AutoStakkert woes

[DavidWright2010]

I'm following the step-by-step instructions that Jon555 wrote on cloudynights, and I'm having a problem. I first tried just 4 Moon images (first quarter Moon, DLSR images, all high quality). The workflow with pipp and AS worked OK, although I had missed a checkbox that converted color to monochrome. Even with just 4 images, the S/N improvement was obvious.

So then I did the 14 images I had at hand (going forward, I could easily have 50-100 images). No problems were encountered, and at first glance the stacked image was ok, but looking more closely, there were vertical smears maybe a few hundred pixels long, and some craters were doubled.

LHS is a single frame, RHS the stack:



Here's the full-sized output file (FWIW, it's not centered in the frame, like all the sub were!?)



Any thoughts?

David

 

[Jeffry7]

Hi David,

I am not sure about the streaking. The only time I have seen a similar artifact it was because I was running on Linux and had misconfigured Wine, a program to allow Windows applications run in Linux. I am guessing you are not running on Linux?

As for the other issue, are you taking these pictures on an equatorial mount? If not then you may be suffering from field rotation.

In a nutshell, unless you are on an equatorial mount, as you take pictures the moon will appear to rotate in your images. The longer you take pictures, the more rotation will be apparent. Autostakkert does not take this into account. (There is an advanced option that is supposed to help with this, but it has never worked well for me.)

You may be able to tell just by looking. Open up the first and last image from your set and compare them.

 

[nighthiker]

Hi!

I haven't had that problem and do most of the processing with AutoStakkert. I don't use PIPP as Autostakkert can also do the centering. I haven't had any problems with it. The only exception were some old - low quality - images from a lunar eclipse taken over several minutes. So that problem is related to me.

I'm taking several hundred images from a fixed tripod but in a short interval (1-2 min) to prevent the problems with the rotation, using a Canon R with 100-400/4.5-5.6 plus 2x TC. I pre-cut the area of interest in the RAW converter and give the converted TIFF's to Autostakkert which does the centering of the images. Please see the - long - video from Emil on his Autostakkert homepage for details. The 'double structures' are probably a problem related to bad centering of the moon images and/or the alignment. See the video of how to select and QC the selection of alignment points. Have you checked that PIPP is doing a proper job? Autostakkert has several options for doing the image stabilization and centering.

 

[DavidWright2010]

Hi David,

I am not sure about the streaking. The only time I have seen a similar artifact it was because I was running on Linux and had misconfigured Wine, a program to allow Windows applications run in Linux. I am guessing you are not running on Linux?

Windows

As for the other issue, are you taking these pictures on an equatorial mount? If not then you may be suffering from field rotation.

Equatorial - Star Watcher EQ6-R Pro

In a nutshell, unless you are on an equatorial mount, as you take pictures the moon will appear to rotate in your images. The longer you take pictures, the more rotation will be apparent. Autostakkert does not take this into account. (There is an advanced option that is supposed to help with this, but it has never worked well for me.)

You may be able to tell just by looking. Open up the first and last image from your set and compare them.

There is no progressive image rotation.

See my other update below.

David

 

[DavidWright2010]

Hi!

I haven't had that problem and do most of the processing with AutoStakkert. I don't use PIPP as Autostakkert can also do the centering. I haven't had any problems with it. The only exception were some old - low quality - images from a lunar eclipse taken over several minutes. So that problem is related to me.

I'm taking several hundred images from a fixed tripod but in a short interval (1-2 min) to prevent the problems with the rotation, using a Canon R with 100-400/4.5-5.6 plus 2x TC. I pre-cut the area of interest in the RAW converter and give the converted TIFF's to Autostakkert which does the centering of the images. Please see the - long - video from Emil on his Autostakkert homepage for details. The 'double structures' are probably a problem related to bad centering of the moon images and/or the alignment. See the video of how to select and QC the selection of alignment points. Have you checked that PIPP is doing a proper job? Autostakkert has several options for doing the image stabilization and centering.

I will look at video.

The centering out of pipp looks pretty good.

I guess I don't understand the quality score. My images ranged from 100% to 81.24% (clearly normalized values, which means that the app 'decides' which is the best.) However, examining the images shows me that image 7338 (quality 82.31%) is the best and 7345 (quality 97.95%) is the worst. See for yourself:



If I remove the 3 worst from the stacking list, the doubled craters are now OK, but the streaks remain:



I had plenty of alignment points (4200).

