I have started playing with DrizzleIntegration recently. My interest was piqued by a discussion on CloudyNights, and I thought I'd participate in some testing. I have the somewhat unique position of having hundreds of frames on many of my images, as I like to stick with my target and get as much data as possible for the best results, and I love faint details. In my back yard, I often get over 200 subs, and I figured that many subs would probably work quite nicely with drizzling, especially with my slightly undersampled image scale of 2.14"/px.
Anyway, in my first few attempts, I was surprised to see that pixel rejection for sigma outliers did not seem to be working. In two of my integrations, I'd used an ROI in both ImageIntegration (when adding the .drz files and ticking "Generate drizzle data") and DrizzleIntegration with the "Enable pixel rejection" option ticked. I drew a preview around the original integration I got out of BPP (where I ticked on Generate drizzle data), and used that preview as a reference in the ROI section of both ImageIntegration and DrizzleIntegration. Both of these drizzle integrations that used ROI had a lot of hot pixels, some satellite trails, and in one this GIANT airplane trail (I've got an airport less than three miles away, planes often fly below 1000 feet directly over my yard... :\). The rejection algorithm in ImageIntegration, Winsorized Sigma Clipping, seemed to do wonders with the standard integration. No sign of hot pixels or trails of any kind.
I finally ended up trying a full size integration (no ROI) with both ImageIntegration and DrizzleIntegration...and voila, the drizzle integration was clean as a whistle. No hot pixels or trails, and twice the resolution of my standard integration.
So I think there is a big in how pixel rejection information is either generated by ImageIntegration when using an ROI...or translated by DrizzleIntegration when using an ROI. I did use the exact same ROI with both, and when doing so, rejection fails. When i do not use an ROI, rejection works. Is there any way this could be fixed? A lot of the time I'd like to use drizzling just for a portion of my image, say a galaxy in the center of a larger field. I often image at 600mm and 840mm to include surrounding satellite and neighbor galaxies and distant galaxy clusters, such as:
These images tend to be sampled barely well enough, or undersampled, and could benefit from drizzling.
