Collecting  Photons

1) Colour data from a polluted location

Images were calibrated in CCDstack and CCDsoft. In brief this is the procedure I followed: initially master dark and bias were subtracted allowing linear scaling of darks to fit for small temperature variation of the sensor (31 dark and bias for each master file). At this point I used a custom made script running in CCDsoft 5.0 to repair two bad columns of my sensor. This procedure was necessary only for the luminance data (bin 1). Finally the appropriate master flat (mean of 31 flat for each master file, acquired at bin 1 also for the RGB data) were applied. After registration a mild DDP of the colour RGB composition is shown here. The image is obtained by stacking 15x8' (for each channel) subs obtained from my site. The image is severely plagued by sky gradients and also by colourful artefacts caused by sky and stray light bouncing around the telescope and arriving at the sensor from odd angles, including light reflected from the not-so-opaque walls of the Wynne corrector. The scattered light is responsible of the horrible artefacts present at the edges of the frame. How I arrived at this diagnosis is a long story that will be told at a different time.  

2) The luminance data are no better

The luminance data are no better (53x8') even if they show the nebula a little more since there are no distracting coloured gradients. It was clear from the very start that my usual way of dealing with gradients and illumination artefacts could not be employed here. Usually, I use the DBE operator in PixInsight that, as all the similar tools for gradient removal, requires a fair amount of sky devoid of nebular features to compute a correct model for the sky background. Here there is not such a thing as empty sky. 

3) The reference RGB data

In may 2009 I brought the telescope to the dark sky of Monte Amiata during the yearly star party. Here I had the opportunity of imaging this area for about 3 hours under really dark sky. This time was not enough to acquire enough signal for both RGB and luminance, but provided an image of the RGB data with only a minimal contamination by sky gradient and no stray light artefacts. Indeed the difference between this image and the two above is striking: there is only a small linear gradient roughly aligned with the north-south direction.

The idea at the basis of the following processing is very simple: I will use this reference image to estimate the background artefacts in the RGB and luminance data I showed above. This "illumination artefact" can be cleared of any residual star images and can be subtracted from data 1 and 2. Finally the data can be composed together. Let's see how it is done.

4) The colour background model

The images shown in 1 and 3 have been scaled to 16 bits and saved in CCDstack. These images have received a mild DDP stretching: all of the following operations have been performed on the stretched images: it is important to notice that while on theoretical ground it would be better to operate with linear images, on the practical ground it is much easier to work with images that have received a mild non-linear stretching. A second important point is that great care has to be exercised to stretch the two images similarly: this is difficult to do because the presence of the background makes the evaluation of the nebula visibility difficult. Some trial and errors are unavoidable: one has to aim to produce a difference image that does not show any trace of nebular features after extensive stretching. The result shown here has been computed after exporting the scaled images to PixInsight and the levels have been adjusted to show better its structure. It can be seen that the colour gradients replicates well those present in the image 1.

5) Cleaning up the background model

On the model there are the ghosts of several stars. These are due to the fact that during the Monte Amiata acquisition it was very windy and that resulted in bloated stars. Closer inspection of the background image shows that the star ghosts are bright at the center since stars from image 1 are more compact, and darker in the periphery.

The stars have been eliminated in Photoshop after repeated application of the "Dust % Scratches" operator (filter/noise menu). I applied it 3 times with scales of 5, 10 and 20 pixels and a low threshold (3). Some residual haloes have been removed with the clone tool. It is worth pointing out that if the large and sharp stray light artefacts were not present, it would have been much better to create a spline fit to the background model (with the DBE operator of PixInsight, for example) instead of working with the difference image.

6) The corrected RGB data

The difference between images 1 and 5 was computed with the pixel math operator of PixInsight. It is very important that the operator allows to compute negative values. This is something that cannot be done in Photoshop. The data resulting from the operation can be automatically rescaled to positive only values at the end of the computation (check the appropriate box on the Pixel Math control panel). This operation could be performed also within CCDstack but I failed to have the "File Math" operator working properly on RGB data.

The resulting image is basically devoid of illumination artefact and the comparison with the original RGB data is striking. Just roll over with the mouse to see a blinking comparison

7) The corrected Luminance data

To apply a similar procedure to the luminance data first I computed an artificial luminance by adding the R, G and B channels of image 3. Then, the background illumination was computed as above computing the difference between scaled and slightly stretched luminance and reference luminance files. Stars were removed as above from the background illumination image. In this case the differences were computed in CCDstack.

Roll over with the mouse to see a blinking comparison between the luminance after and before background subtraction.

8) End result

The total RGB chrominance has been obtained by summing the two RGB data set. Stars were pretty bloated due to the bad conditions of the Mount Amiata data and, of course, because of the binning. The RGB image was processed with an erosion filter through a spatial masks for the stars only, in order to decrease the star size and to increase the colour signal in the star center.

The LRGB fusion was realised in Photoshop and the following processing, carried over in PS and PI was directed at protecting at best the star color and saturation during the several iteration of non linear stretching that were necessary to properly display this extremely faint nebula.