wavelet (frequency) decomposition

Requests for new features and wish-list items.
Post Reply
Scottk
Posts: 17
Joined: Thu Aug 26, 2021 7:18 pm

wavelet (frequency) decomposition

Post by Scottk »

I understand that the sharpening works on various levels of detail. I assume that you are using the wavelet decomposition (frequency) technique. Have you considered extending this so that one could go in the opposite direction and "smooth out" areas of greater detail (higher frequency) so they don't look quite so "crunchy"?
Scott K
User avatar
admin
Site Admin
Posts: 3367
Joined: Thu Dec 02, 2010 10:51 pm
Location: Melbourne
Contact:

Re: wavelet (frequency) decomposition

Post by admin »

Scottk wrote: Wed Sep 22, 2021 4:54 pm I understand that the sharpening works on various levels of detail. I assume that you are using the wavelet decomposition (frequency) technique. Have you considered extending this so that one could go in the opposite direction and "smooth out" areas of greater detail (higher frequency) so they don't look quite so "crunchy"?
Indeed the Sharp module uses wavelet ("multi-scale") decomposition to process detail at different scales.
I think what you are alluding to, is essentially a basic wavelet-based noise reduction algorithm?
Normal and expected Poissonian (shot) noise should only(!) be taken care of during Denoise (when switching Tracking off) and at no other point. It even makes use of this noise intelligently to serve as quantization error diffusion.
It should be noted that, in StarTools, noise is a purely aesthetic consideration. You do not need to "help" StarTools or its algorithms. All important modules and algorithms are noise aware and should not exacerbate noise or react adversely to the presence of normal shot noise.
Similarly, you should not (need to) be undoing the excesses of modules with other modules in StarTools.

If this doesn't help, would you be able to post an example of what you think looks "crunchy"? If the noise is not Poissonian in nature then, indeed, all bets are off...
Ivo Jager
StarTools creator and astronomy enthusiast
Scottk
Posts: 17
Joined: Thu Aug 26, 2021 7:18 pm

Re: wavelet (frequency) decomposition

Post by Scottk »

Ivo,

Thank you for the quick response. I understand what you are saying about the noise reduction. I am coming at this from the perspective of one who has used the wavelet controls or whatever its called in various photo processing applications. I recognize that StarTools works differently. Also I would agree with your comment about noise being a purely aesthetic consideration.

Being a new user, I just need to use it more often and observe how the different modules and settings work on my dataset.

I think this is a great product. Thank you!
Scott K
User avatar
admin
Site Admin
Posts: 3367
Joined: Thu Dec 02, 2010 10:51 pm
Location: Melbourne
Contact:

Re: wavelet (frequency) decomposition

Post by admin »

Scottk wrote: Thu Sep 23, 2021 2:31 pm Thank you for the quick response. I understand what you are saying about the noise reduction. I am coming at this from the perspective of one who has used the wavelet controls or whatever its called in various photo processing applications.
I totally understand where you are coming from, and it is unfortunately necessary evil to "unlearn" some things in addition to learning others. Indeed, if you are used to other software, it is both a curse and a blessing when trying to learn StarTools.

The blessing is that it provides you with a good, fundamental understanding of the algorithms and concepts at work (PixInsight is a particularly good playground to learn the fundamental workings of algorithms and concepts in their most basic/naked/isolated form - if you used this particular package before, this table may be helpful).

The curse, is that nothing in StarTools works the way you are used to. Due to the Tracking engine, everything is time-shifted in the workflow to (for better signal fidelity or better controllability), or relies on statistics, settings and constraints that have been already determined by your previous actions. As a result, the workflows are much shorter, but the tools much more powerful, targeting very specific problem domains. Tools don't replicate functionality from other tools or get in each other's way. This also means that tools then don't need controls to "undo" the effect of other tools - circular processing with tweak-upon-tweak (and "overcooking" an image) is not a thing in StarTools.

Efficiency, predictability and fidelity are the name of the game! :thumbsup:
Ivo Jager
StarTools creator and astronomy enthusiast
Post Reply