AssistedColorCalibration - use in this case?

[astropixel]

Hi. I have read the ACC documentation and Zbynek's tutorial, and experimented with the image below several times.

Not sure how I should proceed

The bright green is typical of all my integrated images, though the finished image turns out OK?

The closest I get to actually seeing what is going on, is to first use BackgroundNeutralisation, which I understand from the tutorial is not optimal.

Many thanks

Rowland

[astropixel]

I read the comments about BN before DBE in the discussion on HP's processing video,  and this seems the only way to first remove the gradient to see what's going on with ACC.

[zvrastil]

Hi Rowland,

in case preview used to evaluate ACC contains gradient, you should indeed first correct this gradient with DBE. In the tutorial, image pixel is described as P = k*(S+B) where S is signal and B is background. It recommends to first find k coefficient and then subtract background. However, you could do DBE beforehand. Considering P = k*S + k*B, you remove k*B with DBE. Then, you can use ACC to find out and correct for k (= do color calibration).

ACC is designed for cases where standard CC does not give satisfactory results for whatever reason (like no galaxy and low number of stars). Using ACC, user is the final arbiter to evaluate CC with help of previewing capability of the process (built-in histogram transform and color saturation enhancement).

regards, Zbynek

[zvrastil]

Some more notes - I think green color is normal result of PixInsight calibration of bayered images (like from DSLR). I have it too, although not so strong (maybe just different STF). I understand it's result of bayer pattern containing twice as much green pixels as blue and red pixels.

Looking at the image, I'd first try the workflow described in tutorial. I think gradient is not so strong to affect it too much.
1) make two previews - one in dark areas to the bottom-left of nebulae, second covering both nebulae
2) select first preview as Background Reference. As it is close to second preview, gradient won't be of much importance
3) make sure you don't have STF applied on your image
4) apply ACC to second preview
5) adjust histogram transformation to see nebulae clearly (reapplying process to second preview to update)
6) boost saturation to 3 or 4
7) adjust RGB sliders to get colors you want
8) apply process to original image
9) do DBE to the image
10) check result with quick histogram & saturation stretch

regards, Zbynek

BTW, ColorCalibration on stars does not work for you?

[Cleon_Wells]

Rowland,

"The bright green is typical of all my integrated images, though the finished image turns out OK?"

I have a modified Canon T1i and in PI I always get a green cast to my converted subs, I assumed this is caused by the White Balance box not being checked in the DSLR_RAW file format converter. When I open a RAW file in PS the Blue channel is higher then the Red channel, which is higher then the Green channel, see attached comparison images, interesting. Another question added to your post. 
Dbynek, I also thought that the green cast is caused by the ratio of G to R and B, pixels but when I open a raw sub in PS the RGB ratios are completely different. 
Cleon

[astropixel]

Zbynek, thank-you. I shall put all of this into practice over the weekend and report back. Thanks also, for explaining the rationale behind extracting the background first. I can approach the problem with a degree of confidence that the results I'm seeing are authentic.

Cleon. Yes, interesting. The distribution shown in your PI example is typical of a newly integrated image - following background extraction RGB tends to be more symmetrical. This seems to improve with linear ATWT noise reduction.

I don't use PS so I can't comment. PI is presenting the data as is, with the RAW converter settings as per the DSLR work flow.

My image has B well out to the right unless ABE is applied with very small sample sizes. B is much more dominant than G then R.

B is so dominant it makes processing quite difficult. Thinking the problem through led me to think white balance issues? and ACC. I'll see how that goes.

[Cleon_Wells]

Rowland, it would be interesting to see what a calibrated raw dark sub and calibrated raw Flat sub looks like on your system. Here’s an example of what my Calibration setup does to a Canon T1i raw Dark and raw Flat sub, see image. The DeBayered c_Dark.fit sub shows hot pixels which are hard to see in this screen shot. The black point in the Dark STF stretch is zero.
Cleon
PS Where is your camera’s White Balance set ?

[astropixel]

Had to dig around for 1000D/XS flat and dark subs. These are bias subtracted only. Not sure about the pink? Clean sensor!

