Excellent drizzling explanation! However, it misses some very important caveats. 1. Drizzling doesn't come free - there IS noise injected by drizzling, so you do reduce your SNR - there is no free lunch :/ 2. Drizzling should ONLY be considered when you are definitely undersampled, have dithered a lot, and have a lot of data. If you are at good sampling, or oversampling, your are degrading your image when drizzling by injecting noise without getting anything in return 3. In my experience drizzling on OSC images can cause artifacts 4. On a mount like the star adventurer, you can only dither in RA - this impacts drizzling to some extent One of the biggest advantages of drizzling is when you are so undersampled that your stars are blocky. This can be fixed by drizzling (sacrifice SNR), then resampling down (gain the SNR back). After a period of drizzling every stack I touched, I realized it didn't bring that much of an advantage AND it increased my processing time by a lot, so I stopped using it - except in severe undersampling situations.
Good point about reducing SNR! I am these using it with data dithered in both axes with a proper GOTO mount and a 300mm scope which if I’m doing the math correctly makes me undersampled with my mirrorless and dslr OSCs. Haven’t noticed artifacts. I’m gonna pin your comment - important stuff here indeed, thanks!
If your stacking using the WBPP script just click on generate drizzle data in the light tab. Then after you open your master light, go to all processes and click to integrate drizzle data, then direct to files to load from registered folder created by the WBPP script. This is only a real advantage if your original data in under-sampled. The idea behind drizzling is to reduce a higher pixel scale to the correct sampling rate.
Just discovered your channel. It was a suggestion on a RUclips page. AWSOME! I watch many astrophotography videos, especially for the Skywatcher. Where have you been hiding? Everything I've watched in the past hour is new to me, and very sensible for me. Maybe I just haven't found it yet, but settling up PHD2 with only one axis is not intuitive. Fortunately Peter Zelinca has such a tutorial. You've made me a fan!
It only provides benefit if you have dithered during acquisition AND you are undersampled. Look up your telescope and camera to see if you are undersampled, otherwise it is just eating disk space and cpu hours for no real benefit.
How well is this working out for you? Are you able to get the reach you were looking for? I am about to start trying to dither and drizzle for myself to compare my results. Keep up the great work, Kamil.
Absolutely correct technical information. It’s 100% effective. More and more software bases are recognizing the importance of drizzle integration. 100% of the people who drizzle agree on its importance. Fine video explanation.
want to buy an old version of Nikon ED AF Nikkor 300mm 1: 4, i just bought a skywatcher tracker. Do you recommend this 300mm old lens is good in astrophotography? I have an old D7000 nikon
Hey Kamil! Wonderful overview video of Drizzle integrations mate, nice one! :-) It's certainly something that can be used to your advantage in many cases! Clear skies :-)
Thanks Luke, exactly. But this is not something you could naturally think of that even exists in software - I had no idea about this until just recently and it blew my mind 😃
Sorry, but when you say “turning this 25mm lens into a 50mm lens…” what exactly does that mean and how do you achieve it? Thanks. I’m a noob that just bought a Lumix G7. Haha 😅
Well, with drizzle you "sacrifice" tons of time and you need to have NASA PC 😉 My last drizzle integration in total took me like 5h (weighted batch pre-processing, subframe selector, calibration and drizzle integration) and I just had 200 frames
It does not (literally) double your focal length, but improves quality of your photos like you would have 2-4x focal length. But this effect is limited only to situations when your photos are undersampled. In general it works pretty well. Such effect is also used on analysis of poor quality cctv video when you have low res cctv. Basically your video is a series of photos and by math and analysis you can make face or plates super sharp. Same thing is applied here. You have series of undersampled photos with some "drift" between then, and by math, software makes final image way more sharp, like you would have x2 or x4 better resolution.
6:10 No, it doesn't. You get nearly exactly the same result with bicubic smoother upscaling. The actual advantage of the angular resolution might be lower than 1.1x. Keep in mind that improving the angular resolution by 1.1x is already like comparing 14.5 megapixels and 12 megapixels. A true increase in the focal length would also make more stars visible (per angle) and would make other stars look smaller. As one can see, the stars are not smaller in the shown example. There are also smartphones (for example Pixel 3a) that use this technique and I compared the results with bicubically upscaled dng files in perfect light conditions multiple times and there is nearly no difference regarding the angular resolution. The digital zoom of a Pixel 3a can produce smoother (but artificial) contours (due to machine learning) and there can be a better signal to noise ratio (due to a longer total exposure time), which can occasionally make more details visible, but this doesn't help at base Iso equivalent exposures (exposed to the right). I also tried the 3x Drizzle mode of Deep Sky Stacker (with hundreds of undersampled frames) and it again confirmed my experience that it is nearly useless in comparison to bicubic upscaling. Drizzle is basically useless in comparison to the tripod high resolution modes of certain cameras. Certain Olympus cameras have a handheld high resolution mode and there can be a tiny advantage in the jpgs, but that's just due to the better signal to noise ratio (longer total exposure time) and due to less spatial noise reduction, it is not a true, noticeable increase in the angular resolution.
