lights to stack: 75%

Yep, that is what APP is doing. It is throwing away almost all of my blue frames when I select 'lights to stack: 75%'.

This is not right. It forces me to go through the file list and sort them by quality and then uncheck the ones with the lowest scores. Not very friendly.

You could process the Red, Green and Blue separately in that case, then it will just stack 75% of the best red, green or blue frames. Then you use the combine tool to combine them.

Thanks for your reply.

I tried that already. APP makes the dimensions of each stack slightly different which causes APP to fail when I try to combine them. StarTools also doesn't like it.

What you could do is load them all in, register them together, save the registered frames and then process them separately. Doesn't that work? I never tried it like this myself though, but it's worth giving that a shot.

I could give that a try. Thanks!

Hope it helps! If not, don't hesitate to ask again and we'll have a deeper look into it.

Vincent,

Last night a fellow AP'er recommended that I not rely on image quality algorithms. Instead he said that I should visually inspect each frame and discard those that don't meet my exacting standards! So this morning I did as he said. I reduced 16 frames in each LRGB channel to 12. I went through the process and integrated 100% of them and I found that there was no noticeable difference with respect to star roundness! The only difference is that the 16-stack is brighter than the 12-stack which is understandable.

I think I understand what is going on. One of the first tutorials I watched was the first by Sara Wager. She made it a point to change the "weights" control from "equal" to "quality" in the Integrate module but she kept "lights to stack" at 100%. This did not make sense to me at the time since my experience is with DSS. I now understand that APP will use the quality score of each light frame to weight the contribution it makes to the stack.

I will still follow my friend's advice and discard frames but I will limit it to those that are truly awful. I still maintain that adjusting "lights to stack" to anything less than 100% can lead to color imbalance.

Thanks for your help.

Manually selecting the frames is fine as well ofcourse. Personally I never do that anymore though, during the acquisition phase I already reject any sub that's not properly guided within pretty strict tolerances (no interest in anything less than strict as I always throw them away during stacking anyway and it saves time to spend on a good sub). After that the quality setting in APP is really good in taking all kinds of statistics and throw away the last few frames that are a bit less still, I tend to stack 90-95% as I already have most frames at least very round. Other software packages aren't as thorough with automatic statistics I notice, or more difficult. This workflow I use has never failed me and doesn't require a lot of extra time.

But in the end, it's all about what you want. 🙂

Everyone's equipment is different. For instance I am using a 50-year old GEM, no go-to, just a DIY stepper motor on the RA axis with PEC. "Passive guiding" I call it. I rely on the accuracy of my polar alignment and periodic error correction. I can get 200s exposures but I like to keep it under 60s. Perhaps one day I'll get a modern mount.

Ah right, that doesn't work then indeed. In that case errors can be quite big between frames. I still think doing that automatically will work just fine due to the very good statistics in APP, it's really different (more advanced) from other packages (Mabula spend a huge amount of time on the maths behind it all and it was the basis for the rest of the program). Either way, manually or automatic will work though. Good luck with the rest of the processing!

I appreciate it. APP is so much better than DSS. My images are much crisper.

Hi Vincent,

Re: "What you could do is load them all in, register them together, save the registered frames and then process them separately. Doesn't that work? I never tried it like this myself though, but it's worth giving that a shot."

Alas, so close but still failure due to different dimensions. I was "this" close to success but for some unknown reason APP decided to make the blue stack one pixel greater.

All of my images are 1920 x 1080. I followed your directions for creating and saving registered lights for all channels. All of the registered lights, everything single one, is dimensioned 2038 x 1163. Then according to your directions I processed each channel by loading the registered lights. Everything was working perfectly until I got to the blues. The dimension of the blue stack is 2039 x 1164. All of the other stacks are 2038 x 1163!

So then I thought maybe I could sort each channel by quality and uncheck the worst 10% but alas when I click the quality column header there is no sorting performed.

Oh well there are other ways of doing this that make it increasingly complex.

