Canon Sensor Banding Removal

[thewormofautumn]

Hi all

Just purchased StarTools and as a total newbie to this hobby, I love how much easier (even if I don’t entirely understand it!) it is compared to Photoshop.

I’m processing an image taken last night of Orion, and can’t seem to get rid of the sensor banding from my Canon 5D IV.

I’ve followed the excellent official video here:
https://www.youtube.com/watch?v=IDY3ov2dQe4
But this is what I’m seeing when I create the mask.

First, the autodev:

Second the Wipe preview (Dark Anomaly Filter at 4). Looks even worse!

So I cancelled it, and cropped away the edges, and added a mask

And then did the Wipe again:

I'm not seeing anything remotely looking like a good image here at this stage

The bands are still there, but when I watch the video above, they vanished. Just trying to figure out how (if possible) to get rid of these. I’ve edited the file through to conclusion, and can get some good results, but those bands remain.

The TIF is from a 109 x 2min subs, ISO1600 with darks (46), flats (30) and bias (29) all included

Help appreciated!

edit: apologies, didn't screencapture the mask

[Burly]

Maybe you could post file your trying to process using dropbox so others can have a better look .

regards dave

[thewormofautumn]

Hi Dave

No problem, here you go:

https://www.dropbox.com/s/glt9om2ecir3aa5/ST.tif?dl=0

[admin]

Hi,

I just had a quick peek at the dataset.

Can you tell us how was this stacked? Are you 100% sure this dataset has not been stretched before?

The flats appear to cause some anomalies in the corners and different channels; something is very wrong here.

Normally, the Synthetic Dark/Bias feature in the revamped Wipe module would be able to take care of the banding quite well.
It still does to a degree - it takes care of the central banding (I used the Defective Sensor Rows setting), but less so for the top;

StarTools_2752.jpg (260.15 KiB) Viewed 5780 times

In the end, I settled for this;

StarTools_276.jpg (301.9 KiB) Viewed 5780 times

Note that I didn't even use a mask here. After a more reasonable stretch (rather than the intentionally harsh diagnostics stretch in Wipe), you should get a more reasonable image, though the top "bar" may still show up.

[thewormofautumn]

Hi Ivo

Thank you for looking into the data. I don't know why I can't get rid of those bands from stacking (I thought that's what the point was!)
Below are my settings in DSS

and

It's good you managed to remove the middle band, but I guess I'll have to go back to the drawing board and figure out what happened

I would have loved to have this as an image: I'm new to this, and it was my first time shooting Orion and only second time shooting a nebula seriously.

Also, is there any "performance" improvements coming down the line? I have a 3900X, 32GB RAM and each step does take a long time to do (don't know if I'm being unrealistic)

Thanks!

[admin]

Thank you for looking into the data. I don't know why I can't get rid of those bands from stacking (I thought that's what the point was!)

Stacking and calibration, indeed.
Dithering should help, but maybe you can have a look at a single frame and see what the banding looks like in a single frame?

I would have loved to have this as an image: I'm new to this, and it was my first time shooting Orion and only second time shooting a nebula seriously.

It's pretty good, deep data otherwise and that's a text-book M42/Running Man. Great coloring with the teal/green O-III emission in the core. Excellent.
Hopefully (barring any serious issues with your camera) it's just a question of sorting out your calibration frames (flats and bias) + dithering, and you should be golden!

Also, is there any "performance" improvements coming down the line? I have a 3900X, 32GB RAM and each step does take a long time to do (don't know if I'm being unrealistic)

Welcome to AP processing. StarTools is an unapologetic resources hog and will happily bring a Ryzen 9 or i9 to its knees with its super heavy processing. Basic Photoshop this is not...

However, the upcoming 1.7 version will be offloading many (suitable) operations to your GPU. Even something modest like an RX 570 will blow past your 3900X for something like deconvolution. So, if you haven't done so, give the 1.7.x versions a try. Virtually all modules in 1.7 are accelerated in some way. How much depends on the nature of the algorithm (e.g. GPUs suck at simple if-then-else branching, but chew through taking the square root of a number like there's no tomorrow). You can read more here.

Let us know how you go!

[thewormofautumn]

Stacking and calibration, indeed.
Dithering should help, but maybe you can have a look at a single frame and see what the banding looks like in a single frame?

Thanks. I'll look at dithering next year. I don't have a guide scope yet, just trying to learn and master one thing at a time
I did take a few frames from the lights/darks/bias and flats to see what the banding looked like. Weirdly it didn't exist so much in the darks or bias. Don't know why, it's like it was a second camera.

It's pretty good, deep data otherwise and that's a text-book M42/Running Man. Great coloring with the teal/green O-III emission in the core. Excellent.
Hopefully (barring any serious issues with your camera) it's just a question of sorting out your calibration frames (flats and bias) + dithering, and you should be golden!

Thank you It's a fun journey! I cleaned that band up in Photoshop's healing tool in the end. Not ideal as it shifts star positions, but it did remove the banding so I guess that's something positive to come from this.

Welcome to AP processing. StarTools is an unapologetic resources hog and will happily bring a Ryzen 9 or i9 to its knees with its super heavy processing. Basic Photoshop this is not...

However, the upcoming 1.7 version will be offloading many (suitable) operations to your GPU. Even something modest like an RX 570 will blow past your 3900X for something like deconvolution. So, if you haven't done so, give the 1.7.x versions a try. Virtually all modules in 1.7 are accelerated in some way. How much depends on the nature of the algorithm (e.g. GPUs suck at simple if-then-else branching, but chew through taking the square root of a number like there's no tomorrow). You can read more here.

Let us know how you go!

Awesome to hear! I'll give it a go. I've got a 2070 Super ready and willing!!!
Great program

[thewormofautumn]

Just bringing this one back again. Seems that my sensor has these bands by default. Not much I can do besides use a different camera.

Here are stretches of my 3 main DSLRs.

5D IV on the left, 1D IV in the middle, and 5D II on the right
Both 5Ds have the issue. The 1D has a weird green tinge on the bottom. Ugh

Trying to not spend money right now as well, but this doesn't help.

[admin]

Just bringing this one back again. Seems that my sensor has these bands by default. Not much I can do besides use a different camera.

Here are stretches of my 3 main DSLRs.

5D IV on the left, 1D IV in the middle, and 5D II on the right
Both 5Ds have the issue. The 1D has a weird green tinge on the bottom. Ugh

Trying to not spend money right now as well, but this doesn't help.

Hmmm... that sucks. Though if the lines are always in the same place, bias frames should take care of that. Also, dithering between frames should spread out the errors. Hopefully the Wipe module can help as well as your last line of defence!