StarXTerminator 2.x: I’m sold

I’ve had mixed luck with StarXTerminator. When I first came across it, I found it worked better than StarNet for the images I was processing. Those were with the RedCat 51 at 250mm F/L. It was wonderful, so I bought a license and considered worth every penny.

Then I started doing some wider field images with my Rokinon 85mm F/L camera lens. Both sets of images were with an APS-C sized sensor, either the ASI 2600MC or a Canon T6i. At 85mm F/L, I was getting ugly block artifacts with StarXTerminator, so StarNet was the tools of choice. It wasn’t perfect, but it was definitely usable.

Then I saw Adam Block’s video/interview with Russ Croman, StarXTerminator’s author, discussing the recent update. I’m usually pretty diligent about keeping PixInsight up-to-date, but I hadn’t pulled the most recent version, so time to click that update button.

Since I was in the middle of processing a set of images of NGC 1499 (the California Nebula) taken with the Rokinon 85mm, time to feed those in. I had already done so with the old version and it had those ugly block artifacts. The new version with the new AI did a much better job, no blocks. Using the two alternate AI datasets, the version “11 lite” and “11 lite no-noise” both had mild to moderate block artifacts. The “11 lite” used with larger block eliminated the block artifacts, but also started smoothing out structure in the dust clouds, something I definitely did not want to do.

Above you can see the blockiness appear and disappear as you roll-over the image.

In both images there are four small areas that look like star halos have crept through, two in the upper left, one in center-left lower half, and one in the upper-half on the right. Let’s look at those more closely toggling between the original and the starless result at something close to 1:1 (close because I think the image is probably getting downscaled for your brower; the original capture was 1:1).

That’s Not a Star

The top left bright field is shown below. The mouse rollover should toggle between the original view and the starless view.

Surprise! It’s not a star or halo after all, but likely a bit of the dust that is near a star and more illuminated than the rest of those dust filaments. A definition score for StarXTerminator.

Hmm, Maybe It’s a Star

The next one to the right a bit less clear.

This really looks to my eye like a star embedded in the dust cloud. But…I should probably create a 3D contour of what this looks like because while it looks like (and quite likely is) a star embedded in the dust cloud, it’s not popping out quite like the other stars. I’ll give this a pass. Maybe I’ll complain later in the processing, but for now, I’m okay with this one.

Is That Even There?

If you’ll scroll back up to the original comparison of the AI 11-lite and AI 11 images, you’ll notice that there are only three bright spots; the one in the center lower-half isn’t there at all in the blocky image. Of course, there are some other details lost in that image, too, but still.

This is totally a win for StarXTerminator. Those two lobes are definitely not star halo, and definitely not stars, but they are definitely in the original as illuminated dust clouds.

Nope, That’s Not a Star Either

The final bright spot on the right side of the image

is yet another win for StarXTerminator. That bit of illuminated dust almost certainly has a star buried in it, but it’s not a star and, even better, some of the stars nestled right up against that cloud are removed very cleanly.