Well… I don’t like posting incomplete work, but I feel this has its own entity being a H-alpha image. I state that it’s incomplete in that I have not completed the acquisition of R and G chrominance data.
So, for your viewing pleasure, I present
IC4628 in Ha - a faint, but very large emission nebula located deep within the tail of Scorpius.
The nebula is surrounded by beautiful open clusters which will really come out with the RGB chrominance data. At lower right is NGC6231, a rich open cluster known as the “Table of Scorpius” and shines at mag 2.6. The final composition of the image will be Ha:Ha+R:G:B which I will complete once I’m satisfied I’ve acquired sufficient data to make this image aesthetically vibrant.
Some info on this Ha image...
This was acquired over two nights (9th and 21st June) and has a total exposure time of 110minutes (just under two hours) using 10min subs. Yes, 10min subs. Typically, Ha images require much larger sub exposures to improve the S/N ratio, however this is defined by filter and telescope specifications. As the Ha filter I’m using (Custom Scientific 10nm) allows a higher penetration of energy than a 6nm or 3nm, shorter exposures are achievable. In addition to this as I’m imaging at F/5, the scope is reasonably fast – further reducing sub exposure time. The downside of the 10nm filter is that bright stars tend to bloat, though this is easily fixed in image processing. I have not done anything about the stars until I commence working with the chrominance data.
A few words on the processing this image…
Due to the dim nature of this nebula, I used MaximDL’s pixel math and multiplied all pixel values by 3. This increased the nebula brightness, but also white clipped the bright stars as they became higher than 65,500 counts. This is a compromise I’ve made and generally a good one as the stars will eventually be formed from the RGB information. Why did I increase the nebula brightness (or the entire image brightness) using pixel math? Simply, algorithms in particular, deconvolution don’t work well on dim (low count data). So, this image was then pushed through CCDSharp for two iterations of LR deconvolution. DDP in MaximDL – let MaximDL determine background level, then dropped it my 200 counts. MaximDL had the mid levels at 1800 counts – a little too high so this was dropped down to 1,200. Low pass sharpening as part of the DDP process was also applied. I feel this compliments the deconvolution step – well after blinking the images in MaximDL it appeared to make a positive improvement, and of course a negative one – noise. To reduce the noise, I used the PS colour range tool and adjusted fuzziness to only select highlights (details), then inverted the selection and ran the reduce noise filter. This cleaned the dim areas reasonably well. I could have perhaps been a little more aggressive on the noise reduction settings. I’ve discovered that many techniques I use for chrominance data simply don’t work well with Ha data. Actually, I’ve processed this image purely to represent the Ha nebulosity. I will process another Ha version when I start working with the chrominance data. The learning curve continues… (if you've read through this far, give yourself a pat on the back).
Hope you enjoy.
All comments welcome.
It is certainly an interesting object. I've only seen this object visually once and found it quite difficult to resolve. Of course, I wasn't using any special filters at the time. Maybe next time
Though I do have the upper hand with MaximDL's advanced tools.
...... typical, get new things sorted, then get nothing but TV nights.... ho hum..
Now have to wait a few days for delivery. Finally, a clean optical axis is on the way (I hope) for the FSQ.
