From a bracketed set of exposures, create a single High Dynamic Range image.
Reverse chronological order (most recent at the top)
Figure 1 - a 4-bracket set of exposures, 2EV apart.
Third attempt, blur size 5 pixels, and adjusted one of the shifts by one pixel which really mitigated the red edge. [At this focal length, 140mm) the Moon shifts enough, 3 pixels, in 2 seconds, that I have to move it]. I half-solved the problem of dimming in the middle of the crescent. I needed to multiply the short exposures, but oddly not by the difference in EV. Perhaps some operator error in there? In GIMP using curves it is surprisingly hard to brighten the earthshine without blowing out the crescent.
Next step:
Look into eliminating blooming before compositing
Older stuff below
Here's the second attempt (with a mask blur of 9 pixels). Obviously there should be no dimming in the middle of the crescent. I have yet to bother going after the bloom (one step at a time). Simply using -auto-level when converting from an average of 5 frames. And I have not tried to minimize grain or do any other processing.
Older stuff below
Progress! Thanks to snibgo's help (https://www.imagemagick.org/discourse-server/viewtopic.php?p=172589#p172589) , I've been able to "import" my Canon CR2 raw images into XYZ linear colorspace!
On the left is XnView's conversion from CR2 to tiff. On the right is ImageMagick's conversion from CR2 to XYZ linear colorspace, then down into sRGB. You can see for XnView a bunch of "bleeding"artifacts on the earthshine side of the blown out pixels. Promising!
Next steps are:
1) combine two of the exposures, then convert back to normal display range.
2) Mask the blown-out pixels before compositing
3) Look into creating a good enhancement curve
4) Look into eliminating blooming before compositing
ImageMagick commands (from http://im.snibgo.com/lincamraw.htm#scref ):
Install notes: [Windows!]
Older below:
This is a tough subject. I'm not happy with enfuse output. Part of the challenge is eliminating the blooming/bleeding, and the other is doing this in batch because manually doesn't work for 500 sets in one time-lapse event.