Excellent drizzling explanation! However, it misses some very important caveats.
1. Drizzling doesn't come free - there IS noise injected by drizzling, so you do reduce your SNR - there is no free lunch :/
2. Drizzling should ONLY be considered when you are definitely undersampled, have dithered a lot, and have a lot of data. If you are at good sampling, or oversampling, your are degrading your image when drizzling by injecting noise without getting anything in return
3. In my experience drizzling on OSC images can cause artifacts
4. On a mount like the star adventurer, you can only dither in RA - this impacts drizzling to some extent
One of the biggest advantages of drizzling is when you are so undersampled that your stars are blocky. This can be fixed by drizzling (sacrifice SNR), then resampling down (gain the SNR back).
After a period of drizzling every stack I touched, I realized it didn't bring that much of an advantage AND it increased my processing time by a lot, so I stopped using it - except in severe undersampling situations.
Good point about reducing SNR! I am these using it with data dithered in both axes with a proper GOTO mount and a 300mm scope which if I’m doing the math correctly makes me undersampled with my mirrorless and dslr OSCs. Haven’t noticed artifacts. I’m gonna pin your comment - important stuff here indeed, thanks!
I think DSS also has a x2 and x3 drizzle effect
Yes it does, I think its better but only because DSS is free :P
Where is it inside DSS, I cant find it
@@bralampt There's a 2x or 3x drizzle checkmark when you are registering the subs. You can't really miss it.
After many years of only half understanding the terms, at last I truly understand! Thank you for the clear explanation.
As always: videos with super helpful tips ... you're great Kamill. We await the video with the explanation in PI of this technique ... please. Cheers!
Actually also worth mentioning: with drizzle also additional noise is re-introduced. So better to have plenty of subs before considering drizzle.
Good point Jens, thanks!
Please do a second video with step by step instructions on this please! I’ve used PI for a while and didn’t know about this
Will do!
If your stacking using the WBPP script just click on generate drizzle data in the light tab. Then after you open your master light, go to all processes and click to integrate drizzle data, then direct to files to load from registered folder created by the WBPP script. This is only a real advantage if your original data in under-sampled. The idea behind drizzling is to reduce a higher pixel scale to the correct sampling rate.
I've heard those terms, but had no idea what they meant. Wow! Thank you, Kamil! What useful information!
Thanks, Kamil. I use APP and it does have a drizzle function. I was just unsure if I should use it. Now I will give it a try.
Astro Pixel Processor also has drizzle :)
Just discovered your channel. It was a suggestion on a RUclips page. AWSOME! I watch many astrophotography videos, especially for the Skywatcher. Where have you been hiding? Everything I've watched in the past hour is new to me, and very sensible for me. Maybe I just haven't found it yet, but settling up PHD2 with only one axis is not intuitive. Fortunately Peter Zelinca has such a tutorial. You've made me a fan!
Many thanks, I appreciate the comment! And yeah, I was learning how to guide with SA from Peter as well 😀
It only provides benefit if you have dithered during acquisition AND you are undersampled. Look up your telescope and camera to see if you are undersampled, otherwise it is just eating disk space and cpu hours for no real benefit.
Hey kamil thanks for your tutorial 😇😇
Great stuff, I think you’ve earned a subscriber 8)
How well is this working out for you? Are you able to get the reach you were looking for? I am about to start trying to dither and drizzle for myself to compare my results. Keep up the great work, Kamil.
Absolutely correct technical information. It’s 100% effective. More and more software bases are recognizing the importance of drizzle integration. 100% of the people who drizzle agree on its importance. Fine video explanation.
thanks - must look into that
You always big deep, thanks, your my go to guy for tech talk, 👍
Leave it to a Polish person to present this kind of info in a complex yet understandable way. Dziękujemy ci!
Hi, do you have any videos about comparing images of deep celestial objects taken by a telephoto lens (say, a Canon) with an APO?
Gee thanks Kamil for this info….please do a Pixinsight tutorial in how to integrate with drizzle.❤️😊👍🏼
You integrate as you would with non-drizzled data. That then updates the driizle files. Then use the drizzle integration process.