I don't understand this. Doesn't AS choose the ''best" feature when there are several versions (due to turbulance)? What to the quality numbers mean?

I get much better results in a focus-stacking app (Helicon Focus 7, nfi) in that it chooses the sharpest feature. But Helicon Focus does not signal average.

BTW, I'm seeing (relatively) big focus changes in this example b/c I didn't just fire off 14 exposures - I shot 3 or 4, waited a bit, adjusted focus, shot a few more, and so on. If I thought that was the issue, I could re-shoot using the interval timer and get 20-30 images over the space of 20-30 seconds.

David

 

[atigun]

I found that the key to avoid those vertical/horizontal artifacts in Autostakkert was to set the area for the "alignment points" to maximum (400 pixels) and use surface mode for the stacking:

https://www.dpreview.com/forums/post/64497983

 

[DavidWright2010]

I found that the key to avoid those vertical/horizontal artifacts in Autostakkert was to set the area for the "alignment points" to maximum (400 pixels) and use surface mode for the stacking:

https://www.dpreview.com/forums/post/64497983

I'm using AS 3.0.14 (x64). The max size is only 200, but makes things much worse - like in no alignment at all.

Surface stabilization (instead of planet) is better - only a few 'streaks', but the stacked image seems to be about the sharpness of the worst sub-exposure. Aside from the noise issue, a single image is far better.

I think I'll take another data set, via interval exposures. The individual exposures should be more uniform.

Thanks for your help.

David

 

[nighthiker]

Hi!

Surface stabilization should be used if the whole frame is only a surface (e.g. part of the moon), not a whole planet, moon, sun. There is somehow an option to prevent AS from re-aligning. So if your pics are already aligned e.g. by PIPP or manually, you can switch that off and let the fine-adjustment do by the alignment point. It's also helpful to have different-size alignment points and I often add some manually in 'empty' areas.

I normally do a testing with e.g. 12%, 25%, 50% and 75% (plus sharpening) of all input images for stacking and check for the best result. I also often use the drizzle x1.5 option for the final run.

 

[DavidWright2010]

Hi!

Surface stabilization should be used if the whole frame is only a surface (e.g. part of the moon), not a whole planet, moon, sun. There is somehow an option to prevent AS from re-aligning. So if your pics are already aligned e.g. by PIPP or manually, you can switch that off and let the fine-adjustment do by the alignment point. It's also helpful to have different-size alignment points and I often add some manually in 'empty' areas.

I normally do a testing with e.g. 12%, 25%, 50% and 75% (plus sharpening) of all input images for stacking and check for the best result. I also often use the drizzle x1.5 option for the final run.

I will look for that “no alignment “ option.

David

 

[nighthiker]

Hi!

Surface stabilization should be used if the whole frame is only a surface (e.g. part of the moon), not a whole planet, moon, sun. There is somehow an option to prevent AS from re-aligning. So if your pics are already aligned e.g. by PIPP or manually, you can switch that off and let the fine-adjustment do by the alignment point. It's also helpful to have different-size alignment points and I often add some manually in 'empty' areas.

I normally do a testing with e.g. 12%, 25%, 50% and 75% (plus sharpening) of all input images for stacking and check for the best result. I also often use the drizzle x1.5 option for the final run.

I will look for that “no alignment “ option.

David

Click 'Surface' and 'Disable Stabilization'.

 

[DavidWright2010]

Hi!

Surface stabilization should be used if the whole frame is only a surface (e.g. part of the moon), not a whole planet, moon, sun. There is somehow an option to prevent AS from re-aligning. So if your pics are already aligned e.g. by PIPP or manually, you can switch that off and let the fine-adjustment do by the alignment point. It's also helpful to have different-size alignment points and I often add some manually in 'empty' areas.

I normally do a testing with e.g. 12%, 25%, 50% and 75% (plus sharpening) of all input images for stacking and check for the best result. I also often use the drizzle x1.5 option for the final run.

I will look for that “no alignment “ option.

David

Click 'Surface' and 'Disable Stabilization'.

I took a new image series, with all 16 pictures taken within 20 seconds, then I skipped pipp and let AS do everything. This worked without 'smears' in the composite image, and seems to correctly choose the sharpest portions of each image:



I also did wavelet sharpening using RegiStax 6.

So, Autostakkert is working for me now. Thanks to everyone who offered help.

David