In astrophotography, an image will be all the more successful as the exposure time and the number of individual exposures are high. All this to achieve the best possible “Signal-to-Noise Ratio” (or SNR). If a single amateur astrophotographer can stack several hours of exposures, several astrophotographers should be able to stack… many more!

The original idea comes from the project organized by Yuriy Toropin on M51, a processing of which is visible here: http://www.bcsatellite.net/bao/m51.html / https://astronomy.ru/forum/index.php/topic,74202.400.html

The idea of stacking images from different sources can pose many problems, among which:

  • The size (or sampling) as well as the orientation, which will change from one setup to another.
  • The amount of signal, which varies depending on each instrument/imager combination.
  • The calibration data (dark/offset) that change depending on the imager.
  • The different optical effects from one instrument to another: vignetting (and flat field), chromatism, diffraction spikes.
  • For color: the calibration of the RGB channels in order to obtain a “balanced” image.

How to proceed

To successfully combine raw images taken by several astrophotographers, a common target object must be determined — one that is accessible to several amateur astronomers using different imaging equipment. Then, a few imaging “rules” must be defined to assemble similar content for the same object (orientation, field of view, etc.). And finally, all the images are combined using pre-processing software to produce an image with a high SNR.

Implementation

I had attempted to initiate something within the French-speaking community Webastro (see the discussion on the forum). But it didn’t really take off. To validate the principle, a test had been carried out by combining two very different sets of images: one taken with an 80mm refractor (Orion 80ED) and the other with a 280mm SCT (C11). Each with a different CCD camera. The test focuses on M27, the Dumbbell Nebula, in the H-alpha layer. Here are the starting raw frames:

The exposures are as follows:

  • 80ED side: 6 raw frames of 10 minutes
  • C11 side: 3 exposures of 5 minutes

Many thanks to Goto24 for lending his raw frames!

Pre-processing protocol

First of all, one should avoid using the automated pre-processing workflows built into software (such as WBPP in PixInsight), at least because calibration varies from one camera to another. The procedure is as follows:

  • Calibration performed independently for each set of images (dark/offset/flat).
  • Manual upscaling of the smaller images (using an integer factor to avoid excessive resampling) so that all raw frames have a similar sampling — in order to help the alignment software.
  • Alignment, using as reference an image from the set with the highest resolution.
  • Signal adjustment of each set. The histograms must be similar on an unstretched image and with the same bit depth (16-bit). To achieve this, in the pre-processing software, one can:
    • Set the images to the same mean value
    • Apply a proportional coefficient to the histograms (this corresponds to a linear stretch). One can use dedicated functions such as the “LinearFit” tool in PixInsight
    • Align the sky background — this corresponds to an “offset”
  • Integration, using a good rejection algorithm to account for the noise and artifacts that will differ between sets of images.

Here is an integration test for each set, as well as a combined integration. The test shows that the resulting image is indeed a combination of both sets, with advantages and drawbacks:

  • Intermediate resolution between the two setups (resolution is lost compared to the set with the finer sampling).
  • Improved signal-to-noise ratio compared to a single set.
Cliquez pour agrandir

Results

The most convincing attempt was made on M81. Here we have three sets:

  • Two nearly identical sets, using a CFF 300 at f/5.4, and a KAI-4000 sensor.
  • One set using a 250/1200 Newtonian and a KAF-8300 sensor.

The total exposure time is 6 hours in luminance, plus 1h30 in color over a single night, which is quite good for mobile imaging! The composite is visible >>here<<

The “simplest” case of compositing is well known and consists in assembling the color layers at a lower resolution compared to the luminance layer. This allows “binning” the colors, which was interesting in the CCD era but less so with CMOS. On the other hand, one can now use a second, lower-resolution optical system for color (typically a refractor mounted side-by-side with an astrograph) in order to capture the full LRGB palette in a single session.

We have a few examples with M109, NGC 772, M63, M101, and M1.

Finally, since we have two twin configurations with the 2 RCs, we can easily combine our raw frames, as with M33 where we quickly accumulated 35 hours of exposure for a fine LRGBHa palette.