Hmm, NOW I might consider Pixinsight. Thanks for the video!!
It’s well, well worth it!
Yeah, 100% agreed Digit975!
Though it costs EUR 230, so not exactly “for free” unless you already own it, but cheaper and more effective than 2x converters 😊
Do you have any information about the large lens over your shoulder? Videos or and pictures you have taken with it?
The fat one on the far left of the frame?
@@kamilkp that's the one
Not yet but I’m planning to make a video about it at some point. It’s a vintage catadioptric lens
@@kamilkp looks similar to a 800mm mirror lens I was looking into recently
It’s probably the same lens. Catadioptric = mirror
want to buy an old version of Nikon ED AF Nikkor 300mm 1: 4, i just bought a skywatcher tracker. Do you recommend this 300mm old lens is good in astrophotography? I have an old D7000 nikon
Hey Kamil! Wonderful overview video of Drizzle integrations mate, nice one! :-) It's certainly something that can be used to your advantage in many cases!
Clear skies :-)
Thanks Luke, exactly. But this is not something you could naturally think of that even exists in software - I had no idea about this until just recently and it blew my mind 😃
I used to program computer games and I could upscale an object in real time by using an anti-aliasing routine.
Many thanks. Do you know if there is something in Siril?
Not that I’m aware of
Siril has drizzle. Just use the processing script that has drizzle incorporated.
Drizzling works only on color cameras. Right?
i was expecting something related to visual astronomy ...a trick which can improve magnification without using Barlow and without sacrificing visual
Sorry, but when you say “turning this 25mm lens into a 50mm lens…” what exactly does that mean and how do you achieve it? Thanks. I’m a noob that just bought a Lumix G7. Haha 😅
Hello, I have a Mac. Which app could I use to do that?
Thanks
PixInsight
I thought it was for Pc only.
Thanks
just wow!
want to double your FL, easy, use a micro 4/3 camera instead of FF.
Correction on Peter Zelinka, with a 'k' not a 'c'. Sorry Peter! Who by the way also has a wealth of astro knowledge.
Well, with drizzle you "sacrifice" tons of time and you need to have NASA PC 😉 My last drizzle integration in total took me like 5h (weighted batch pre-processing, subframe selector, calibration and drizzle integration) and I just had 200 frames
Yeah but you can leave it running overnight and let the computer crunch all the numbers while you sleep
I have seen the video but still don't understand how you doubled the focal length.
It does not (literally) double your focal length, but improves quality of your photos like you would have 2-4x focal length. But this effect is limited only to situations when your photos are undersampled. In general it works pretty well. Such effect is also used on analysis of poor quality cctv video when you have low res cctv. Basically your video is a series of photos and by math and analysis you can make face or plates super sharp. Same thing is applied here. You have series of undersampled photos with some "drift" between then, and by math, software makes final image way more sharp, like you would have x2 or x4 better resolution.
@@mehow357 That makes sense, I get it now. Thanks!
Where is the line drawn between this and spray painting glasses and a beard on?
... and increase exposure at lower ISO to enhance the signal to noise ratio.
6:10 No, it doesn't. You get nearly exactly the same result with bicubic smoother upscaling. The actual advantage of the angular resolution might be lower than 1.1x. Keep in mind that improving the angular resolution by 1.1x is already like comparing 14.5 megapixels and 12 megapixels. A true increase in the focal length would also make more stars visible (per angle) and would make other stars look smaller. As one can see, the stars are not smaller in the shown example.
There are also smartphones (for example Pixel 3a) that use this technique and I compared the results with bicubically upscaled dng files in perfect light conditions multiple times and there is nearly no difference regarding the angular resolution. The digital zoom of a Pixel 3a can produce smoother (but artificial) contours (due to machine learning) and there can be a better signal to noise ratio (due to a longer total exposure time), which can occasionally make more details visible, but this doesn't help at base Iso equivalent exposures (exposed to the right). I also tried the 3x Drizzle mode of Deep Sky Stacker (with hundreds of undersampled frames) and it again confirmed my experience that it is nearly useless in comparison to bicubic upscaling. Drizzle is basically useless in comparison to the tripod high resolution modes of certain cameras. Certain Olympus cameras have a handheld high resolution mode and there can be a tiny advantage in the jpgs, but that's just due to the better signal to noise ratio (longer total exposure time) and due to less spatial noise reduction, it is not a true, noticeable increase in the angular resolution.
Before I watched it, I thought another clickbait video that will present a gimmick.
Couldn't be more wrong