Oskari Nikkinen:
This thread is a real treasure trove of misinformation, OP if you are still reading this thread disregard everything about the hallucinations regarding sampling rates written above.
Says you. Nothing I said was misinformation. I've been doing this a lot longer and I have a lot more experience with different imaging systems. I am not right about everything but in this case I have experience and data to back me up. I have spent many hours imaging the same objects with different systems on the same nights. I can certainly say whats ideal and what isn't.
1.19" is not optimal for small galaxies.
1.19" is not optimal for small galaxies.
1.19" is not optimal for small galaxies.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Here is an image captured at .33" resolution and next to it is the same data binned 3x to .96" resolution. The seeing was 2.6" which is not good by any means. I am sampled 8x my seeing! and yet my resolved detail is much better than what It would be at only 3x which is considered optimal. When It comes to astrophotography being oversampled is better and even more so on small objects.
 |
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Matthew Proulx:
Here is an image captured at .33" resolution and next to it is the same data binned 3x to .96" resolution. The seeing was 2.6" which is not good by any means. I am sampled 8x my seeing! and yet my resolved detail is much better than what It would be at only 3x which is considered optimal. When It comes to astrophotography being oversampled is better and even more so on small objects.
But is it really better? Certainly not by a factor of 3. The pixelated look for the zoomed in image comes from the fact that PixInsight uses a very simple interpolation method (i believe nearest neighbor, not entierly sure) to show values other than 1:1.
I downloaded this image, cropped the right side off, binned x3 and then upscaled to 300% using Lanczos-4 interpolation and put them side by side again. Result below:
To my eyes they look quite similar now, if not nearly identical. I would say no detail was lost by binning x3, at least not in any significant capacity but obviously this is not a robust test since it was done on a .PNG instead of proper data.
I think you should let this one go, your arguments are making less and less sense.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Oskari Nikkinen:
Matthew Proulx:
Here is an image captured at .33" resolution and next to it is the same data binned 3x to .96" resolution. The seeing was 2.6" which is not good by any means. I am sampled 8x my seeing! and yet my resolved detail is much better than what It would be at only 3x which is considered optimal. When It comes to astrophotography being oversampled is better and even more so on small objects.
But is it really better? Certainly not by a factor of 3. The pixelated look for the zoomed in image comes from the fact that PixInsight uses a very simple interpolation method (i believe nearest neighbor, not entierly sure) to show values other than 1:1.
I downloaded this image, cropped the right side off, binned x3 and then upscaled to 300% using Lanczos-4 interpolation and put them side by side again. Result below:
To my eyes they look quite similar now, if not nearly identical. I would say no detail was lost by binning x3, at least not in any significant capacity but obviously this is not a robust test since it was done on a .PNG instead of proper data.
I think you should let this one go, your arguments are making less and less sense.
It does not look similar at all. I can see shadows and lines that don't exist on the right. Not only that, in Native resolution there would be far less to discern without zooming in 3x. Check yo spectacles. This is one I WILL NOT LET GO.
You're also missing something very important!!! I have pixels here to work with. Had that data been captured at a lower resolution, the noise would not only be higher, the post processing would have been vastly different. Imaging oversampled I can do much more in the way of processing and if need be, bring it back to a lower resolution.
If you really want to prove me wrong, go image this with your telescope and then come back to me. I will be waiting.
 |
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Matthew Proulx:
Had that data been captured at a lower resolution, the noise would not only be higher,
Again, the opposite of thruth. You would have higher signal to noise ratio with a lower resolution. This is all very simple astro 101, not sure how you keep saying things that are obviously not true.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Matthew Proulx:
Here is an image captured at .33" resolution and next to it is the same data binned 3x to .96" resolution. The seeing was 2.6" which is not good by any means. I am sampled 8x my seeing! and yet my resolved detail is much better than what It would be at only 3x which is considered optimal. When It comes to astrophotography being oversampled is better and even more so on small objects.
Hi,
What you see when you zoom 6:1 in an active Pixinsight view is not a good indicator, since the scaling algorithm is optimized for speed, not result quality.
Here what happens when I produce the right image by scaling the left image down to 33% and then back 300%, but using the Resample process with Mitchel-Netravali interpolation (which mimics more accurately how a different sample size would look):
The only detail that was lost is noise (mostly at <1px scale), due to the inherent smoothing effect of the interpolation. If you were imaging at 0.96" you wouldn't have that noise to begin with. In fact, you can scale down to 25% and back to 400% (to mimic sampling at 1.28") and the result will be practically identical.
You need to go at 6x (almost 2" sampling) to start having a noticeable loss of detail which means you actual seeing was closer to 4".
Proper sampling gives you the maximum detail you can attain. There is no way you can do better. Undersampling sacrifices detail, while oversampling sacrifices SNR.
If you are aiming for maximum detail and cannot be 100% perfectly sampled (which is almost always the case, given you have a constant focal length and a pixel size which can at best change to 1/2nd or 1/3rd or 1/4th), then slight (or even moderate) oversampling makes better sense, because noise reduction is much easier than restoring detail. But you cannot do better than what optimum sampling would give you. Best you can hope for is getting the same level of detail (slightly less actually), with a bit of extra noise left over from noise reduction.
There's a reason they are calling it a theorem. 
