Quote:
Originally Posted by jase
Depends on the QE of the chip. If you're an narrowband enthusiast, you'd probably be nuts if you didn't go with an NABG cam. Unless of course, you want to do fast lens work where blooms are guaranteed.
Could be a few things Greg...
Do you perform normalisation manual i.e. select the highlight and background of the image manual? Normalisation is critical for good data rejection. I don't use auto (may once in a while). If you're not happy with the image weightings, you can always re-normalise. Sub weights determine the preference of the subs in the stack (specifically the data it will use to build the master frame).
When it comes to registration, Nearest Neighbor is the go as it preserves pixels, thus the noise statistics therefore making the outlier detection easier. Though it only works well with large stacks. For all others, go with Bi-Linear if the subs only need shifting or B-Cubic if they need to be rotated or scaled. In your image processing routine, avoid double registrations...particularly with luminance data as it will lower the resolution by skewing, rotating, etc the image data again.
As for the rejection algorithm, either STD Sigma Reject or Poisson Sigma Reject. The latter is preferred as it work very well on small or large stacks. Based on stack size and data quality, you'll need to tweak the sigma multiplier. Anywhere between 1.5 to 2 works well, lower if the data is bad as this will make the algorithm more aggressive. Blink the subs to see what has been rejected. If you still see "nipples" as mike put it on stars or a satellite/aeroplane trail not highlight in red (rejection marker), the sigma value is not low enough or the sub weightings need to be recalculated (re-normalise them again).
Apologies for this off-topic post...umm, yes, Ha filters...they're great!  |
....MAN.. I have no idea what ya talkin about Jase
but it sure sounds good...compared to my way too simple processing...and nipples
If there aren't too many subs with nipples and sattelite trails...wouldn't it be better just to discard the worst ones? All that fudge processing has to affect the final image quality, surely? You don't get anything for nothing as they say.