Page 2 of 2

Re: An approach to subduing Halos?

Posted: Sun Dec 09, 2018 2:16 am
by admin
prusling wrote:Thanks, Ivo, please could you post the log from this approach
It's essentially what I posted (unfortunately, the Layer module does not log the key presses at current).
The main takeaway is that the luminance mixer is just another way to get at a dataset blend that does not show halos, the only difference between this approach and simply choosing one channel, is that it does retain coloring (it separates luminance and chroma information). This may, or may not, be desirable. It's useful if you don't have separate channels, for whatever reasons (e.g. DSLR, OSC).


However, to apply the technique you're trying to apply in StarTools, just do the following;

1. You will get best results in all cases by prepping both datasets individually (e.g crop stacking artefacts and use Wipe) before doing data transfers/blending between channels. Use the Restore functionality to restore to Linear, Wiped before saving each dataset; this allows you to visualise (and work with) a stretched version (e.g. use Develop, AutoDev, etc.). This way, you shouldn't really need to match sky backgrounds, etc; Wipe will have done that for you.

2. Once you prepped both datasets, create your mask and save that away.

3. Now load the problematic data set (which should be linear if you followed instructions above), open the Layer module, open the "good" dataset in the foreground (which should also be linear if you followed instructions above). Load the mask you created. Apply a little Mask Fuzz if you wish. Upon clicking "Keep" you should now have a fixed up version of the problematic set. You now can verify if all is well by applying a temporary stretch. Save away this new linear dataset.

4. You can now combine as usual.

Re: An approach to subduing Halos?

Posted: Sat Dec 29, 2018 12:46 pm
by prusling
Hi Ivo, thanks again for your feedback on this. Please could I make a feature request (assuming it makes sense!) - when working with the Layer function on non-stretched data (including data previously restored to linear/wiped form) it would be very useful to have a (temporary) stretch operation within the function that could be applied to foreground, background (and the result) as it can be hard to visualise and tune the results otherwise. What do you think?