Processing a very noisy luminance image

[mstriebeck]

Hi,

A month ago when comet Lovejoy was high in the sky, I took some images ... and since then try to process them. The main challenge is that our backyard is very light polluted and at the time, the moon was up pretty high too.

Here is the calibrated and stacked image: https://drive.google.com/file/d/0B0KcEqw-v71VMWlBVEJBUldtMmM/view?usp=sharing

With a lot of trial and error, I could process it to this: https://drive.google.com/file/d/0B0KcEqw-v71VdXlrRzBRS3ZvLTg/view?usp=sharing

Here is what I did (mostly along the lines of this tutorial http://lightvortexastronomy.blogspot.com/2013/02/tutorial-nebulae-pixinsight-workflow.html):
1. DBE
2. Noise reduction (I tried TGVDenoise or ATrousWaveletTransfor - both with similar results)
3. DDP in Nebulosity (I couldn't figure out how to stretch the image with HistogramTransformation or DigitalDevelopment)
4. Back in Pixinsight: further noise reduction with ACDNR
5. Enhancing faint details with LocalHistogramEqualization
6. Further enhancement with ATrousWaveletTransform
7. Some minor stretching and enhancements with CurvesTransformation

What annoys me the most of my processed image is that the tail became really blotchy. I wonder if any of the masters here have any other ideas how to better process a suboptimal image as this.

Thanks
     MarkS

[avastro]

Hi Marks,
Not a master at all, just my try.
DBE, just some samples in the right top corner, MultiscaleMedianTransform noise reduction,histogram transformation and curve transformation to stretch carefully,TGVdenoise using a mask to reveal only the background, histogramT and curveT to tweak the final image.
Best
Antoine

[mstriebeck]

Thanks a lot Antoine. When you say that you used a mask for TGVDenoise, do you use it as local support or as a straight mask?

    MarkS

[avastro]

Hi Marks
The same mask was used to protect the comet, the tail and as local support.
I duplicated the comet picture and applied HistoTranform to reveal the comet core and the tail getting the background more or less black, then using AtrousWavelet and unchecking the first and second layers to smooth the mask. The mask is applied to the main picture revealing the noisy part of the picture.
That could be done also with range selection after a Histogram Transform, and ATWT to smooth.
Some preview in the picture can help you to set correctly the strength and the automatic convergence in TGV.
Antoine
http://www.astrosurf.com/avastro/