2023-01-19: APP 2.0.0-beta13 has been released !
!!! Big performance increase due to optimizations in integration !!!
and upgraded development platform to GraalVM 22.3 based on openJDK19
We are very close now to releasing APP 2.0.0 stable with a complete printable manual...
I'm just nearing the end of my 30-day trial, and I'm scrambling to try out as much as I can. I'm a little confused about the drizzle settings, specifically bayer drizzle. Scale is pretty straightforward, but droplet size has me baffled. I've seen so many recommended settings (scale 2 & droplets 2.25px, scale 1 & droplets 1px, scale 1 & droplets 2-3px) but I'm not sure I understand what droplet size to use.
Would droplet size and scale be related to the resolution of my CCD? Using astronomy.tools CCD suitability, I can see that I'm slightly undersampled (2.46"/pixel). Would I be able to use this information to arrive at a droplet size to approximate a sampling rate of 2"/pixel?
Given that the ideal CCD pixel size for my telescope would be 3.49µm and mine is 4.30µm, I tried a scale of 1.2 and a pixel size of 1.62. This looked fine, but not particularly sharp. I tried again with a scale of 1.2 and a pixel size of 0.81. This was noisy and dark (with the same stretch applied) and I could see the mesh pattern. I'm currently trying again with a scale of 1.2 and a droplet size of 1.21 to split the difference, but I'm really unsure about the best settings.
I know there are a few forum posts on Bayer Drizzle, but there doesn't seem to be an agreement on a best approach. One fellow reports excellent results with scale 1 and droplets 1px, so I might try that next, but I would really appreciate some direction here, and maybe some recommended settings. Thanks in advance!
Hi Mark @mgermani,
Drizzle is explained in detail in this post, have you seen this post? If not, please do, it might explain a lot. I think It might explain the meaning of the droplet size somewhat?
Bayer Drizzle is like an extension of Drizzle in the fact that only the CFA pixels are drizzled into the result. So no debayering of the data is done which is normally required to fill the CFA holes in R,G,B channels. So with Bayer Drizzle, the R,G,B channels are incomplete. The CFA holes are there. This has the consequence, that if you relate Drizzle to Bayer Drizzle, you would double the droplet size with Bayer Drizzle compared to Drizzle to get the same results in terms of image reconstruction/sharpness and noise levels in the result.
The post to which I refer should make it clear that there really is no such thing as ideal settings for Drizzle nor Bayer Drizzle, because it depends on factors that will be different for each image project that you do ;-).
The settings depend clearly on the sky quality when you shot your images, it depends on your optics, it depends on how much data that you have. It depends on the filters that you use, it depends on your personal taste for sharpness and noise, it depends on the imaging target I think even.... So please don't fixate yourself on possible ideal drizzle settings for your imaging setup, because I think it does not exist 😀.
Of course, if you shoot multiple targets and then play with the drizzle scale and droplet size, you will get a good feel for what works for you and this can be rather personal because it relates to your taste for sharpness and noise in the result. If you want more sharpness, then you need to realize that you will also get a more noisy result. So there is a fine personal balance where you want to balance sharpness and noise. For some object, sharpness is not so important, for others it can be really important 😉
Keep these 3 practical rules in mind:
- the lower the droplet size, the noisier the result for sure ! At some point you will start to see drizzle artefacts even...then you have set the droplet size too low, or you simply don't have enough data to make it work with such a droplet size.
- If drizzle works for your data because you meet the drizzle requirements well (undersampled and dithered data with plenty of it !), the lower the droplet size, the more sharp the result will be as well.
- At some point for a given droplet size, lowering the droplet size further will no longer increase sharpness, it will only create a noisier result...
In addition, I would strongly advise you to not relate this to a very theoretical approach of your ideal CCD pixel size, because in real life and taking all factors of influence of your imaging setup and imaging conditions, the nicest settings for scale and droplet size will be different each time for a new project 😉 (It will also for the choice of drizzle kernel as well !)
So, my advice is to take this with a practical approach instead of a theoretical approach. For instance, If you Bayer Drizzle, I would simply keep scale at 1, and make 3 stacks with different droplet sizes, 1.5 pixel, 2.0 pixel and 2.5 pixel. Then study how this affects the results for noise and sharpness.
If a low droplet size of 1.5 looks nicest, then keep the droplet size at 1.5 and increase scale to 1.5 maybe.... And again compare 😉 That should help you to find the nicest settings for the data set that you are processing.
A nice way to find the good settings is to not integrate the full composition of your data. Use the composition crop mode in 6)Integrate to only integrate/stack a small part of the field of view 😉
Ok, brilliant! I hadn't really thought about how drizzling compares to debayering, but I think I understand just how complex this actually is now - not reducible to an equation. I guess it also depends on the pixel-colour matrix of the camera, too.
Thank you so much for your thorough reply. I hadn't realized you could specify a smaller section of your FOV to stack. This will save me so much time, as I am primarily using drizzle right now to try and image really distant galaxies that my 360mm has a hard time with. The Leo triplet takes up less than a quarter of my FOV, so I'll definitely be using this feature!