Lunar data, what % of frames to stack

[turfpit]

In this post viewtopic.php?t=8810 - the % of frames to stack came to light.

When processing with Autostakkert!4, stacks of 25, 50, 75 and 85% were produced.

stack_results.JPG (154.59 KiB) Viewed 124 times

The stack logs contained an entry for stack quality (this used to be 0 in earlier versions).

stack_quality.JPG (39.25 KiB) Viewed 124 times

This is a summary of the stack quality figures.

quality.JPG (9.1 KiB) Viewed 124 times

All 4 stacks were processed identically in BiggSky with blind deconvolution (PSF from ROI) & gamma=1.5.

The 25% and 85% stacks were evaluated (visually) as the extremes were more likely to make any differences stand out. Each stack was loaded into GIMP and displayed at 50%.

stack_comparison.JPG (138.96 KiB) Viewed 124 times

Each stack was loaded into GIMP and displayed at 100%.

stack_comparison_100percent_zoom_in_GIMP.JPG (120.44 KiB) Viewed 124 times

Each stack was loaded into GIMP, displayed at 50% and contrast of 0.050 applied.

stack_comparison_contrast_applied_in_GIMP.JPG (171.7 KiB) Viewed 124 times

In the comparisons above both images seem to present the same level of detail, particularly in the crater bottoms and number of craterlets displayed around the images. It does seem to come down to the personal preference for brightness of the image.

Based on my own experiences after processing 0.5Tb of lunar data captured in May, I am now stacking any frames which are above 50% quality on the graph but YMMV.

Autostakkert_Quality_graph.JPG (11.32 KiB) Viewed 124 times

The stack anything over 50% quality worked alright for me here viewtopic.php?t=8807.


Well you have now got a systematic way to evaluate outcomes for your lunar images.

Dave

[admin]

Hi Dave,

a very near thing in the 25% vs 85% comparison - I think you would have to do some sort of computational sharpness analysis (like the contrast focus score in SharpCap) to really tell if one has more detail than the other. Stacking a bigger fraction of the frames ties in with experience from the planetary live stacking in SharpCap, which can actually give good results in many cases stacking *all* frames. For fainter targets like the planets where noise is a bigger issue, only taking 25% of the frames means 2x the noise level in the stack, which can limit the scope for sharpening before the noise becomes intrusive. Filtering is, I think, most useful when the seeing is variable with bad spells mixed in with steadier conditions.

cheers,

Robin

[turfpit]

Robin

The JPGs posted in the forum are degraded. Each stacked TIFF is 3.3Mb but even with those I cannot discern any real difference by eye between 25% and 85% stacks. All the images in viewtopic.php?t=8807 were stacked using all frames above 50% quality , which seemed to turn out alright.

Planets are a different case but I haven't the same experience in that area.

Dave

[turfpit]

Something occurred to me overnight (isn't sleep wonderful for focussing the mind) - the quality graph has ~85% of the frames above the 50% quality line. This is probably a set of data captured during very good seeing. I will try the same exercise when I come across some more variable quality data.

Autostakkert_Quality_graph.JPG (11.32 KiB) Viewed 93 times

Dave

[admin]

Hi Dave,

yes, that's a useful thing to note, because it indicates that the seeing variation wasn't just a random 'normal' distribution - in fact you can see that in the graph there was one really bad patch at about 37.5% of the way through the file, another near the beginning and an extended poor patch near the end. I expect that the 85% selection is removing these quite nicely and leaving the rest. If you had instead mainly poor seeing with a few good patches, maybe 15% would be a good choice...

cheers,

Robin

[turfpit]

Robin

I am sure some bad seeing will be along soon to test this out. I might come across a suitable distribution graph as I reprocess old data. I will add to this thread when I find some.

After having ploughed through some 0.5Tb of May data, I am left with the impression that stacking frames above 50% quality seems to consistently produce decent results.

Dave