Using DBE on Individual Colors

[llpastro]

Is there a down side to using DBE on color masters before creating the RGB?  Someone on another forum said this should not be done.

Larry

[vicent_peris]

Hi,

Yes, a big downside. In a greyscale image you have the brightness information about the color. In a color image you have the brightness and the color information about the gradient. Please consider that the most important when you try to correct residual gradients is not the tool you're working with but to learn to analyze the gradients. It is very important to carefully learn and have practice looking at the gradients of your image in order to correct them as well as possible. If you correct the gradients separately, you're analyzing only part of the properties of the gradient and it will almost never be efficiently corrected; you'll probably need then to join the RGB channels and further correct the residual gradients.

Take into account that this can be very difficult in some images where you have very dim and diffuse nebulae, and even dangerous since you can be introducing color variations that came from placing the samples in the wrong place; think that, in those cases, sometimes you have more information about the nebulae in the color than in the brightness component.

There can be few exceptions, that are not the general rule. For instance, a well defined gradient in a single channel in a completely red nebula where almost all the extended object is red and it's very faint and precisely located in the green. It can work then. But it's not the usual case, and it's not the way to learn gradient correction techniques. I can say that, after teaching to correct hundreds of images from the students in my workshops, I have only needed to split the channels 2 - 3 times.

On the other hand, I always advise to correct the gradients before the background neutralization process. This way, you have a flat background and, after running DBE, you can safely place a good reference preview in your picture. It doesn't matter if the sky background levels are very different in each channel since the statistics are calculated separately for each color channel.

Please, don't put my words on stone. I'm trying to communicate here that, as a starting point, it's much better to work with the color image (it could be useful to visualize the separate color channels as well from the Channel Selector). You'll find a good solution without needing to split the color channels in 95% of the cases. Then, analyze your result and analyze your problem; you could need to split the channels, as I did those 2-3 times!

I think that user in the other forum came to one of my workshops since he was giving exactly the recommendations I usually give in my workshops.


Best regards,
Vicent.

[pfile]

its possible he took your class. that user made it sound like it was a sin against humanity to do DBE on one of the channel masters. i essentially said that the world would not end if you do it but you will probably end up doing DBE again since your eye will more easily see the gradients as colors.

imo the only problem that crops up here, and this is probably only relevant in high LP conditions, that the absolute value of the backgrounds of each channel master could vary wildly. this might cause you to oversample the background in one or more channels during DBE. so it probably makes sense to linear fit your masters before combination so that the levels are comparable during DBE.

rob

[vicent_peris]

Hi,

I say "never" or "always" very few times. And I always tell people that I try to teach them how to think by looking at the picture, I don't try to give them my rules in stone. As I told, I needed very few times to first split the RGB channels and apply DBE separately. I have corrected really wild gradients from light pollution working with color images.


Best regards,
Vicent.

[pfile]

i wasn't attributing his absolute language to you, just saying that he might have attended the class and then extrapolated from there.

rob

[vicent_peris]

Hi,

Yes, I know. Trying to summarize, you can apply DBE to each channel separately, but it can be difficult and, sometimes, dangerous and inefficient. Start with the color image, analyze the gradients in the RGB image and separately in each channel (if needed). Then, decide if you need to apply DBE to the color image or to each channel. The latter can be simply because you need very different samples. But, 99% of the time, it will be easier and safer to apply to the color image. And safer here is very important, because these gradient correction techniques are always a guess. That's why, when you have a very deep image, you need also a clean master image as free from gradients as possible. At a certain point, you'll almost never know if that's a gradient or a color variation of the object.

Best regards,
Vicent.

[Duncan]

I'd be interested to know if the process that I've been trying out with a couple of images sounds sensible... it has worked well but that could be a fluke.

The images that I am working with have gradients that are quite different for each of the color channels (the data was captured over multiple days, and the object was in a slightly different location in the sky for each set) and I am applying DBE separately to each channel.

I created a set of samples using the luminance image, and spent a lot of time doing multiple trials on it to see what I get. I look at the extracted background and resulting image (subtraction mode) using boost mode STF to see if the background is even, moving sample points around until I get a result that looks ok. Then I generate the final DBE luminance and save it, drag the DBE new instance to the workspace and close everything.

For each of the color planes I then open a new instance of DBE, and drag the saved setup back onto it and run it again. So I'm running DBE separately on each channel, but with the same positions of sample points (the channel images are already  co-aligned at this point).

Any thoughts on this? As I said, this has worked pretty well for the images I have been using but that could be situation-specific.

[llpastro]

Vicent, what about the situation of creating a synthetic luminance?  If I wait until the color masters are added, I have only gray for DBE to work with.  Will the result conflict with the DBE version of the RGB?

Larry

[vicent_peris]

Hi,

I would then compose the RGB image, correct the gradients, then split the RGB channels, save the three primary color images and compose the synthetic luminance with ImageIntegration. This way, you'll have a synthetic luminance without gradients.


Best regards,
Vicent.