Hi @oopfan and @vincent-mod,

I hope I explained that properly. Allow me to clarify. I am not asking for separate percentages. I am saying that if I have 100 L frames, 100 R frames, 100 G frames, and 100 B frames, I want APP to accept 75 L frames, 75 R frames, 75 G frames, and 75 B frames. I don't want 100 L frames, 100 R frames, 100 G frames, and 0 B frames!

Thank you for bringing this forward, I will make a note of this to correct this in either multi-channel and/or mult-session mode 😉

I rarely throw away images. Only when they are very bad when compared to the other images. And the analytical results in APP are robust in this regard, that they will show quickly which frames are bad.

In most cases, I integrate with the quality weights, then the influence of bad frames is greatly reduced, shar shapes are hardly affected and by still keeping these frames, you are still reducing overall noise in your integration. So I would actually advise not to throw away too many frames, especially if they are only slightly worse than the others. Off course, if you want the best possible stars, you need to throw away all frames with stars that are not perfectly round and sharp. On the other hand, very slightly (usually not visible) worse stars with clearly lower noise is a good choice I would think.

Visual inspection of all images is something I actually not do... because there really is no need for this in APP, at least I think so. If an integration does shows a weird artefact, just by looking at the analytical results by plottting these in graphs (right mouse click in the bottom panel) you will find the culprit very fast usually 😉

Mabula

Hi Mabula,

My situation is unusual in that I have a 50-year old mount. I have a stepper motor on the RA axis but nothing on DEC. My Pi code incorporates periodic error correction. I rely on accurate polar alignment primarily. No active guiding. I can get round stars but still need to throw out 10%-20% of the frames to get the best results.

When I do single-channel imaging, like SN 2019np in Luminance, I can rely on APP to reject 10%-20% by quality. The problem is multi-channel. APP has a tendency to reject more of the blue subs because of the low SNR of blues.

Thanks for great software!

Brian

Mabula,

I rarely throw away images. Only when they are very bad when compared to the other images. And the analytical results in APP are robust in this regard, that they will show quickly which frames are bad.

You said "they will show quickly which frames are bad." Is there a way to export the file list + quality score to a spreadsheet? If that were possible then I could sort by quality and then channel.

Posted by: oopfan

Mabula,

I rarely throw away images. Only when they are very bad when compared to the other images. And the analytical results in APP are robust in this regard, that they will show quickly which frames are bad.

You said "they will show quickly which frames are bad." Is there a way to export the file list + quality score to a spreadsheet? If that were possible then I could sort by quality and then channel.

 

 

Hi @oopfan,

I will fix the sorting of the frames as well as part of your request here:

I hope I explained that properly. Allow me to clarify. I am not asking for separate percentages. I am saying that if I have 100 L frames, 100 R frames, 100 G frames, and 100 B frames, I want APP to accept 75 L frames, 75 R frames, 75 G frames, and 75 B frames. I don't want 100 L frames, 100 R frames, 100 G frames, and 0 B frames!

If you are working in multi-channel / multi-session mode, the quality selection slider should work per filter/session as you indicate.

With the right mouse button on the frame list panel, you can sort on the analytical results. This then should also work per filter/session, right ?

Mabula

Hi Mabula,

With the right mouse button on the frame list panel, you can sort on the analytical results. This then should also work per filter/session, right ?

Yes, that would be very helpful.

I will fix the sorting of the frames as well as part of your request here:

Yes, thank you. I think there is one other user who asked for this. In retrospect I hope we are not crippling your design. I would like you to do this because you see value in it. I come here with a set of biases I learned from other software 🙁

Perhaps if I present this example I can learn a bit more of the internals of APP:

Really simple scenario: 1 frame in each of the 4 channels having these quality scores: L: 40, R: 20, G: 10, B: 5. When I combine these four stacks using the RGB Combine tool and I choose 100% contribution in each channel, will that single frame in the B channel contribute 25% to the final image, or 6.67%, calculated as 5/(40+20+10+5)? If the answer is 6.67% then won't my image have a yellow cast?

