Gradient Merge Mosaic - Linear or Stretched?

[RadarBoy]

Hi Folks;

I'm playing with GMM for the first time as I am working on a 6 panel mosaic of the North America/Pellican area in Ha and OIII.

Some of my panels are merging better than others while linear data.  I suspect this may be due to me setting the gain higher by accident on my first panel of the grid. 

Is it better to stretch each panel, then run GMM, or is linear better?  If its linear, is there a good way to match histograms prior to running GMM?

Thanks!

Ryan

[RadarBoy]

I may have largely answered my own question with the video found here:

http://www.youtube.com/watch?v=H2r3ZYqCaP8&list=UUZvdEb-EGgL_rPx2iGuv17w

The linear balancing problems largely disappeared once I created a "master mosaic" using star align as suggested by the video.  Still, I'd welcome any further input/experience using GMM.

[georg.viehoever]

LinearFit can help with balancing the statistics. GMM can compensate additive offsets, but not multiplicative, non-linear differences or noise patterns.

You can use GMM on linear images -but there is a higher likelihood that you get artifacts around bright stars at the border of images. Workaround: Remove critical stars from all but one contributing image.

You can use GMM on non-linear images. You are less likely to get star artifacts, but it is much more difficult to get the different panels sufficiently similar since not even additive differences are linear anymore.


Georg