Best way to remove background color in OSC images

joelshort

Well-known member
I've been imaging with mono cameras for years but recently started using an OSC (ASI2600). After calibration, the background in the images is completely blue. This is because my flat panel light source is weak in blue so the calibrated image gets over corrected.

My question is, what is the best way to correct the background? Is a simple pass with BackgroundNeutralization good enough, or is there something else I should be doing that results in a cleaner image?

Thanks in advance.
 
Is a simple pass with BackgroundNeutralization good enough

yes - unless you have insane gradients in your image, BN (or PCC with the background section configured properly) should remove the cast.

even if you do have gradients either of those should get you into the ballpark for a good DBE. you don't want the channels to be too mismatched when running DBE on an RGB image.

rob
 
yes - unless you have insane gradients in your image, BN (or PCC with the background section configured properly) should remove the cast.

even if you do have gradients either of those should get you into the ballpark for a good DBE. you don't want the channels to be too mismatched when running DBE on an RGB image.

rob
Thanks Rob. Can you point me to a tutorial on PCC, specifically on how to configure the background? I use PCC but one thing that I have never understood or liked is that after running PCC the background is really dark. In other words the histogram is shifted to the left so there isn't any room to play with on the left side. Not a huge deal but I'm sure I'm not using the PCC background settings properly.
 
it is pretty much the same as BN - what i do is just hover the mouse around the background and look at the values in the 3 channels, and then pick a background tolerance level that is slightly higher than the highest channel.

IDK about the post-PCC problem you are seeing - i don't think i have ever noticed that. STF with channels linked after PCC always looks pretty normal to me.

rob
 
So I've always just used BN with the default settings, which has "Rescale as needed" by default. If I change that to "Target Background" and put the target as slightly above the brightest channel (blue in my case) the result is pleasing.

There doesn't seem to be an equivalent to that in PCC. So PCC doesn't correctly do background neutralization as I understand it.

The attached files show the histogram after PCC and the histogram after BN.
 

Attachments

  • PCC.jpg
    PCC.jpg
    300.3 KB · Views: 143
  • BN.jpg
    BN.jpg
    295.6 KB · Views: 122
BN is one of those tools which can never be used with defaults as every image is different...

it's really only the "upper limit" control that you need to be tweaking. you shouldn't have to change modes or set the target background for it to work right.

PCC has the same exact limit controls and if you configure them right (slightly above highest) it should neutralize the background. OTOH there should not be any harm in running BN and then going into PCC. i guess you can even turn off background neutralization in PCC if you run BN first.

rob


bottom of PCC window; note it has to be checked actually do the neutralization.

Screen Shot 2020-06-17 at 10.57.14 AM.png
 
Something's not right, or I'm doing something wrong. I cannot get pleasing results no matter what I do with the Upper Limit. A tiny variation will result in either an unchanged background or severely clipped.

Could you take a look at this image and play around with it? It is a stacked image (no rejection) that has only been cropped.

On another note, I'm going to have to invest in some serious data storage. These files are absolutely huge. That single stacked image is 550mb, and the calibrated subs are each 750mb!
 
Joel,

Did you mean to display your Histogram in 8-bits?
What is your plate scale?
What is this object?
-adam
 
it seems fine to me? see attached screenshot with linked-channel STF applied after PCC. i guess there is a slight gradient in the image but i am not familiar with the target.

rob

Screen Shot 2020-06-17 at 8.27.52 PM.jpg
 
oh i should say though that there are a lot of hot pixels left hanging around and they seem to be all lined up (no dithering?). anyway seems like maybe more aggressive rejection is in order (or cosmeticCorrection to the calibrated subs)

rob
 
On a side note, for general information: In the next version 1.8.8-6, the ImageCalibration tool will allow applying different scaling factors computed separately for individual RGB channels of CFA flat frames. This option will be enabled by default and will solve all of these color cast issues with CFA data. Note however that these issues are just cosmetic; as Rob has shown, the initial color cast is no obstacle to perform a perfect color calibration with PCC.
 
Sorry guys, somehow I missed these responses.

Joel,

Did you mean to display your Histogram in 8-bits?
What is your plate scale?
What is this object?
-adam

I've been using PI for years and I am ashamed to admit I never noticed the histogram bit scale. Changed it to 16bit.
Scale is 1.6, pixel size 3.76
This object is vdB111

it seems fine to me? see attached screenshot with linked-channel STF applied after PCC. i guess there is a slight gradient in the image but i am not familiar with the target.
I think my problem with PCC was that I had changed the lower limit and that was messing things up in relation to the upper limit. Once I moved the lower limit back to zero and the upper limit to 0.03 I get the same results as you.
However, my point above is that after PCC the histogram is nearly flush up against the left side. There is almost no room at all to work with. But, I guess that's not really a problem...

oh i should say though that there are a lot of hot pixels left hanging around and they seem to be all lined up (no dithering?). anyway seems like maybe more aggressive rejection is in order (or cosmeticCorrection to the calibrated subs)
Yes, this image has not had any kind of pixel rejection at all. And you are correct that this image was not dithered. I run a tandem scope (side by side) setup and as of yet I do not have the ability to coordinate dither between the two scopes/cameras. If you recall I had some questions regarding some issues I was having with CosmeticCorrection a few weeks ago. You helped me with that but the solution was complicated and I'm lazy. I basically abandoned CC in favor of a bit more aggressive rejection and it has been ok.

On a side note, for general information: In the next version 1.8.8-6, the ImageCalibration tool will allow applying different scaling factors computed separately for individual RGB channels of CFA flat frames. This option will be enabled by default and will solve all of these color cast issues with CFA data. Note however that these issues are just cosmetic; as Rob has shown, the initial color cast is no obstacle to perform a perfect color calibration with PCC.
This is good news Juan! Yes, it is purely cosmetic but I will appreciate this nonetheless.
 
However, my point above is that after PCC the histogram is nearly flush up against the left side. There is almost no room at all to work with. But, I guess that's not really a problem...

yes i don't think that is a problem. as long as no clipping has occurred then it should not matter if the histogram is up against the left edge. not that PCC can cause clipping, i just mean that in general.

Yes, this image has not had any kind of pixel rejection at all. And you are correct that this image was not dithered. I run a tandem scope (side by side) setup and as of yet I do not have the ability to coordinate dither between the two scopes/cameras. If you recall I had some questions regarding some issues I was having with CosmeticCorrection a few weeks ago. You helped me with that but the solution was complicated and I'm lazy. I basically abandoned CC in favor of a bit more aggressive rejection and it has been ok.

yeah i remember, just wasn't 100% sure when posting.

rob
 
yes i don't think that is a problem. as long as no clipping has occurred then it should not matter if the histogram is up against the left edge. not that PCC can cause clipping, i just mean that in genera
So I think the problem there was the 8bit thing in HT. It appeared that the image was severely clipped but in reality it wasn't.
 
:) :) :)
I thought was the issue (the 8-bit representation of your data)...but did not want to assume before looking at your image.
That is why I asked that question first.
Pretty subtle eh?
-adam
 
Back
Top