Cheers,
D.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Oskari Nikkinen:
Matthew Proulx:
Had that data been captured at a lower resolution, the noise would not only be higher,
Again, the opposite of thruth. You would have higher signal to noise ratio with a lower resolution. This is all very simple astro 101, not sure how you keep saying things that are obviously not true.
The signal at .96" or binned 3x to .96" in software will be the same. The SNR will not be. You figure that one out.
Go image this at 1.19" please, come back with the same result. I bet you can't. Lots of blah blah but not a lot of putting up.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Matthew Proulx:
Here is an image captured at .33" resolution and next to it is the same data binned 3x to .96" resolution. The seeing was 2.6" which is not good by any means. I am sampled 8x my seeing! and yet my resolved detail is much better than what It would be at only 3x which is considered optimal. When It comes to astrophotography being oversampled is better and even more so on small objects.
Hi,
What you see when you zoom 6:1 in an active Pixinsight view is not a good indicator, since the scaling algorithm is optimized for speed, not result quality.
Here what happens when I produce the right image by scaling the left image down to 33% and then back 300%, but using the Resample process with Mitchel-Netravali interpolation (which mimics more accurately how a different sample size would look):
The only detail that was lost is noise (mostly at <1px scale), due to the inherent smoothing effect of the interpolation. If you were imaging at 0.96" you wouldn't have that noise to begin with. In fact, you can scale down to 25% and back to 400% (to mimic sampling at 1.28") and the result will be practically identical.
You need to go at 6x (almost 2" sampling) to start having a noticeable loss of detail which means you actual seeing was closer to 4".
Proper sampling gives you the maximum detail you can attain. There is no way you can do better. Undersampling sacrifices detail, while oversampling sacrifices SNR.
If you are aiming for maximum detail and cannot be 100% perfectly sampled (which is almost always the case, given you have a constant focal length and a pixel size which can at best change to 1/2nd or 1/3rd or 1/4th), then slight (or even moderate) oversampling makes better sense, because noise reduction is much easier than restoring detail. But you cannot do better than what optimum sampling would give you. Best you can hope for is getting the same level of detail (slightly less actually), with a bit of extra noise left over from noise reduction.
There's a reason they are calling it a theorem.
Cheers,
D.
Nyquist's Theorem is based on analog radio signal. Now lets convert analog signal to digital and then display it on square pixels. Great thats what CCDs do. But theres also the introduction of noise when it becomes digital signal. That's why the optimum becomes 3-4 pixels. Our signal also isn't perfect. Its scattered across optics through the atmosphere. With the processing tools we now have, we can oversample by a large margin and get a better result.
I'm done here. Put out some good pictures because that's all I'm gonna keep doing.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Matthew Proulx:
Oskari Nikkinen:
Matthew Proulx:
Had that data been captured at a lower resolution, the noise would not only be higher,
Again, the opposite of thruth. You would have higher signal to noise ratio with a lower resolution. This is all very simple astro 101, not sure how you keep saying things that are obviously not true.
The signal at .96" or binned 3x to .96" in software will be the same. The SNR will not be. You figure that one out.
Go image this at 1.19" please, come back with the same result. I bet you can't. Lots of blah blah but not a lot of putting up.
Right you are, the SNR will not be the same. Because the captured-at-lower resolution image will have higher SNR due to the same amount of signal being present in both, but fewer pixels and so a smaller effect of read noise will be present in the lower resolution one. This is insignificant if one exposes long enough though, i am at least 60% sure you are aware of that.
Im not saying you should image an object like this at 1.19''/px by the way, not sure where you got that idea. What Im saying is that if lets say 1''/px is oversampled, then any sampling rate higher than that will be objectively worse, and there are some lower sampling rates that could be better.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
Oskari Nikkinen:
Matthew Proulx:
Oskari Nikkinen:
Matthew Proulx:
Had that data been captured at a lower resolution, the noise would not only be higher,
Again, the opposite of thruth. You would have higher signal to noise ratio with a lower resolution. This is all very simple astro 101, not sure how you keep saying things that are obviously not true.
The signal at .96" or binned 3x to .96" in software will be the same. The SNR will not be. You figure that one out.
Go image this at 1.19" please, come back with the same result. I bet you can't. Lots of blah blah but not a lot of putting up.
Right you are, the SNR will not be the same. Because the captured-at-lower resolution image will have higher SNR due to the same amount of signal being present in both, but fewer pixels and so a smaller effect of read noise will be present in the lower resolution one. This is insignificant if one exposes long enough though, i am at least 60% sure you are aware of that.
Im not saying you should image an object like this at 1.19''/px by the way, not sure where you got that idea. What Im saying is that if lets say 1''/px is oversampled, then any sampling rate higher than that will be objectively worse, and there are some lower sampling rates that could be better.
I never debated that. I said I'd rather be oversampled everytime. As I said for most people in average seeing, the best sample rate is .75". I want to be in a range where in my best seeing night I'm not giving up data. If you want to image small objects you need to give yourself the best possible chance. When seeing drops to 1.5" and it will no matter where you live at least once a year, its ideal to be at .5" or better. Now if you're willing to give that up thats on you. I have a hard time believing anyone serious about this imaging in rural sky or even b4-b5 is getting 3" seeing year round. I would not want to limit myself to that.
|
You cannot like this item. Reason: "ANONYMOUS".
You cannot remove your like from this item.
Editing a post is only allowed within 24 hours after creating it.
You cannot Like this post because the topic is closed.
6.06
3.15
6.89
