issues with drizzle

[georgian82]

Hi guys,

For some reason my drizzled image looks very strange...almost as if it had a film layer over it. The integrated (stacked) image looks good (the way I expect it to look like) but when I drizzle it 2x2 it looks weird. I have tried processing the data with and without local normalization that the results are the same.

Here is a link of both images in dropbox:

https://www.dropbox.com/sh/w8vjgbmzh8z6lht/AAAy82cgTMXT6wFH9y-rYULLa?dl=0

There is something very interesting that happens after I stretch the image. When I go to extract the lightness in order to create a mask, the lightness is not stretched at all. I am not sure if this is related to the issue I am having with drizzle.

Thank you for your input!

[drmikevt]

I am not exactly sure what is going on here, but:

 - I was able to stretch your image (using a standard STF transferred to Histo. Transformation, and then applied to the image). 

 - After stretching, I was able to exact a lightness, and it was stretched (non-linear) - like normal.

 -One issue is that you need to crop off the border areas of the image - use Dynamic Crop to do this. 

 - Another issue might be your subs.  Did you go through each sub visually (using the Blink process) to look for images that were all washed out?  It looks like you may have one or more subs that have captured too much sky glow, maybe from clouds?  Another way to do this is to open all images in SubFrameSelector and then look at the median and noise values - you should see the outliers.  Eliminate them, then integrate and drizzle the stack again. 

If you then crop the resulting image, I think all will be back to normal.  How were you trying to stretch your image?  Again, everything seems normal when I stretch it.  The border areas look like they don't stretch, but they do - they just have very little signal.  I hope this helps!

Mike

[georgian82]

Hi Mike,

Thank you for your reply and for taking a look at my data.

I am not sure why you had to stretch the image as it was already stretched. I stretched it using the screen transfer function before I saved it. When I try to extract the lightness out of it (after the image is stretched) the lightness image is dark and not stretched.

Does the drizzled image look right to you?

Yes, I am aware that the image needs cropping. This was a multiple night session, that's why there is discrepancy between the light frames. That being said, I have not bothered cropping it at this point because I am more interested in figuring out my other issues first.

I did go through every single light frame prior to uploading the data in PI. I did not use the Blink process though but I am sure all the frames were clean. Those two nights were extremely clear...no chance of even a single cloud passing by and I image from a fairly dark site so the chance of skyglow is minimum with my setup at 3 minutes and f/4. Having said all of that, I will go through the data again just in case I missed something.

Thanks again!

Sebastian

[pfile]

STF is, as the name implies, only for screen display. it does not permanently stretch the image. you have to apply the HistogramTransformation tool to the image to really stretch the data.

XISF, being a pixinsight-native format, can remember that a file had the STF turned on and applied when it was saved. so when you close and reopen it, it appears to be stretched, but in fact it is still in a linear format.

rob

[drmikevt]

Yes, so, just to check myself I re-downloaded one of your images. You will see that when you open it in PI, there is a green line next to the name tab on the left of the image.  This tell you that the image is linear, i.e. - it has not been stretched.  In this case, we would expect the extracted L image to also be linear (not stretched).  If you saved the image as a .fit instead of .xsif, you would see that when you re-open it, it will be black. 

If you have questions or confusion around the stretching of images, the tutorial about this on the LightVortexAstronomy website will help tremendously. 

Mike

[georgian82]

Hi Mike,

I mistakenly thought that I had saved the STF by applying it to the histogram as I usually do (but obviously I didn't)...thanks for pointing that out.

In any case, non of this explains the issues I am having with the drizzled image.

I am currently redoing everything from the start as I suspect the issue might be in the way I calibrated the dark frames. We'll see...

Cheers,

[drmikevt]

Good luck - its hard to say if the drizzled image looks right without seeing the difference between it and your integrated image.  Does it look hazy?  Sure it does, but its a very dense star field.  I don't see any major artifacts, though (but you have some funky diffraction spikes and bit of a coma issue).  If you still have trouble, post both the drizzle and integrated image, and your subs if possible - that might allow folks to better help.  And again, make sure to use Blink - making sure that you are applying the same histogram to all images (rather then applying a custom one to each).  If you looked at the images using a tool that automatically applies an STF, then you can easily miss when one image is much brighter then the others because the auto-STF will try to make it look the same. 


Mike

[georgian82]

Hi Mike,

Thanks for the input. The link I posted is for the integrated image and for the drizzle of that integrated image. They look nothing like it...

Anyways, I will keep working on it.

Thanks again.

[sharkmelley]

I took your non-drizzled image (which is linear data), removed the STF and gave it a quick ArcsinhStretch.  The resulting screenshot is attached and the colours look fine.

On the other hand your drizzled image was already stretched i.e. non-linear and the colour balance appears completely wrong.  Again I've atttached a screenshot.  It is impossible to fix the colour balance in its current non-linear state.   In addition, the red and blue channels appear to have become clipped in the background sky areas - that's why parts of the red nebulosity are missing in the outline of the Heart Nebula. What processing steps did you perform on the drizzled data?  That's probably where the problem lies. Do you have the original linear drizzled data?

Mark

[georgian82]

I took your non-drizzled image (which is linear data), removed the STF and gave it a quick ArcsinhStretch.  The resulting screenshot is attached and the colours look fine.

On the other hand your drizzled image was already stretched i.e. non-linear and the colour balance appears completely wrong.  Again I've atttached a screenshot.  It is impossible to fix the colour balance in its current non-linear state.   In addition, the red and blue channels appear to have become clipped in the background sky areas - that's why parts of the red nebulosity are missing in the outline of the Heart Nebula. What processing steps did you perform on the drizzled data?  That's probably where the problem lies. Do you have the original linear drizzled data?

Mark

Hi Mark,

Thanks for taking a look at the data...I appreciate it!

Unfortunately, I don't have the original linear drizzled data...

These are the steps I took along the way before drizzling, following Lightvortex's tutorial (which I have been using successfully in version 1.8.4):

1) Make superbias
2) Calibrate dark frames with superbias
3) Integrate calibrated dark frames to create a master dark
4) Calibrate light frames with superbias and master dark
5) Debayer calibrated light frames
6) Register calibrated/debayered light frames (clicking the option for generating drizzle data)
7) Integrate calibrated/debayered/registered light frames (adding the drizzled data)

After all of that I do drizzle 2x2

I really don't understand why would the drizzled image look any different from the non-drizzled...I have tried different settings during the calibration step and also the integration step and nothing seems to work.

Let me know your thoughts.

Thanks again!
Sebastian

[sharkmelley]

Unfortunately, I don't have the original linear drizzled data...

Let me know your thoughts.

My thoughts are: never throw away important data, especially when you are trying to track down a problem 

Can you create a new 2x2 drizzled stack, so we can take a look?

Mark

[georgian82]

Unfortunately, I don't have the original linear drizzled data...

Let me know your thoughts.

My thoughts are: never throw away important data, especially when you are trying to track down a problem 

Can you create a new 2x2 drizzled stack, so we can take a look?

You are very right about that! Lesson learned for sure. I have tried integrating this image so many times with different settings that I was compiling too much data so I decided to throw it all away and start fresh.

Let me work on it and I will upload new images as soon as I can. If there are any particular settings that you think I need to be using during calibration or integration (for pixel rejection for example) please let me know.

Cheers,

Mark