Thank you.

Mabula,

I continued working through the tutorial videos last night and discovered that there is in fact a way to sort the file list by quality. On the one hand, videos are nice but on the other hand they are very much sequential and not searchable (but there are exceptions that I've seen elsewhere.) Either way I found how to sort by quality. The reason why I didn't discover it sooner is that, in my opinion, the user interface is non-standard and counter-intuitive. One does not expect to right-click on an item in a list and get options that operate on the entire list.

So I have been working on NGC 4565. I've got a total of 74 minutes of lights in LRGB. Previously I integrated all stacks 100% weighted by quality. The result was very encouraging, much better than I could get with DSS and StarTools, but as of yet the final image is still too noisy for sharing -- more image capture is required.

This morning I experimented with sorting the file list by quality. It was a little awkward but I was able to deselect the worst 10% in each channel. A lot of those frames had very elongated stars due to poor tracking. When I completed registration and integration and then verified that in fact 10% of the total integration time was taken evenly out of each stack, I combined RGB and proceeded with the final image processing steps.

I have to admit that you are performing some magic. I cannot tell the difference between the original 74m integration and the new 67m integration after removing the worst 10% of frames evenly across all channels. I was expecting dramatically rounder stars, better detail in the dust lane, at the expense of a slight increase in noise. In my opinion it is not worth the effort I went through. APP is magic!

Thanks.

Posted by: oopfan

 

Perhaps if I present this example I can learn a bit more of the internals of APP:

Really simple scenario: 1 frame in each of the 4 channels having these quality scores: L: 40, R: 20, G: 10, B: 5. When I combine these four stacks using the RGB Combine tool and I choose 100% contribution in each channel, will that single frame in the B channel contribute 25% to the final image, or 6.67%, calculated as 5/(40+20+10+5)? If the answer is 6.67% then won't my image have a yellow cast?

Thank you.

Hi @oopfan,

Very good question and scenario, and you are right, I need to fix this as soon as possible for better image selection in the multi-channel/session modes.

Thanks 😉

Mabula

Posted by: oopfan

Mabula,

I continued working through the tutorial videos last night and discovered that there is in fact a way to sort the file list by quality. On the one hand, videos are nice but on the other hand they are very much sequential and not searchable (but there are exceptions that I've seen elsewhere.) Either way I found how to sort by quality. The reason why I didn't discover it sooner is that, in my opinion, the user interface is non-standard and counter-intuitive. One does not expect to right-click on an item in a list and get options that operate on the entire list.

So I have been working on NGC 4565. I've got a total of 74 minutes of lights in LRGB. Previously I integrated all stacks 100% weighted by quality. The result was very encouraging, much better than I could get with DSS and StarTools, but as of yet the final image is still too noisy for sharing -- more image capture is required.

This morning I experimented with sorting the file list by quality. It was a little awkward but I was able to deselect the worst 10% in each channel. A lot of those frames had very elongated stars due to poor tracking. When I completed registration and integration and then verified that in fact 10% of the total integration time was taken evenly out of each stack, I combined RGB and proceeded with the final image processing steps.

I have to admit that you are performing some magic. I cannot tell the difference between the original 74m integration and the new 67m integration after removing the worst 10% of frames evenly across all channels. I was expecting dramatically rounder stars, better detail in the dust lane, at the expense of a slight increase in noise. In my opinion it is not worth the effort I went through. APP is magic!

Thanks.

Hi @oopfan,

Thank you very much

Tthat is the beaty of good analytical calculations to derive the quality score per frame, (which is based on noise, star density, star shape (size and roundness) ) and then using that for integration weights.

I my own experience with APP, I rarely stack less than 100% of my frames I only remove extreme outliers which are easily found when sorting the list on quality, noise or star shape.

Cheers,

Mabula