I am imaging with my FSQ-106, my image scale is 2.65" so I never bin the RGB to keep the image at a higher resolution that my skies do give me. If I bin 2x2 with the FSQ than I end up with 5.3" which just doesn't do well, at least not for me.

[Nocturnal]

You are right Steve. Image scale is important too.

[papaf]

So, if resolution gets too low, I should avoid binning? What do you think this limit is? In my situation, I have 3.17" per pixel already when non binned.

Thanks

Fabio

[Nocturnal]

Hi Fabio,

so binned you'd be at over 6" per pixel. That's low resolution by most standards. I don't think the SNR benefits of binning care what the angular resolution is. So technically it really doesn't matter.

The key point is still, given the same amount of total exposure time, which one is better:

- RGB   1:1
- RGB 2:2 + L 1:1

Intuitively I'd say that the RGB 2:2 'blurs' the image even if 1:1 L is applied. Not sure if this is true. If it is then the 2:2 image scale matters. If it isn't true then the image scale is immaterial.

I wonder if the following experiment would be valid

- take RGB image
- extract L
- downsample RGB 2x
- extract R, G and B from RGB
- combine L with R, G and B
- compare to original image

If the images look identical under close scrutiny then image scale is immaterial. If they do not then image scale matters.

Finally the final publishing scale also matters. It makes little sense to worry about pixel resolution for a 3000x2000 image if it will be published at 800x600.

[sreilly]

Sander,

Not sure we are talking the same language here. When I refer to luminance I'm referring to data taken through clear/luminance filter, not extracted from a RGB image. The fact that more light is gathered through a clear/luminance filter should result in a stronger more intense image for a given time. Now I know some who have taken straight RGB images taking images through red, green, and blue filters then adding those for a synthetic luminance but I'm not sure that's the same. In my case were are talking 10-12 hours of luminance data with maybe an additional 6-8 RGB data added to get the L+RGB image. In this case I can say from my experience, the luminance is definitely the most detailed of the two with a stronger SNR. Adding the RGB for the color signal alone is the main purpose of the RGB data for me. That's what I've heard over and over many times at many imaging conventions by highly regarded speakers. That said, it is still personal taste that should prevail in each person's imaging. While I admire many imagers, I strive to learn from others while developing my own style taking what I like from others and applying that to my images.

It will also be subjective to the software used as to what the final data can produce. As I've seen in PI, the need for additional software, other than acquiring the data, is far less than any other software I've used. Some will say that no additional software is required but I haven't gotten there yet. I still use PS, I already own it so I might as well use it, to create the web image JPEGs but that's about all. I still like seeing what the final size of the image will be when I generate the JPEG and having control over that before creating that image. It also gives me a visual for artifacts that may be introduced when choosing the quality or percentage of compression.

[papaf]

Sander,
I think RGB taken at 1xbin will yeld the best result. In fact, the entire point of taking a separate L channel is to save time, not to obtain a better result. To me, at least, this has always been clear.
The fact that I have to travel to a dark site every time plus bad weather forces me to try and save as much time as possible. So, ignoring for a moment the loss (probable) of precision, I just wanted to hear from you guys if this (rgb bin2x) was worth a try or not. It seems so, so the next time out I'll give it a go.
I tried something similar with what you suggest as a test, as you can read from a previous comment. Although it seems to work fine, I have some glitch around bright stars. They have color bleeding, but to me it looks like the result of RGB digital resample more than anything else.

Fabio

[lucchett]

Hi Fabio,
this is also my point.
LRGB make sense if you bin the color data, and only to save time.
At least with modern tools like PI, that work separately on luminance and crominance.

I used to shoot LRGB with RGB binned because I also travel to a dark site and that was the best option, especially during summer (2-3hours luminance 1x1 and the rest color data bin2).

I've tried to move to RGB 1x1, but I ended in getting only one image in two consecutive days. The image processing is easier as far as color is concerned, but the noise control is much more difficult.
getting 3h of luminance in a summer night is certainly possible, shooting 9h of RGB in two nights, in the field, it is more difficult.
not to mention that with less luminance signal you cannot use powerful tools such us deconvolution.

I think the trade off is between details and color accuracy.


So, in my personal opinion, if you have an observatory, shoot RGB bin1 till you get the signal that you need.
if you go in the field, both have pro's and cons (and the sensitivity of your ccd can dictate the choice)