TVG Denoise - creating unwanted contour lines around galaxies

[RickS]

A clipped lightness/luminance mask is often useful.  If you have a grayscale luminance then stretch if linear and then clip the blacks with HistogramTransformation.  If you have RGB then extract CIE L*, stretch if linear, and black clip it.

These masks are good for noise reduction (inverted), saturation boost for RGB images, sharpening, etc. 

Cheers,
Rick.

[dpaul]

Rick

Thanks for this, I've never used such a mask but will try

David

[dpaul]

I've just been using a mask to protect a galaxy (M66) whilst using MLT for noise reduction. This was on both RGB al L images (still Linear).
The masks in each case were created by creating a clone then using range selection to only include the galaxy. I also made is blurred  it with the smoothness slider.
In both cases I still some subtle contours - see attached images, the RGB is still linear and the L is stretched but in both the contours are visible.

How can avoid these contour lines?

Thanks

David

[wimvb]

Did you apply deconvolution before noise reduction? These contours around saturated stars are an artefact of deconvolution.
You can try noise reduction on a non-deconvoluted image.
The method described by Jon Rista has never resulted in such artefacts, as far as I know.
If you have the possibility to attach/link to the stacked but otherwise unprocessed lum master (xisf format), I can have a look at it.

[dpaul]

Hi Wim,

I think I did deconvolute before noise reduction (could be wrong). I also made a mask to protect the galaxy by using arcsinhstretch on a clone and stretching / black level until everything but the galaxy was clipped. This I believe may have also create weird contours.

Either way, I've had another attempt, definitely convolution done first. Also a mask created with a light extracted clone then range selection to just show the galaxy and made is blurred with the smooth slider.

Tthe attached result seems a lot better, no contours.

The only issue I still had was creating a star mask and doing MT to reduce star size. No matter what star mask settings I chose, I couldn't include the brightest star near the galaxy. So some of the stars are still a little bloated but at least not too over-processed now. Advice on star masks would be useful.

Thanks

David

[wimvb]

To include stars on a bright background:
Extract luminance from the stretched image. Apply HDRMT with layers down to 3. Then create a star contourmask from this image. You may need to blur (convolution) the image a little to produce a good mask.
General settings for contourmask:
Large scale and small scale = 0 - 1
Compensation = 1
Smoothness = 8 or less
If you need to soften a contourmask, it's faster to do so afterwards with convolution.

[dpaul]

Hi Wim,

Thanks for the note -

Still struggled to get a decent star mask from the stretched luminance image. HDRMT did help reduce the dominance of the galaxy core but I still barely detected the nearby brightest star.

In the end I redid the whole thing (to get better colors, etc) and taking a bit more care. Everything was in the order that you suggested except I did MT just prior to stretching which was much easier with the star mask generation. Also I did arcsinhstretch for the RGB then a little bit of HT too to stretch a bit more - better contrast. It meant the RGB was more balanced with the HT stretched luminance Image. The end result seems to be better and stronger colors.

Incidently, I tried PCC and color calaibration and the latter was the one final chosen, the colors seemed better 'on this occasion'. Sometimes I get the opposite.

See attached final image, not perfect - still a bit noisy. The reason is that the mask I used for the galaxy when reducing noise was probably too wide - better if I'd just protected the inner spiral arms.

Thanks

David