Elephant's trunk (IC 1396A) is a very interesting interstellar gas region in the constellation Cepheus with emission in Ha, OIII and SII.

To the right see my first image of the actual 'trunk', covering an area of about 1°6' x 45'. The data is used as image C in the mosaic below.

The hole nebula is more than 3° x 3°, so I decided to create my first narrowband mosaic with this nebula (my RGB mosaic of M31 can be found here).

Creating a mosaic with Ha and OIII images combined to a color image posed some questions for me:

  • When is the best point in the workflow
    • ...to go from linear to non-linear images?
    • ...to combine Ha and OIII to a color image?
    • ...to extract the background?
    • ...to do noise reduction?

Besides the workflow challenges, I faced additional problems with the images:

  • Due to an issue in the JNow vs. J2000 settings between Stellarium, APT and EQASCOM, the images were not properly placed (see image to the right)
  • Weather and moon changed a lot during the sessions. For images A and D, I only had half of the data I needed, and I am still waiting for clear skies. In addition, the moon has an impact on the OIII data, especially in image A.

 

I tried several options and found the following solution which fits best to my setup (suggestions always welcome, please use the comment section in the blog).

Image Capturing and stacking

All images were captured using an ZWO ASI 294MC pro, Skywatcher 200P on AZ EQ-5 with an ZWO dual-band filter.

Imaging time varied between 100 min and 288 min, with 4 min subs.

Stacking was done using SiriL with the OSC_Extract_HaOIII script, to which I added background extraction for each image.

Tip for rookies: Use BatchFormatConversion script to transform all your files from .fits to .xisf.

Stacking results (upper row: Ha, lower row: OIII)

A

B

C

D

Cropping

As can be seen in the images above, some images contain large areas which need to be copped. These were caused by a slightly different view prior and after the meridian flip.

I also realized that it is very important to crop areas with bright stars in overlapping zones in all images but one. Merging the mosaics later with GradientMergeMosaic fails with bright stars, even with a large feathering radius - so bright stars should only reside in one image.

Tip for rookies: Apply the same cropping on Ha and OIII by cropping one image, open the history explorer of the image, then drag the crop icon on the other image. This way, Ha and OIII are cropped to the same region.

Cropping results (upper row: Ha, lower row: OIII)

A

B

C

D

Noise Filtering

As two of the image sets (A & D) had a much lower exposure time due to weather conditions, their SNR is much lower. This might not be obvious in the isolated images, but becomes visible in the merged mosaic.

I decided to reduce their noise visually by comparing them with the 'better' images. ACDNR was applied to the nebula only, using a mask for stars and dark areas. Parameters were adjusted until the noise was similar to the images with longer exposure time.

Tip for rookies: You can combine masks using PixelMath e.g. by max(mask1, mask2).

 

Filtering results (left: unfiltered preview, right: filtered preview)

Tip for rookies: You can copy a preview from one image to another by just dragging the preview tab to the left border of the target image.

Plate Solving the Images

In the next step, the images need to be plate solved using the ImageSolver script. With this step I struggled most. I tried to save calculation time by combining the corresponding Ha and OIII images to a pseudo-RGB with PixelMath, but the script often refused to solve these images. Plate solving also failed with non-linear Ha and OIII images, maybe because the nebula is too dominant after stretching.

So I used the script on the linear images, adding all 8 images to the list and using the standard parameter set.

Preparing the Mosaic

The script MosaicByCoordinates places the images at the plate solved position on a black background, whose width and height depends on the mosaic and is calculated by the script. The manual of the script proposes to do a background extraction in advance but with my image set, I received better results by applying this step to the whole mosaic.

 

Results for Ha

Merging the Tiles

Prior to merging the tile with process GradientMergeMosaic, the histogram was stretched with STF and HistogramTransformation. This seemed to work better than on linear images, at least I had noticed more artifacts on bright stars when applying GradientMergeMosaic to linear images.

It is important to know that GradientMergeMosaic handles overlapping images in the way you order them in the list. The bottom image will be placed on top. This can create results of very different quality if you have large overlapping zones. So either shrink the overlap or play with different image ordering – maybe this can already improve the result.

 

Results for OIII for different stack orders

Stack order ABCD

Stack order BDAC

Stack order BCDA

Final results for Ha (left) and OIII (right) with Standard Parameters

Background Extraction

Background extraction is maybe the most challenging part. The whole mosaic is filled with nebula, it is hard to say if there is a gradient at all. When using DynamicBackgroudExtraction, the process always flattened out some interesting areas. So I skipped background extraction completely.

Removal of Stars & Star Mask

Starnet2 (accelerated with CUDA) was used to remove stars from Ha and OIII after cloning the images. The OIII image was used to create a star mask.

 

Results for Ha (left), OIII (center) and star mask (right)

Combining Ha and OIII with PixelMath

The starless Ha and OIII images were combined with PixelMath as described here. Parameters used:

INCLUDE = 0.35, REJECT = 0.99, SLOPE = 1, DIFFTOHA = 0.2, MIX = 20, BOOST = 1.5

Then a slight ArcsinhStretch was applied-

In tile A (upper left corner) we can see a circular image artifact from the OIII channel caused by the full moon. I hope this will go away with additional subs. Also, the gradient merge was not perfect.

The Ha and OIII images with stars were also combined with PixelMath. This creates the stars we will combine with the nebula in the next step.

Combine Nebula and Stars with PixelMath

Using the star mask, the nebula was merged with the star image using PixelMath.

Color Adjustment with CurvesTransformation

In a final step, stars were boosted a bit using the lightness channel in CurvesTransformation (left) and the nebula colors were adjusted in red and blue channels (right).

With some noise reduction post-processing in Photoshop 2022 using Topaz DeNoise AI, the final image looks like this.

Obviously, there is still lots of room for improvement. Additional subs are needed for each tile (and a monochrome camera with a filter wheel would also help a lot) - but for me, it was important to work out a suitable workflow, speeding up the process once I have collected more data.