Preprocesing problem with "Incompatible image geometry"

[STEVE333]

I'm trying to use PI to do Preprocessing manually for my Canon T3 RAW images.  I've been following a YouTube tutorial (Richard Bloch).  All works well when creating the Bias, Dark and Flat frames.

First I create master Bias, Dark, and Flat frames. When I try to use the ImageCalibration process to correct the RAW images I get the Incompatible image geometry error message.  The Bias, Dark and Flat frames all load in a normal manner with the image being wider horizontally than than it is tall.  However, if I load a RAW uncorrected image it loads rotated 90 deg CCW (now taller than it is wide) which must be the reason for the error message.

This is very frustrating!  I've tried different Format Explorer DSLR_RAW settings but with no success.

I'm running PI ver 1.8 on a Dell laptop.

Any ideas?

Steve

[pfile]

i think the format hint "raw cfa" will cause the various processes to ignore the camera orientation flag in the exif. if you were not using that before, you should discard the masters and start again with that hint in ImageCalibration and ImageIntegration.

rob

[STEVE333]

Thanks Rob, I'll give it a try.  I'll let you know how it works.

Steve

[STEVE333]

Rob - Do I just type  -  raw cfa  - in the Format Hints area? Do I need any quotes?

Do I type it in the Input hints or Output hints or both?

Thanks

Steve

[pfile]

no quotes - input hints only. and only in processes where you are loading CR2 files directly... not necessary once youre working with xisf files.

rob

[STEVE333]

I tried the "raw cfa" in Input hints and ran ImageIntegration on my Bias files.  The output was a Gray file with no RGB color.  If I open one of the RAW Bias files it has a "purplish" color.  I believe the integrated file should still have the same color as the RAW image, just less noise.

Are there some settings in Format Explorer or somewhere else that I need to set?

Steve

[pfile]

that is the correct behavior. under certain configurations of DSLR_RAW, the file is debayered for you when it's opened and so you've got a 3-plane RGB file.

however, for calibration of OSC images, you need to keep all the calibration subs (and masters) in some kind of undebayered, raw format. that means CFA format (a greyscale image) or RGB Bayer (a 3-plane image; each channel looks like a checkerboard because the green and blue pixels are missing from the red channel, the blue and green are missing from the red channel, etc.)

only after you have calibrated your lights should you debayer them into RGB format. bias, dark, flats should never be debayered.

this is one possible flow for OSC images:

ImageIntegration (raw cfa) of bias CR2 files -> master bias in CFA format
ImageCalibration (raw cfa) of dark CR2s with master bias selected -> calibrated master dark in CFA format
ImageCalibration (raw cfa) of flat CR2s with master bias selected -> calibrated master flat in CFA format

ImageCalibration (raw cfa) of CR2 lights with master bias, master dark and master flat selected, "calibrate" unticked everywhere in the master sections -> calibrated lights in CFA format

BatchDebayer of calibrated lights -> RGB files of calibrated lights

StarAlignment of calibrated lights in RGB form -> registered lights in RGB format

ImageIntegration of registered lights in RGB format -> integrated image


rob

[STEVE333]

Rob - Everything works up to and including the Batch Debayering to create RGB images.

When I perform the StarAlignment and then ImageIntegration the Integrated (stacked) images don't align well with each star showing up as two stars displaced vertically from one another (one much brighter than the other).  I only used 6 images to speed up the testing, so, I didn't expect any stacking problems. 

I loaded the same 6 Debayered files into DSS and it stacked them with no misalignment.

Not sure why I'm having such a difficult time with this.

Steve

[pfile]

it could be that the stars are a little soft or a out-of-round enough that PI is not able to find them with the default StarAlignment settings. you can first try opening one of the problem subs, then switching StarAlignment to "Detected Stars" mode and apply SA to the image. you should get a new image showing which stars were detected.

another possibility is that you've picked up a boatload of hot pixels. the above test will show that. if that's the case you can increase the "noise scales" in the Star Detection tab of star alignment.

if you're seeing too few stars, you can tweak the Log (sensitivity) down to a larger negative number, and the Peak Response up to a higher number.

rob

[STEVE333]

Thanks Rob - I tried increasing the Noise scales from 0 to 2, and, it worked!  Perfect alignment for the 6 images.  I will now try it on all of the images.

Your help is much appreciated.

Steve

[pfile]

you're welcome ... that's good news! i forgot to mention there is also a hot pixel rejection control in there as well.

rob

[STEVE333]

Rob -Sadly, bad news.  I registered and integrated 85 (5 min exposure) images, and the result was a badly smeared final stacked image.

Not sure why it worked with 6 of the images but not with all 85.

I'm a little burned out on fighting this Preprocessing with PI.  For now I'll be sticking with DSS for the stacking but still processing with PI.

Thanks again for your help along the way.

Steve

[STEVE333]

Update - I went back and tried to stack a few images, and, although the stars didn't all line up the Hot Pixels did! 

In the StarAlignment process I changed the Hot pixel removal setting (in the Star Detection section) from the default value of 1 to 2 and left all other settings at their default values.  That solved the problem!!!  All 83 images stacked successfully with no smearing.  WHEWWWW!!!!! 

Hopefully this setting will work for my image in the future.

Steve

[pfile]

thats good, yeah i should have mentioned the hot pixel rejection control earlier in the thread... forgot about it.

rob

[STEVE333]

Thanks Rob - It feels like I just climbed a mountain!  However, the view is nice from here.  The noise in the PI stacked image appears somewhat less than in the DSS stacked image.  The PI stack has also removed several satellite trails that are visible in the DSS stack.

I have noticed that some of the stars in the arms of M51 that were red when processed with the DSS stack are no longer red with the PI stacked image (see attached images).  Any ideas what happened to the red, and, how to get it back? (Should this be a new thread?)

Steve