cybereye wrote:Hello Ivo!
I was just wondering if you could provide some references/papers regarding how the denoise is done when I turn off the tracking. The main thing I'm coming to grips with is getting the optimum grain-size to do the denoise. Is bigger better (depending on data of course) or is there a point of diminishing returns...?
Apologies - a write up on all the new features in 1.3.5 (similar to the write up explaining the color module) is sorely needed.
The grain size thing is new, but it's pretty easy to understand (once someone has explained it
StarTools' philosophy is one of "don't bother the user with asking for a setting, unless it's a matter of aesthetics or subjective interpretation". Grain size is one of those settings.
What this parameter allows you to specify, is the size of noise grain that is visible in the image - it's purely a visual thing (with definite ramifications for the denoise stage though). In the denoise setup screen you blur the image until you cannot see any noise grain anymore, the more you blur, the bigger the noise grain that is 'swallowed' by the blur. You'll find the same grain size parameter on the actual denoising screen - it governs the exact same thing. The reason it's also on the setup screen is that it allows for a more easy to see aid when determining a good value for it.
Once StarTools knows what size of noise grain you ('the observer') can no longer see, it uses this information during the denoising process; because it knows that noise grain larger than the one you specified can not be seen, it can better retain structure that is larger
than that noise grain (since anything larger than the noise grain specified cannot be noise). It further gives the denoising algorithm a surface area over which it can safely redistribute energy that was taken away (denoised), so that the image will retain the same amount of energy. The latter stems from the assumption that 'noise' is just spatially misallocated energy - still a valid measure when spread over a larger area. This assumption helps retaining and emphasizing larger scale structure using the 'noisy' energy.
Hope this helps in the meantime. Do let me know if you have more questions about the denoise module - it's arguably the most advanced module in StarTools and, as far as I know, blows any competing algorithms out of the water for the simple reason that it has vastly more information (Tracking) about your image than other algorithms (which just look at the image 'as-is').