GOALS AND PRESENTATION
It would be easy to produce a "standard looking" image with this data. Truly minimal processing would be required, but in this case and at this time, for me that wouldn't be nearly as fun neither challenging.
The goals when producing this image were very precise. Just a glance at the image should reveal what these goals are, and I am somewhat satisfied with the results - meaning the goals were achieved to an extent. Clearly the main goal was to reveal any subtle and faint data - the data that typically sits right above the noise. A secondary goal was to do this while containing galaxy brightness and at the same time, being able to bring out small scale details in such galaxies.
To enhance details in the galaxies, a dynamic range compression process was applied to reduce brightness in the larger galaxies, as well as wavelets-based HDR enhancements. Preserving background illumination was accomplished by very careful gradient reduction and further non-selective
There may be some people who find all the "dust" in the background distracting, or the small details in the galaxies somewhat the result of processing and not a "natural" depiction of the field, but ...In order to appreciate this image you need to understand the goals set for it, then conclude whether those goals were met or not, rather than whether you would have aimed for different goals.
DETAILS IN THE DUST
In order to better see the "dusty background" I have also prepared a monochrome image that you can see here:
Click on the image for a larger version.
It may be surprising to see an image with all this dusty appearance in this direction, out of the galactic plane. To be honest, that is not for me to judge, but this is what came out of my data, and the processing involved was as careful as possible. Data capture wasn't perfect but all the data was taken from the same location, during the same hours on three different nights, and on nights I was very discriminative as to whether I should capture data for this project or not: all nights displayed an SQM reading of magnitude 21.7 or higher at the zenith at some point during the session, and were all consistently above 21.5 at any time.
Although most of the dust above the background you see in this image is likely from our own galaxy, if you look closely, on top of NGC 4435 (the "eyes"), you may see a thick faint tail moving towards 10-11 o'clock, and that is most likely pulled from that galaxy, not a foreground cloud. The "1" arrow in the image below points to this area. There's also some even fainter strikes visible (barely) going from M86 towards NGC 4435. The lines to the right of the "2" in the image show where these strikes are happening. You will need to go back to the larger version of the above image to better discern them (the green lines in the image below are covering the most visible strikes).
There's also some intergalactic "fluff" - though very diffuse in the image - from M87 to NGC 4461 and NGC 4473. And of course, the well-known interaction between IC 3481 and IC 3483 is clearly visible towards the bottom of the image:
Other than that, I personally cannot tell whether other faint signal is intergalactic, it belongs to our Milky Way or, suffice to say, might be an enhanced artifact during capture or processing.
The hardest part in processing this image was the gradient removal process. It's not that the gradients were complicated or severe. In fact, they were very smooth and subtle, thanks to the dark skies of the DARC Observatory, and if I hadn't gone after the fainter signal, it would have been quite easy to deal with it, but there was a great mix of very light gradients from so many frames, and it took me a lot of trials and analysis until I was convinced I had built a background model that would mainly subtract gradient signal and nothing else (and nothing more). Although it's quite possible that not ALL of the faint but visible data is 100% accurate, I believe most of it is. Here's some of the data and processing involved to "flatten" the master luminance data.
Below you can see the master luminance after being cropped (to remove "bad" edges due to misalignment between frames) and with nothing else done to it but a non-linear stretch tailored to reveal the gradients in the image:
And here you can see a non-linear stretch version of the final background model applied to the image:
As you can see the background model is very smooth and gradual even after a strong stretch - obviously when the background model was subtracted, it was in linear form and visually it looked like a completely dark image. PROCESSING
The processing of the data did not include at any time any curves transformation, DDP nor selective or masked histogram stretch. All data stretching was done by unmasked and non-selective non-linear histogram adjustments.
More specifically, after the gradient removal, a rather standard process was done to the luminance/lightness data, that mainly included:
- Slight deconvolution, with masks at three levels: local, global and external. The purpose of the local and global masks was to avoid the Gibbs effect, and the external (lightness-based) mask was used to avoid applying deconvolution to areas low in SNR, and increase the deconvolution effect as the SNR improved.
- A first non-linear unmasked histogram stretch.
- Masked ACDNR (noise reduction). The lightness-based mask here served the opposite effect of the external mask used for the deconvolution: applying the noise reduction only on areas low in SNR and gradually reduce the effect as the SNR improved.
- A second non-linear unmasked histogram stretch. This is because after the noise reduction, our histogram has been altered, and so it allows for a new adjustment.
- Masked HDRWT. This is the process that reduced brightness in the larger galaxies and also revealed some of the data and details in them. The lightness-based mask here is also helping avoid ringing.
- A third non-linear unmasked histogram stretch. Again, we readjust the histogram because after the HDRWT process we have a different scenario that allows a new adjustment.
- Slight masked sharpening using wavelets. The lightness-based mask here kept noise from being sharpened.
- Very light masked morphological transform. This process reduced overall presence of the largest stars, bringing them back to their form as they were prior to the last histogram adjustments. A star-based mask is necessary with every morphological transformation process, otherwise we would be applying the morphological changes to structures that shouldn't be affected by it.
- A last histogram adjustment.
- A last masked Laplacian sharpening.
In between some of these processes, I did integrate scaled before/after images via PixelMath at some stages during the processing, to apply a process only so slightly...