[astropixel]

Here's the ACC efforts. Same ACC application, but different approach to processing - left or right - saturation in R???

Both images are an improvement on previous attempts with CC. The peppering of white stars was virtually unseen and the stars colours quite uniform.

I kept G at 1.00 and adjusted R and B to come up with;

R 1.809
G 1.000
B 1.609

EDIT: B 1.67 is closer to the mark.

I guess it's down to what do you think?

[Cleon_Wells]

Rowland, I think your Calibration files have some problems. The color in your processed image looks pretty good considering the calibration files you are using. All I want is images that impress my wife, come to think about it, could Juan write a Wife Impress script for the next big update.  8>)
I have an image that shows what I think are average Calibrated raw subs for Dark and Flat image.
I would not modify the raw files with the DSLR_RAW preference, this can lead to/add to the confusion of getting good _c.fits files.
I would take a hour and create new test Cal files----20 bias, 10 darks, 10 flats and see if you can Cal a Dark and Flat raw sub that looks closer to mine.
Cleon

[astropixel]

EDIT: These are the 5D MKII files applied to the image. Sorry, I misunderstood your earlier request and posted 1000D files.

I would not modify the raw files with the DSLR_RAW preference, this can lead to/add to the confusion of getting good _c.fits files.

DSLR_RAW preferences are set for straight RAW conversion - no white balance, AHD etc. That's how I always do it.

I would take a hour and create new test Cal files----20 bias, 10 darks, 10 flats and see if you can Cal a Dark and Flat raw sub that looks closer to mine.

The dark files (40) were taken automatically in the same sequence as the lights at the same temperature, give or take a degree, which is well inside dark scaling capabilities.

The bias files (50) were taken under the same conditions, except that the sensor would be cooling down.

Similarly the flats, except on this occasion I used 3 only - this was experimental. Usually, I take 20 - 30 as in the previous files.

I follow the calibration and integration instructions at http://pixinsight.com/forum/index.php?topic=2570.msg19019#msg19019 and acquire the raw files under controlled conditions.

To display the attached files,  it was necessary to convert them to greyscale, debayer, background extract and STF.

I believe that I am doing the right thing here. Not sure about the pink in the flats, but suspect it has to do with my light box, which is an EL panel with a diffuser.

A wife impresser - are they ever impressed by our astrophotographical antics.

[Cleon_Wells]

Rowland., I'm calibrating the Dark and Flat raw subs with the MasterBias and have a problem with the Dark_c level, I hope to get to the bottom of this problem with the Dark subs today.
The Flat subs look correct, Debayer Flat level is reduced by the Debayer Bias level correctly.
Cleon
PS I didn't realize your image of the Dark and Flat were not Debayered.

[astropixel]

Cleon. I think the calibration thing comes back to acquisition and temperature on almost every occasion. I can trace all my earlier difficulties back to that. I now work on the principal, if in doubt take lots of frames, which tends average things out.

All the images have been debayered for display. But I admit that you are examining things more closely. I'm happy to get past the colour issues at the moment. Not a brilliant image but on closer examination the star detail has improved exponentially thanks to Zbyneks ACC tool.

Cheers

Rowland.

[Cleon_Wells]

Rowland, here's a comparison of uncalibrated and calibrated Calibration files. There are 20 Bias 100@1/4000, 10 Darks 400@120sec and 10 Flats 400@1/250 subs that I used to create the Master files and 1 M52 light 400@120sec under the Moon.
Cleon
PS I don't know why I had problems calibrating the Dark subs, I restarted PI and had no problem, go figure.

[astropixel]

Equally valid from what I understand - if the master files are not calibrated before integration, check the Calibrate boxes during lights integration.

I normally take the flats at iso100 (which means a set of 100bias frames also, just for the flats) and the bias and dark frames to calibrate the lights at the same iso as the lights. I notice that your bias is iso100 - I'm interested from a learning perspective, because I don't know if it matters all that much. Guess I play it safe.

Cheers

Rowland