My first "real" project - M81 - strange vertical noise lines

B

bennich

Member
Hi All
I am working on my first project with M81 and are making good progress....I think.
I am imaging with an unmodified Canon EOS50D and are trying hard to get the correct dark frames with the correct temperature, flats and dark flats.
Currently I am processing with WBPP and matching my dark frames with my light frames, within a few degrees celsius.
My issue is that I keep seeing these vertical lines in my stacked image. I was diggin' around on the forum and came across the "CanonBandingReduction" script and tried it on my image. Without too much luck.
I started looking through my masterframes, and found the same vertical lines in my Dark frames. I now ran the "CanonBandingReduction" on those - and managed to completely remove the lines both vertical and horizontally.
Now I was hoping to reintegrate my data - WITHOUT - the aforementioned annoying lines - BUT NO!!!!
They are still there - I don't get it :cry:
Any kind of input would be greatly appreciated.
I have dropped my master data and the latest stack here: https://drive.google.com/drive/folders/1aVPDrw5f5B1to7I3i3no8uY81symrrhw?usp=sharing

I have also added my DBE script, to save you a bit of time if you want to check it out.
 

Attachments

  • Original_WITH_MATCHED_DARK_TEMPS_AND_CLEANED_DARKS_DBE_SPCC.jpg
    Original_WITH_MATCHED_DARK_TEMPS_AND_CLEANED_DARKS_DBE_SPCC.jpg
    482.2 KB · Views: 98
it's hard to do on the subframes because the banding is so subtle. and its a little hard to do on the master light because the image registration process will likely have rotated the subframes frames a little bit here and there.

this is pretty typical for certain canon cameras, unfortunately, so making CBR work on the master light is probably the only way out of this.

i'll look at your image and see.
 
can you upload the integration of the non-cleaned subexposures? this could have made the problem worse.
 
a couple of things - the script is designed to remove horizontal banding, so you have to rotate your image by 90 degrees first.

however, after that the banding is not perfectly horizontal. the script will mess up the image, or not do anything, if the banding is not almost exactly horizontal. i had to use dynamiccrop to rotate the image by 0.85 degrees to make the bands mostly horizontal. 0.85 degrees was not perfect, but close. it could be that there is no perfect rotation because the individual subs were all rotated by slightly different amounts during registration.

i then ran CBR with strength = 1.7 and highlight protection of 1.5 and it mostly cleaned up the image.

i think you need to go back to the non-CBR darks/lights and then run CBR on the original image.

rob
 
HAHA, It actually seems that if I overdo it first with CBR - setting amount = 2.7 and highlight protection = 4.5 and THEN run it again with amount = 1 and highlight protection = 2 - it becomes really good with an absolute minimum of banding.
Will test a bit more
 
Last edited:
I reintegrated the data, without cleaning up any darks with CBR.

The result is disturbingly similar to the integration with "cleaned" darks, with only slight variations in the distribution of the banding and coloring.
Applying CBR as described in the above message, even it out even further.

Data is here: https://drive.google.com/drive/folders/1aVPDrw5f5B1to7I3i3no8uY81symrrhw?usp=sharing
The reintegrated file is called: REINTEGRATED_NO_CLEANED_DARKS.masterLight_BIN-1_4770x3177_EXPOSURE-600.00s_FILTER-NoFilter_RGB_autocrop
 
You must log in or register to reply here.
Back
Top