Open Cluster NGC 1647 in Taurus and Comet C/2022 E3 (ZTF)

13 February 2023

When I first heard about comet C/2022 E3 I checked its position only to find that my house blocked the view. I had given up trying to get a picture of the comet when my neighbour and his son came over last week and asked if I had seen it. When I said it was on the wrong side of my house they corrected my mistake. A quick check with SkySafari confirmed that I was wrong and the comet was in the perfect position for me to photograph it.

Two days later the skies cleared and I had my telescope out for my second ever comet image. This time with a telescope instead of a telephoto lens. It was pure luck that on this particular night the comet was close to the open cluster NGC 1647 and that both were perfectly in my field of view.

I asked some friends how long to expose and whether to guide for the stars or for the comet. With their answers I decided to use standard star guiding and a rather short exposure of 30 seconds. I took 303 single exposures, of which I used 240 in the end, as the others were just too early in the evening and therefore too bright to be usable.

With that, the image acquisition part was done and quite standard compared to "normal" deep sky images. But when it comes to processing fast-moving objects, it's a different story, and it took some time and new software to get a satisfactory result.

The approach I used is described by Adam Block (YouTube) and consists of the following steps:

1. Do a normal processing of the image using WBPP in PixInsight.
2. Separate the stars from the background in the resulting image and keep only the stars.
3. Comet-align the registered frames that come out of the WBPP script using the CometAlignment process.
4. Remove the stars from the comet-aligned images.
5. Stack the comet aligned images.
6. Combine the comet-aligned stack and the stars from above to obtain the final image.

The key step is step 4, removing the stars from the comet aligned frames. Adam uses StarXTerminator by Russel Croman but mentions that Starnet++ should also work. Since StarXTerminator is a paid program that's been on my shopping list for quite a while, but I hadn't got around to buying it, I tried the process using Starnet++ (V2). Creating the starless versions of the comet-aligned images took about 20 hours and ran over one night. But the stacked result was far from optimal, certainly better than stacking the original comet-aligned stars, but there were still quite a few artefacts. To be fair, I could have played around with the stacking parameters for a bit longer and possibly achieved a better result, but since I was going to buy StarXTerminator eventually, I figured now was as good a time as any.

Removing the stars with StarXTerminator was quite a bit slower than using Starnet++ (I don't have an nVidia graphics card and can't use CUDA acceleration): It took about 2 days to remove the stars from the comet-aligned images, but the result was worth it. The stacked image was much easier to process than the one created with Starnet++.

However, the image was not perfect:

The stacked image shows a hole in the comet's tail, and I have found that all starless images have this particular hole to some degree. I suspect this is due to the fact that there are stars in this position that have been removed by StarXTerminator and not quite correctly replaced with cometary tail content.

Various tests with the inpainting brush of Affinity Photo showed promising results, but left me with a bad feeling. So I kept looking for alternative approaches to get rid of the hole in the comet trail and came up with the idea of using a stack of comet-aligned, but not star-removed subs - just for the problematic part of the tail:

This was the solution I was looking for and the rest of the process was quite straightforward.

To sum up the whole comet imaging experience: I learnt a lot, got some new software that I had wanted to buy for a long time, got lucky with my field of view and ended up with an image that I really like.