Astro Pixel Processor

Trying again with a mosaic of Eta Carina

One thing I found when working on the Eta Carina mosaic was the large amount of stacking artefacts around the edges of the frames. For a normal image this is not a problem as you just crop those bits off. When you’re building a mosaic, however, you need all the image and can’t lose any of the borders. Also, the stacking artefacts can cause problems when you come to stitch the panels together. As I found out…

We finally had a clear night arrive, after so many nights of cloud or rain. (As I write this, the rain is coming down again. One clear night of imaging then solid clouds and now rain…)

As I mentioned in the other mosaic article, I have installed the image flattener / reducer to bring the scope back to a focal length of 357mm and to ensure the stars are not elongated in the corners of the frame. I didn’t use any filters, either.

I collected what I feel to be the minimum required amount of data - one hour per panel. I would have collected more except for the small fact that it was a Sunday night and I had to be at work the next morning. (Even so, I got to bed at 02:00, and then had to be up at 06:00…)

However, the problems started appearing when I tried to stack the images. SiriL 1.0.6 still crashes when I try to remove calculation points when doing the Background Extraction, and Siril 1.2.0b2 didn’t appear to give consistent results when doing BE across all 4 panels. (Kind of obvious when you think about it - it’s processing each panel separately with no knowledge of the other panels.) Also, the stacking artefacts along the panel borders were (to put it mildly) terrible.

I was able to get 3 of the panels stitched together in Affinity Photo, but was not able to align the last one. Even spending a large amount of time manually moving and scaling and rotating the last panel failed.

Introducing APP - AstroPixelProcessor

A question on the IIS forum had multiple options provided - PixInsight, AstroPixelProcessor and even Microsoft ICE. I did some looking around, to see that both PI and APP offer free trials and offer complete mosaic stacking and stitching workflows. APP looked to be the easier option, so I signed up for the free trial after reading a few web pages and watching some tutorial videos.

Overall, I am very impressed with APP. Yes, it has a steep learning curve, but for the most part, the default values work well. I expect that APP experts will tweak the settings to get the most out of it.

I used APP to stack, integrate and register each panel (I missed a setting, so should go back and do it all again when I have an evening) with excellent results - and NO artefacts on the panel borders. Oddly enough, the mosaic had some very strange ‘swirls’ of stars in one quarter where something went a bit haywire. Not to worry, I then opened Affinity Photo to use it’s ‘Panorama’ tool.

The image below is the end result of the post-processing, and scaled down from the initial mosaic. I’m still not happy with how it’s turned out, and might go back to rework it all from scratch and spend more time on the post-processing. It really is my weakest area.

Image details:

• Gain: 50
• Offset: 10
• Temperature: 0 degrees C
• Exposure: 20 seconds
• Lights: 180 (60 minutes of data per panel)
• Darks: 50
• Flats: 50
• Dark Flats (Bias): 50
• Stacked and pre-processed in APP, then stitched and processed in Affinity Photo

---

---

APP - Gum15 Reworked

After my success with the Eta Carina mosaic, I wondered how well APP would cope with the Gum15 data that I collected. (Remember, I had to discard several hours worth of data because I could not get it to work with the data from the dark sky site.)

APP can handle multi-session imaging, where data is collected over several nights. This is perfect, as I had 4 sessions worth of data.

I fed all 56 gigabytes of data (900 light frames and a couple of hundred calibration frames) into APP, checked the various settings and hit the ‘Go’ button.

I don’t know how long it took to process everything as I went off and did other things, but when I returned I had a very good looking image waiting for me. I suspect that the high speed NvME drives, fast CPU and large amount on RAM in the processing machine helped.

APP had stacked all 4 sets of data, including aligning images where the camera rotation was slightly different between imaging sessions, without any problems.

I did very little post processing in Affinity Photo - a crop to remove the not-so-good parts around the outside of the image, and then a small amount of tweaking things like brightness and colour saturation.

Something tells me that APP will be worth the purchase price when the trial expires in a few days.

---

APP - Eta Carina with filters Reworked

Continuing with my re-working of old data with APP, I looked at my favourite image of Eta Carina - that taken with two sets of data. The Optolong L-Enhance filter data set and the UV-IR Cut filter data set provided a good way of looking at how well APP could process two sessions of data that were wildly out of alignment.

I knew from the Gum15 work (above), APP could easily digest multiple imaging sessions, so it was a chance to see how well it could align images.

As it turned out, it handled it very well indeed.

This just shows how important it is to not make changes to your optical train…

After all the usual post-processing, I ended up with the following image. I’ve not yet done a side-by-side comparison with my older image. I suspect that major differences between them will be painfully obvious, and caused by differences in how I did all the post processing.

---

For fun - a starless version

There are some tools around that will remove the stars from your image. The idea being that you can separate the stars out, carry out your processing on the nebula layer without impacting the stars and then re-add the stars into the image (after processing them to your liking).

The new (beta) version of SiriL has integrated support for ‘starnet++’. The image below is it’s output. As you can see, not all the stars have been removed, with four large stars left behind. If I was doing serious post-processing, I’d remove these and rely on them being put back with the star layer.

---

Now for a DSLR image - NGC55

I’ve been looking for older image data to see how well APP deals with it.

I’m pleased to say that APP has handled it very well. The data here is from 2022, before I got my dedicated astro-camera and was using my DSLR. Oddly enough, for this data set I had taken the whole quartet of lights, darks, flats and bias frames. Usually I had only shot lights and darks, hoping that whatever processing software I was using would cope with the lack of flats and biases.

Galaxy season is approaching, and I can’t wait to try this one again with the ASI183MC-Pro to see how it turns out. (I just need to remember not to use the L-Enhance filter on it!)

The big difference between the two images is the very heavy blue / cyan tint in the old image. I was still getting the hang of my software. In trying to make an image that looked good, I have pushed the brightness too far and into the blue zone. Nothing wrong on the whole with that, as I can’t decide which image I like the best.

It is tempting to go and rework the APP stacked image and see how far I can push it to look like the old image.

As the saying goes - a good astro image is one that you’re happy with. So much of the processing comes down to personal preference, the colour calibration (or lack thereof) and how our eyes see colour. What looks good to me might look terrible to someone else. It’s very much a matter of taste and how we go about processing our images.

• Astro