Newbie Question - deBayer problem with 40D?

[another_mikey]

Hi,

I just downloaded the trial of PixInsight a few days ago and so far like it a lot. I think I am having a problem I might have aklready seen mentioned here but it wasn't clear to me how to fix it from the previous thread.

I have processed a few stacks using lights and darks and using the PreProcessing script. Although I have been able to use the DBE tool to 'fix' this, I am thinking that there is really an issue somehow with the script as the image comes out bright lime green to start with. I am assuming this is an issue with the program settings being wrong for the actual filter mosaic in my 40D?  I also noticed that my final output was 'flipped' (starting at the wrong end of the file?)

I have played a little bit with a few of the settings (Up-bottom FITS, changing the bayer matrix order from RGGB to some other choices) but as of yet have not been able to solve either of these problems. Interestingly, if I use the deBayer tool by itself with the default settings the output file background color looks ok. So perhaps this is a script issue? I can try to run all the steps that the script runs individually and see what happens but given what a common DSLR I am using I am assuming this must be a fairly well known issue that has either been solved (if it is user error) or at least documented (if it is not).

Can anyone help me out?

Thanks bunches,

ML

[pfile]

are you starting with CR2 files or fits files?

[another_mikey]

Thanks for the reply!

I am starting with .CR2 files and have the CFA checkbox checked.

ML

[georg.viehoever]

I have a Canon EOS40D. PreProcessing works finde for me. The colour tints disappear after doing BackgroundNeutralization and ColorCalibration. For a preview, use STF in Auto-mode with decoupled color channels.

[pfile]

okay. if you are starting with CR2, i think all you'll need to do is make sure CFA is checked. the BPP script is going to write out FITS files of course, but the initial calibration should occur against the CR2 and so there should be no confusion about up-bottom or bottom-up. as long as you don't change the settings midway through (like you make calibration masters with one FITS setting and then go in and change it) i think that part should be okay.

as Georg says, weird color casts can happen - for me, i get them because i use a CLS filter - but it's true that funny colors can be a result of wrong debayer settings. to my knowledge, all canon cameras use the RGGB matrix, so you are probably all right there. if you zoom in on your calibrated images and they look normal (that is, no weird "screen door" look) then you should be fine.

one thing you can do to get a 'preview' of what the data should look like is to do AutoSTF with the channels uncoupled. in the STF process window, the channel coupling is the little icon that looks like 3 RGB blocks stacked up into a pyramid. turn that off and you'll get a mostly neutral view.

[another_mikey]

Thanks to all that replied. So summarizing with what I have heard so far:

• The RGGB shoud be fine for my camera
• Using the STF tool could verify that I am decoding okay
• The color casts are somewhat 'normal' in a lot of cases. This still seems a bit strange to me, especially since this looks BRIGHT lime green (i.e. there is nothing subtle about it)

I still have the issue of getting a flipped image when running all the steps using the PreProcessing script, so that it seems as if the image has been flipped about the horizontal axis. Here is an example after running thru the PreProcessing script. Examining any of my .CR2 raw files, M42 is in the lower right (as it looked when the data was being taken), but the output of this script puts it in the upper right, as shown in the following link. And in the output, before I used DBE it was also bright lime green:

http://mikeyl.zenfolio.com/p341668809/h4c96b3f8#h4c96b3f8

Any additional thoughts on what I might be doing wrong to mess up the orientation of my image?

ML

[astropixel]

Hi Mike.

RGGB is correct for just about any Canon camera, which typically come with a standard Bayer matrix.

Deselect 'up-bottom' I recall it flipped my images too. I'm not sure at what point in the process this happens, and while not ideal, Main menu bar > Image > Geometry > whichever, will right the image.

The bright green cast is normal and is removed with background extraction - wysiwyg, nothing is hidden from view with PI - as far as I know... Try ABE. Irrespective of background extraction, after stretching, it is usually necessary to apply SCNR Green Average Neutrual (default settings) to clean up before further processing.

No comment on STF - I use it to have a look at what I've got and move on from there. You can do 'CTRL click A' to set the light and dark points and drag the instance to the bottom bar of the HistogramTransformation tool, which adopts the same values as STF.

Rowland.

[pfile]

i agree with rowland - it seems likely that the fits orientation checkbox in the BPP panel causes the flip. for some reason i ended up checking this and then since my calibration masters were written out with that orientation, i had no choice but to keep using it (or re-make my masters, i guess.)

it's not that big of a deal, as long as you are aware it is happening. you just need to mirror the image across the horizontal axis, and as you've seen you can check what's happened by opening one of the CR2 subs and looking at the image.

i think your results look fine, so probably everything is working okay. alls you gotta do is flip the image.

[another_mikey]

Thanks for the continued input. I think I have corrected my problems, and it still seems like th script may have a bug. Based on my comments below, see what you think:

I used the debayer script to produce my .fit files from my darks and lights using all stock options. Then I used the PreProcessing script for the rest of the processing, again with all stock options and with the CFA option unchecked. I then got a good combined light frame, with the correct orientation and NO lime green tinting. So I think there might be something going on with the way the Preprocessing script handles the debayering options, but in any case I am a happy camper now as I can use this workflow going forward and appear to have no other issues.

Thanks again for the support!

ML

[astropixel]

Mike. A little less busy, after fruitlessly attempting to recover 2.5hours of data, I recall that pixel rejection rates are significantly higher using RGB frames for calibration. The point of the BPScript sequence is to avoid using RGB files during calibration. As you know RAW is just a file and technically not an image.

The BPScript up-bottom selection is intended for users of amateur CCD cameras, as far as I know and forces correct orientation. It may be deselected. The BPScript was originally intended for CCD images - the authors also made it available for DSLR_RAW processing.

Pixinsight tools are essentially standalone and offer quite a bit of tweaking. As you may have gathered from the DSLR_RAW workflow, the BPScript throws all that together as a matter of convenience and I might add, comfort. The downside is that you get a green/red/blue cast depending on whether a filter is attached or not. The upside as I understand it, is that calibration is more accurate.

Regards,

Rowland.