Quote:
Originally Posted by Pilgrim
Thank you for kind words, Mathew, David, Adam, JA, Fred and Marc! 
The problem with algorithmic removal of stars (no matter neural networks or masks) is that while they work well for bright stars, they produce some artifact in border-condition stars and don't work for extremely faint barely-distinguishable background stars(which are a million times more).
The physical stars removal by subtraction of a continuum frame has the advantage that while giving some unwanted artifacts on bright stars,
it works good on medium, works well on weak and perfectly on an extremely faint indistinguishable "porridge" of background stars.
As a result, you get not only the removal of stars, but a much flatter, less noisy background, although you didn't even realize that part of its "noise" were groups of super weak indistinguishable stars barely leaked through the narrowband filter. Another bonus is that after retouching of artifacts, you can substract that starless H-alpha from the continuum image and this give you nearly perfect stars mask for future post-processing operations.
The closer an object is to the Milky Way, the stronger this effect is.
(It is a pity that there is a limitation, the method is suitable only for emission nebulae, in other cases there is no alternative to neural networks or masks-based wizardry.) |
All That you say is true, the continuum filter can be very effective...but. I often image 60 or more hrs with Ha for lum. I have found the continuum exposure time needs to be near or the same as the Ha image exposure for optimum results. Thats a hell of a lot of extra effort just for star removal !. I have found using that extra exposure time on data collection instead, and more fiddling in post to remove artifacts with software star removal gives a better final image. I do get your point though, a purist would go the extra mile for perfection without cheating
.