Proper Use of DBE

[sreilly]

I use a mono STL-11000 with filters for imaging so I end up with individual red, green, blue, and luminance images to compile my final LRGB image. As it stands now I align all my data at once for all the images, save, create the master red, green, blue, and luminance images. I crop the image after creating my rgb combine image. Should I be using the DBE process on each individual master frame, being the red, green, blue, and luminance before combining? Or should I combine the RGB, crop the edges, and then use DBE on the combined image? There should be some logical order to the work flow but all I can say for certain right now is to use the processes that require linear data before using the histogram stretch tool.

Any insights?

[Philip de Louraille]

Seems to me you should not DBE the R, G, B channels by themselves. The combo RGB gives you L. So independently DBE'ing R, G, and B and combining would give you 'tinted' color field (for a lack of a better explanation.)
Seems to me like doing the LRGB combo and then a single DBE pass would minimize 'noise' artifacts due to the filter.

[sreilly]

I put together a little comparison chart showing the different results. The top 3 images are the master red, green, and blue images with DBE applied to each. There are examples of these as a compiled RGB image as well as the masters without DBE before and DBE afterward on the resulting RGB image. Not sure how useful this will actually be but I think the chart shows the difference fairly well.

[Philip de Louraille]

I guess you'd need to include a RGB Combined w/DBE and Levels Applied as well for a more complete comparison but the Individual blah blah .... and Levels Applied looks very similar to the RGB Combined w/DBE. Hard to tell at that scale but it may even look better "visually" (not to say that it is more scientifically correct.)
However, from the small images, it seems that there is still too much green

[sreilly]

It's not really very scientific but as a final point, the finished product is http://www.astral-imaging.com/M33-NGC-598.htm

I need to look at this closer. The funny thing is that the background issues show much clearer with the STF tool. After DBE, Background Neutralization, Color Calibration and then Histogram Stretch it looks much better.

Thanks,

[mmirot]

I believe good results can be achieved on single rgb or lrgb image.
However, I think most of the time it is better the to process individual frames.

 Actually, it may be just easier for me.  8)

Max

[jtalbot]

Steve,

My two cents.  I live in a suburban location with plenty of light domes.  I have been using PixInsights, DBE function since its inception in the LE days.  I find the best results are obtained if you apply DBE to your linear LRGB masters separately.  There is another caveat too.  If you have multiple areas of the sky where the gradients are different and you imaged through the meridian, stack all the images on each side of the meridian separately.   Do your gradient correction and then combine the two in pixelmath.  The gradient contribution to the image gets complicated if you have several different levels of light on each side of your sky.  I have totally different gradients to my east than I do to the west.  This simplifies the application of DBE significantly.

Once you have the separate LRGB masters DBE corrected, combine to a LRGB in your favorite way.

Best,

Jon Talbot

[sreilly]

Jon,

That makes total sense to me. Although I'm fortunate to have a fairly light polluted free area, it still has its issues at times. Combining the two masters, east and west, with pixel math using sum?

Thanks,

[Nocturnal]

Hi,

when you do DBE on a color image each channel is examined separately. Since your channels contain the same targets the sample locations are equally valid for each channel. In other words if a sample lies in the background for R it lies in the background for G and B as well.

So you can save yourself some time by doing DBE once instead of 3 times. You can always view just 1 channel instead of all 3 while placing samples. I sometimes look at just R to find the Ha.

It seems to me that if you have an L frame there is no risk in first combining it with RGB to create your linear master image before doing DBE.

Naturally Juan is the authority on this topic so he'll hopefully jump in if I'm spreading misinformation.

Steve: always crop before doing DBE. 'Crap' borders affect the statistics DBE depends on.

[sreilly]

Yes Sander, that became apparent pretty quickly. I may still have the "PS" approach that's hard to shake but I seem to have better control over doing the RGB and Luminance separately and then combining after doing a dynamic align. Then of course I have a final crop to do.

Thanks,

[Nocturnal]

Hi Steve,

in that case I recommend creating your registered RGB and L images first, then DBE one of them and then re-use the sample set on the other one using a process icon. It will save you some time. At least for me correct sample placement is quite time consuming.

[jtalbot]

Sander,
You have a good point that I forgot to mention.  Once I figure our where my background areas are I save a process icon, then use this same process icon on the other images.

The reason I do DBE on separate individual colors is that for me its much easier to figure out where my "real" gradients are by looking at the individual frames.  I suppose one could do a combined DBE on a RGB image just as well since as you say, the process is carried out on each individual channel anyways.

Jon Talbot

[Nocturnal]

Jon,

we all have our methods, thanks for sharing yours. Like you I like to check individual color channels for DBE but I do that by loading the RGB image and switching the view between R, G and B. The benefit of this method (besides the time savings) is that you can also view L or Saturation or any of the other presentations to help place samples. Load an RGB image and check Image->Display for options.