Anbody seen this before?

[Allinthehead]

Obviously some sort of stacking artifact but i've not seen it before. Any ideas?

[vlaiv]

Have no clue, but it does remind me of JPEG compression.

Base of jpeg compression is 8x8 block (or maybe 16x16 or even adjustable). Algorithm uses DCT over such block (akin to Fourier Transform) to separate different frequencies and then applies loosy transform based on frequencies. Sometimes in very compressed JPEG images you can see such blocks in background if conditions are right. Here is example of it:

Each block is in effect 8x8 pixels (and you can see individual pixels in some blocks) - but it looks pixelated.

Not sure if it's related to above artifact, just reminded me of it.

What is the size of artifact in pixels?

[Allinthehead]

5 minutes ago, vlaiv said:

What is the size of artifact in pixels?

Hi Vlaiv. It's 34 by 34 pixels. I should have added that i through 60 second and 120 second subs into the same stack.

[vlaiv]

Did you have excessive rotation of set of subs to align them?

This might be due to aliasing artifacts, although it would not have so sharply defined edges.

If this gets out of stacking like this, first thing to try is to change interpolation method for registration. Do you know what interpolation are you using now?

 

[Allinthehead]

No excessive rotation and Mitchell-Netravali for interpolation. I've tried stacking the 120 and 60 second subs separately and no issues.

[vlaiv]

Just to be clear, as I'm having trouble following how it appeared.

Above artifact appears in stack when you combine multiple exposures, but when you do separate stacks each containing exposures of particular length - you can't notice this artifact right?

What stacking method are you using? And also if you do any sort of "normalization" of frames - it could be related to it.

[Allinthehead]

3 minutes ago, vlaiv said:

Above artifact appears in stack when you combine multiple exposures, but when you do separate stacks each containing exposures of particular length - you can't notice this artifact right?

Exactly

3 minutes ago, vlaiv said:

What stacking method are you using?

Local Normalisation winsor clip in APP. I'm redoing the stack as we speak without the LN

[vlaiv]

1 minute ago, Allinthehead said:

Local Normalisation winsor clip in APP. I'm redoing the stack as we speak without the LN

I would bet that LN is the problem.

[Allinthehead]

1 minute ago, vlaiv said:

I would bet that LN is the problem.

I think you may be correct. Thanks for the replies and i'll post an update when complete.

[Allinthehead]

1 hour ago, vlaiv said:

I would bet that LN is the problem.

Bang on. I changed my rejection to sigma 

[Xiga]

Glad you got this sorted Richard.

ps - You should post this over on the APP Forums. Mabula will no doubt look into it in more detail and it could well lead to a fix in the next version.