12 bit ser vs 8 bit avi Jupiter

[MilwaukeeLion]

6/14 took 3 ser rgb captures and 3 avi rgb close together.  One minute captures each same roi 640x480 camera asi120mmS. 12 bit ser = 1365 frames each rgb 19% stacked in as3- 8 bit avi = 1750 frames each rgb 15% stacked in as3.  De-rotated videos ran thru as3 again same parameters then final rgb images combined in winjupos, color balanced, opened in ps, resized 200%, same unsharp mask applied and despeckle.  At this point images are pretty similar avi is showing little more detail.  Usually don't go much further just slight wavelets if halfway decent image came out.

Did wavelets, denoise and lowered levels tad.  Can certainly push wavelets much further with 12 bit.  8 bit begins to degrade fast.  12 bit easier to wavelet and keep smooth planet look but not much difference overall.  Could've pushed 12 bit more, quality of overall video prob not worth it.  Maybe better seeing would show more of difference. If not better to take avi for speed alone.  End of experiment.

Clear skies!

[MarsG76]

Correct me if I'm wrong but I always thought that even if you capture 8 bit data, that stacking 500 (or a large number) of the best frames 8 bit frames in a 16 bit container would result with actually 16 bits of data per channel on the stack.... so that would make sense that there wouldn't be much, if any, difference to the final result or stacked data itself.

[JamesF]

To me it appears that the AVI version has a more obvious "ringing" artefact in the enlarged images.  It seems particularly obvious at the limb on the lower left hand side.

James

[MilwaukeeLion]

2 hours ago, MarsG76 said:

Correct me if I'm wrong but I always thought that even if you capture 8 bit data, that stacking 500 (or a large number) of the best frames 8 bit frames in a 16 bit container would result with actually 16 bits of data per channel on the stack.... so that would make sense that there wouldn't be much, if any, difference to the final result or stacked data itself.

Someone wiser than me will need to jump in on that. The avi does degrade quicker pushing the wavelets but possible it came out of as3 crispier. Also possible seeing changed or slight wind gust inside of 10 minutes all were captured.

[MilwaukeeLion]

1 hour ago, JamesF said:

To me it appears that the AVI version has a more obvious "ringing" artefact in the enlarged images.  It seems particularly obvious at the limb on the lower left hand side.

James

Good eye yes avi does show bit of a ring \ more sensitive to sharpening.  

[MarsG76]

57 minutes ago, MilwaukeeLion said:

Also possible seeing changed or slight wind gust inside of 10 minutes all were captured.

I was thinking this also... 

[symmetal]

Craig Stark has a useful article on this The Effect of Stacking on Bit Depth

Towards the end of the article he concludes that stacking even a few 8 bit images reduces the quantization noise due to bit depth considerably as long as there is some noise in the 8 bit images. (As all astro images have.)

Stacking a number of  8 bit images, he found this is the equivalent bit depth of the stacked image.

Stacking 10 gives a bit depth of 10.5

Stacking 50 gives a bit depth of 11.6

Stacking 100 gives a bit depth of 12.1

Stacking 200 gives a bit depth of 12.6

So at a very modest 25 fps, an 8 second video of 8 bit data will yield a stack equivalent to a 12.6 bit image. Stacking more will reduce the overall noise further but it can be seen that the quantization error caused by 8 bits quickly becomes insignificant as soon as a relatively small number of images are stacked (as long as the images contain some noise initially).

I've tried taking full frame 3096x2080 pixel lunar and solar images at 8 bit (30fps) and 12 bits (15fps ) for 30 seconds each, giving 900 and 450 frames respectively and no matter how much processing I did I couldn't find any difference between them.

For planetary I manage 200fps for a 400x400 pixel image so the 6000 stacked frames of a 30 sec 8 bit video makes 12 bit imaging totally unnecessary, and just uses up double the storage space (as they're saved as 16 bit videos).

Alan

Edited June 16, 2019 by symmetal

[MarsG76]

54 minutes ago, symmetal said:

Craig Stark has a useful article on this The Effect of Stacking on Bit Depth

Towards the end of the article he concludes that stacking even a few 8 bit images reduces the quantization noise due to bit depth considerably as long as there is some noise in the 8 bit images. (As all astro images have.)

Stacking a number of  8 bit images, he found this is the equivalent bit depth of the stacked image.

Stacking 10 gives a bit depth of 10.5

Stacking 50 gives a bit depth of 11.6

Stacking 100 gives a bit depth of 12.1

Stacking 200 gives a bit depth of 12.6

So at a very modest 25 fps, an 8 second video of 8 bit data will yield a stack equivalent to a 12.6 bit image. Stacking more will reduce the overall noise further but it can be seen that the quantization error caused by 8 bits quickly becomes insignificant as soon as a relatively small number of images are stacked (as long as the images contain some noise initially).

I've tried taking full frame 3096x2080 pixel lunar and solar images at 8 bit (30fps) and 12 bits (15fps ) for 30 seconds each, giving 900 and 450 frames respectively and no matter how much processing I did I couldn't find any difference between them.

For planetary I manage 200fps for a 400x400 pixel image so the 6000 stacked frames of a 30 sec 8 bit video makes 12 bit imaging totally unnecessary, and just uses up double the storage space (as they're saved as 16 bit videos).

Alan

Great article....

[MilwaukeeLion]

16 hours ago, symmetal said:

So at a very modest 25 fps, an 8 second video of 8 bit data will yield a stack equivalent to a 12.6 bit image. Stacking more will reduce the overall noise further but it can be seen that the quantization error caused by 8 bits quickly becomes insignificant as soon as a relatively small number of images are stacked (as long as the images contain some noise initially).

Good stuff, I tested again with stacks of 600.  Error was applying similar sharpening to different stacks.  Only difference is using 2 to 3 times wavelets on 12 bit stack to get similar detail of 8 bit stack.  That's probably differences in seeing/wind of captures?  I did another test using pipp to sort image quality before as3 vs just using as3 to sort image quality.  There was no difference in final images at least with this Jupiter avi.

 

Edited June 16, 2019 by MilwaukeeLion
pipp test

[Pete Presland]

Interesting comparisons.