Outer corners of image are darker

[sreilly]

and I think it may because of the RC flattener but had thought DBE would help correct. It's particularly apparent in this image of M33 which is a project that has been going on now since Oct/2016. The usable luminance data consists of 153 - 10 minute images and the RGB data has 63, 56, and 75 10 minute images respectively. Seems I can handle the background gradients better in the color data than luminance. A .jpg of the luminance is attached. It's 1/2 original size. I can upload the original calibrated image to dropbox if needed. On the RGB images I ran DBE on each color master frame before combining as an RGB image. When I run DBE on the individual luminance image it looks much better but you can't run that on 153 individual images, at least I wouldn't think. Suggestions always welcomed.

Steve
 

[sreilly]

the RGB image is attached but way to dark for my liking.

[msmythers]

Steve

I think your best choice for this would be to use DBE on your best quality calibrated L sub and then use the Local Normalization tool using that DBE'd sub as the reference for the rest of your subs. If I remember correctly from past postings you would do this at least before Image Integration and maybe before Star Alignment but I'm not sure on that. I can't give you instruction on how to use the tool as I have not had the opportunity yet but I'm sure someone else will help. Juan or Vicent I know can answer this.

As with all new tools it's a learning curve but from what I've seen and understand about the tool it should be a big help with gradients across large image sets or as in my cases gradients that are never the same from sub to sub during most nights.


Mike

[sreilly]

I have to admit I know zilch about  the Local Normalization tool. But it sounds interesting if I can get some guidance on using it. Thanks for the reply Mike.

Steve

[Geoff]

Why don't you just stack the 153 images and perform DBE on the integrated image?

[sreilly]

The results were terrible on the combined image. I set the correction to Subtraction and set the size to 12 with 10 samples per row. Division wasn't any better. Looks great when done but when the STF is refreshed it looks terrible. The attached screenshot is typical for this image. Maybe it's too much galaxy and not enough background?

[bob_franke]

Hi Steve,

With our background models we generally want a smooth transition from light to dark, so give this a try. Use a sample size of 50 to 100 and put one sample at each corner where you can find some background. Then execute with subtraction. I find it is better to use as few samples as possible and I use quite large ones.

Hope this helps,
Bob

[blinky]

Interesting post, I have exactly the same issue with my M33! I'm going to try reducing the samples and increasing th size as per the above post

[ngc1535]

Hi,

I do not see the original data (the JPEG looks clipped to me). However, in general for vignetting and "hot spots" due to flat errors- a simple circular symmetry (axial) will do the trick (likely much better/easier than placing samples in the corners- although it might work OK). In addition why choose subtraction as the method? A flat error is certainly multiplicative in nature (and not "additive" like additional light). So division seems like the best way.

I plan on demonstrating some of these things, including a nice example of Local Normalization, during my AIC workshop.

-adam

[sreilly]

Adam,

I'll be happy to post the original combined image to Dropbox if you want to see the raw data. In this data set it's very evident.

Steve

[ngc1535]

Sure... go for it!

[DMouse]

To be honest looking at the image it looks more like M33 covers the top left and bottom right corners and so therefore they wont be as dark, and the other two corners are just open space, so are darker. I don't think it is down to vignetting. Just a thought.

Clear Skies

Paul

[sreilly]

So the link to the combined M33 file is https://www.dropbox.com/s/5y16hjbqqevqqzj/M33-153-L.xisf?dl=0

Thanks,

Steve

[ngc1535]

Just looked at the image. I agree with DMouse- I don't see the problem. The top right and bottom left corners are darker since they are sky values instead of mostly galaxy glow. It seems the "terrible" bit you are describing is just a choice in how you choose to display the image.

https://youtu.be/B5bRDBGJWeA

[pfile]

since STF is computed based on statistical properties of the image, it's not uncommon for the auto-STF to look really bad after DBE or noise reduction... but the reason is simply that the image has been overstretched. if you save the STF computed before the DBE or NR and apply it to the image post-NR or -DBE, you'll see that it looks fine.

rob