Author Topic: MLT vs MMT  (Read 196 times)

Online ngc1535

  • PixInsight Addict
  • ***
  • Posts: 291
    • View Profile
MLT vs MMT
« on: 2019 August 04 19:33:10 »
Hi,

This topic is something of a recurring theme with bits of information scattered across several forum threads and PI tutorials.
I wanted to offer something of an approach to comparing the differences between all of the algorithms and methods.
In my mind what is necessary is to create an image that contains all (reasonably) possible scales of structures that are typically found in astronomical images.
There exist cyclic functions with decreasing amplitude that can serve this purpose. My favorite is sin(x^2) .
The Pixel math expression I used to generate a field of structures is:

sin(0.5 * ( X()*180/pi())^2 ) + sin(0.5 * ( Y()*180/pi())^2 )

This creates the interference of the two sines- and thereby makes an image of ever-decreasing structures. I came up with the idea because sin(x^2) was used as the exemplar for the ways in which FFTs do not adequately model quickly changing frequencies at all timescales. (There is a limit to the precision one can know about the frequency or time for a signal.)

With this image in hand, you can apply MLT, MMT and other permutations at large and small scales and see the differences. More specifically you can pick a specific spatial size (layer) and see what the layer looks like with each algorithm. The job of interpreting the results is still there...but the differences are quite dramatic.

Juan can best say if this approach is good- or if it is biased due to something special about the function/structures generated.

Anyway this was my attempt at approaching MLT vs MMT in a way that yields something obvious to compare.

If nothing else...the  image is pretty cool... you should give it a try!



-adam