My first Rosette

[Mike in Rancho]

I have not acquired any new data on this (and likely won't before the month runs out), but have been experimenting with various restacking options on the original files. I like experiments.

When it was still bicolor, I did run DSS' Bayer Drizzle on it. Weird results, and for various reasons ST doesn't really like the data. I couldn't get SVD to work worth a darn. And this was even after it has been binned down to clear up the uneven and scattered results of the CFA drizzle.

The SHO versions above were just standard DSS bilinear. I next tried out DSS superpixel, as another non-interpolated option for "real" data, and just left it at the resulting resolution. Which is about the same as the 50% bin I've been using on this target anyway.

Finally, I was going to try those out in ASTAP, but it appears Han removed them for not performing as well as expected or hoped. He did add his own new demosaicing, Simple C, which is what I tried out.

I tried to process and color as identically as I could, sometimes with same settings, and sometimes by eye. But once the underlying data changes up a bit, things start diverging and so perfect matches aren't easy it seems.

This quad compare of a zoomed region has DSS bilinear top left, DSS superpixel top right, DSS Bayer Drizzle bottom left (bicolor), and ASTAP bottom right. Each fully processed, but not necessarily finalized.

Rose quad CFA comparo.jpg (491.73 KiB) Viewed 773 times

It does seem to me that the ASTAP version is a pretty good compromise of colors, stars, and clarity of detail, and ST seemed to like working with the data as well. And for whatever reason, unlike the DSS stacks, I didn't seem to have to run separate wipes on each filter -- Wipe handled the ASTAP composite in one fell swoop -- nor did I need to crop off the left edge due to remnant gradient.

It takes longer, but the quality here may beat out DSS' speed. I'd still put DSS straight bilinear in second place though.

I know we are supposed to just pick bilinear with whichever stacker, but, any thoughts on the new ASTAP demosaic? I will probably use these stacks to put together my final version of the Rosette. I still haven't really looked at what Siril offers for options, but I think it's even slower than ASTAP.

[Mike in Rancho]

Martin's post reminded me to come put a finishing touch on this one.

I ended up using the ASTAP stacking and demosaicing, finding it to be just a bit cleaner than the DSS with bilinear. Same data, I never went back out to add integration, so just the 4 hours of L-eNhance and 4 hours of SII. The G and B were extracted from the L-eN and combined as (2xG)(B) for my OIII, and the SII and Ha were just extracted R's from the SII and L-eN filters, as sort of explained in another thread. I'm still not sure if I am doing my mono extractions right (interpolate setting?), but that will require further testing I haven't got around to yet. Things still worked though.

The color mapping and balance probably isn't too far off from the prior renditions, as I kind of liked it, but I did try to feed back in some more green/Ha. Just a bit, which lightly starts showing up along the inner periphery mostly.

Not sure if this will work? Will try a cross-link where I have the 1600px, 1.7Mb jpg at 100% compression. Crossing fingers...