7th December 2023: added payment option Alipay to purchase Astro Pixel Processor from China, Hong Kong, Macau, Taiwan, Korea, Japan and other countries where Alipay is used.
Poor quality result
Hi, I'm completely new to APP and I have a question about filtering based on image/star quality.
I loaded my observations (FITS) and used the Analyze Stars section to measure my 20 subframes. I know that some of the frames are not great quality and I wanted to find a way to get APP to reject the bad frames for me.
On the Integration section, I set 'integrate' to average and 'weights' to quality. I also set the 'filter' to adaptive rejection and left everything else at the default. I have attached the resulting image - the stars are very poor quality and there are artifacts throughout.
This is only a 20-frame integration but I'd like to integrate hundreds of frames so I need a reliable way to reject the outliers. How can I do this in APP? (I realize I need good quality data however, I'd like to take advantage of auto-filtering because I have hundreds of frames)
Ok, I tried a few things and have a better result (attached), but there are still artifacts throughout the image. Am I missing a setting?
Ok, I processed the same data (20 frames) in other software and I have attached the result, so I know that the data is of good quality. Note the significant difference in the stars and the detail in the nebula - can I achieve something similar with APP?
Very strange. I usually just leave everything on default and then do Normalize. After that there is an analysis graph showing about ten metrics including quality, star shape, etc. I remove any image below the norm (of 1), say below 0.8 and then I do the integrate - also default. I always get a good result without artifacts. You can do this with only lights or you can add calibration frames - if there is anything amiss I do the process without calibration frames to get a better picture of whats in the raw images.
PS: use the latest beta22 - I'm not sure when the automatic analysis was introduced
@astrogee Thanks for your reply. I tried with all default settings and ended up with the first image I posted. I achieved the second image after reading some articles, watching some videos, and making changes to a number of settings. It's still not as good as the third image I posted, which I find odd, because all of the images are based on the same 20 subframes.
I did experience a big win with APP when I used it to process a mosaic. APP handled it really well and produced a nice result, so I know APP is good, but am stumped as to why it doesn't work with the 20 subframes I have been using (the whole dataset is over 200 subframes so I was hoping APP would filter out the poor quality ones for me).
I have the latest version installed (beta 22, I think).
Ok, I figured out what the problem was. I had to adjust the settings for rejection to Winsorized rejection and adjust the kappa low and kappa high values. I got a good result once I found the right values.
Here's the result of integrating over 300 subframes (it obviously needs some more work, but this is better than what I started with):
Ok, it looks like I'm back at square one again. I have a set of 10 FITS files that stack perfectly in DSS. When I try the same set of 10 FITS files in APP, the result is really bad (blurry, smeared stars, etc).
What can I do to improve the result? Should I upload the sample data?
Instead of waiting for someone to reply to me, I decided to ZIP the 10 FITS files along with the resulting FITS files from DSS and APP (in the result folder).
Here's the link to DropBox to download the file:
Hi Erik. Well I tried your data and I am having the same problem. Very strange. I see you are using a smart scope so the exposures are very short (10s). I did notice that about half of the images are not good - maybe due to wind? But then only 5 images at 10 secs doesn’t produce much. For sure there needs to be some rejection of shaky images. Not sure what setting for that. I’ll see what else I can find.
EDIT: Also, I think I would definitely use calibration frames on this if you can get them. The data looks good in the analysis but I'm surprised its finding 500 stars to analyze - probably mistaking them... I don't know.
Well, with <=20 frames you have to use very different rejection settings and good data. Because in this case the formula for outlier identification rely on very few observations and depending on the quality of the subs may produce very unstable results. Maybe even registration is having a hard time to do a good job...
So, if you got >20 lights, why did you try to stack so few of them? Even for testing or comparing you should use some more data. If you want to go for only 5 subs for whatever reason, they have to be flawless, so you have to drop every shaky one (how should a software be able to identify "bad" images with only 5 observations, its not a trained ki or something... :P)
Thanks for looking into this.
You are correct, I am using a Stellina for my observations.
I know the stars are not round and that, within the ten frames I posted, the stars move around a fair bit, but DSS doesn't seem to have a problem registering and stacking the images to produce a good result (refer to the file in result\stacked_using_dss.fts for the output DSS generated for me).
I found a thread that describes almost the same problem, and there was an update to fix the issue - here's a link to the relevant post in the thread:
@ewestermann Interesting - the final results on the link you gave are good but yes there does seem to still be star detection issues. There was a good suggestion to do dark frames with Stellina which could be used for calibration and bad pixel map. This may improve things a lot.
Erik, please accept my apologies for my late response to this issue. Of couse the result that you posted in your first post, looks very bad and not APP-like for sure !
I do recognize the issue here immediately, it is a know issue on my bug list which i will soon address. Your data most not be calibrated well since APP is detecting hot pixels for stars, causing the image alignment (registration) to fail.
The best solution is to supply darks and make a Bad Pixel Map, that should remove most of those hot pixels immediately .
In addition, in 3) Analyse stars, you need to enable the cosmic ray/noise reducer and probably you need to set it to a value of 10, to have it work nicely here.
The reason that i highly suspect that the noise in your data, the hot pixels especially, is the cause to have APP misproduce, is that you also indicate that you need to tweak outlier rejection.
Let me know if this explains it 😉 and solved it.
I will make sure that going forward, APP will deal better with data like this.
I had some difficulties, but I managed to capture dark frames. I downloaded the latest beta (26) and tried to process some of my files using the settings you suggested, and I had dark frames - APP produced an acceptable result. I'll try with some more observations to find out how they work out.
I did however notice that the image APP produces is quite a bit brighter than the DSS image - there is also significant vignetting on the APP image. Does this happen because APP is still in beta?
The mean pixel value of the DSS image is 1.96e-2, while the mean for the APP image is 3.9e-1 so it is a lot brighter.
I have created a ZIP containing two linear FITS files  - one from DSS and the other from APP (see filenames and ReadMe.txt for details) - so that you can compare them. Is there some setting I'm missing in APP?