Registax Stack Size comparison analysis and results

[bus_ter]

How many frames do you need to stack, and what is the optimum number?

Well here is a little experiment to see just that.

I started off with a 2000 frame video of Jupiter taken under very good seeing conditions. Using PIPP I put them in order of quality and reduced (LIMIT) the stack sizes into 2000, 1000, 500, 100, 50, 25, 12 and 5 stacks of frames. The quality ordering should mean the smaller stack sizes have the highest quality frames.

I first processed the full 2000 stack with RS6 and used the wavelets to get as much detail as possible without excessive noise. I then saved these settings. I then ran each other stack through Registax with EXACTLY the same settings. The results are on the left hand column. The results show that larger stack sizes allow more aggressive use of wavelets before noise becomes apparent. I didn't realise just how much difference this would make!

I then reprocessed each stack again, but decreasing the wavelets until the excess noise was reduced to a similar level to the 2000 frame stack. These results are on the right.

Even though the stack sizes are based on quality of frames, there seems to be no advantage in this sample in Limiting the stack size.

The bottom line *More frames allows more aggressive sharpening (wavelets) before noise is introduced*

Original Video

[cgarry]

Good experiment and nicely presented. However, I would think that 2000 frames are not really enough to expect big improvements from using quality selection. I like to start with at least 6000 frames so there is plenty of scope for dropping frames, though it helps to have a camera that does 60 fps!

I would be interested to seeing a similar test with quality weighting enabled in PIPP as that does give a slight improvement in certain conditions but I do not quite understand what those conditions are.

Cheers,

Chris

[Gina]

I think the atmosphere causes not only wobbling but also changes in focus, so selective dumping of frames can dump the out of focus ones leaving the sharp ones.

[bus_ter]

Another link for the images

For some reason I can't get it to show in the original full size