"Washed out" look

[joelshort]

I'm slowly learning PI and reprocessing some old images, both to learn and to improve the images.  I've noticed that I tend to get images that are more "washed out" and I'm not sure if this is just how the images should really look when processed or something I'm artificially doing. 

Take a look at the following image of NGC6914 and compare the reprocessed version with PI and the old version with that other software:


This effect does tend to happen on images that are dominated by nebulosity over most of the photo.  Any thoughts on this and what I am potentially doing wrong?  Whatever I seem to do regarding processing this is how it comes out. 

[Geoff]

Try boosting the saturation.  Apply a mask to cover the background then use the saturation option in Curves.  Thumbnail attached

[guest11353]

My image looked just like that when I finished my workflow yesterday

This process here worked perfectly

http://harrysastroshed.com/pixinsight/pixinsight%20video%20html/pixinsighthomeinter.html

Local Histogram Equalisation - really brought out the contrast - https://sites.google.com/site/welfordobservatory/home

David

[joelshort]

Thanks for the replies Geoff and David.  I think a combo of those two suggestions will produce a more pleasing image.

[Zocky]

Joel,

A small adjustment on CurvesTransforamtion can do wonders

[oldwexi]

Hi Joel!
In your processing the Dynamic BackgroundExtraction did not remove all of the largescale parts of the left side of th Image.
Therefore your tries to make the darker parts more visible were not succesful.

So i did attack this simply by running an ABE first without(!) Change of the Image and kept only the Background Image (the LS part).
From this background image i pulled the L und put this L as mask over the original image.
Than ran a second ABE now with(!) changing the original very minimal because of the mask...
This removed carefully a part of the largescale artefact in the left background of your Image.

Than pulled another L out of this result and pushed it hard with 2 LocalHistogramm equalizations  (64 and 220)
and a little curve this got me the below attached LUMINANZ Mask

Using this LUMINANZ mask INVERTED i ran 2 LHEs, one small and one large, over the color result from before.
Finally a very little tweek with curves got me the below result.


New Image


LUMINANZ MASK



Gerald

[joelshort]

Thanks Gerald.  I figured that the root of the problem was with the initial DBE but nothing I did seemed to correct that left side of the image.  How can I adjust the DBE so that the large scale parts are subtracted in the first place? 

I will try your process and see how it turns out...or (more likely) ask some follow up questions!   

[cdavid]

Wow thats an impressive result!  I had never considered that process method. Thanks for the tips!
Carlos

[joelshort]

Gerald,
I'm trying to follow the steps of what you did, but I'm very new to PI and don't quite get this first step:

So i did attack this simply by running an ABE first without(!) Change of the Image and kept only the Background Image (the LS part).

What setting do I choose in ABE to get the background image (LS part) you are referring to?  I see no reference to "LS" anywhere.

I think I understand everything after that.

[oldwexi]

Hi Joeal!
Took the standard parameter with ABE.
The 3 boxes at the bottom of the process were set to:
Subtraction.
The Discard Background model (what i called LS (Largescale) box was off,
because thats what i wanted to use for pulling the Lumiance out of it and us this as a mask
for the real ABE with standard parameter.
"Replace target "was also off in this first maskcreating ABE step.

Hope this helps

Gerald

[cdavid]

Hi Gerald....I'm also trying to follow along on your method.  I did note one thing that had to be changed in the ABE default values....and that is the output scale for the background needs to be set to 1.0 to generate an image that can be used as a mask.

A couple questions....when you say you do a slight curves..do you mean saturation....RGB...?  Is it a contrast "s" curve or just a slight elevation?

Also when you ran the large and small LHE on the final image....what exactly were your settings?  Was the amount ~.3 to .4?

Thanks!
Carlos

[oldwexi]

Hi David!
The LHE settings were
200 and 0.5  for the large and
34   and 0.3 for the small.

The ABE was set to Subtraction, rest to standard.
The median of this ABE-background Image came out with
0.30, 0.23, 0.24

The 4 points in CurvesTransformation were set on RGB/K to:
x, y
0.07113,  0.07113
0.144, 0.12
0.3158, 0.269
0.502, 0.497

the same Parameter were used for the second Curves Transformation too.
As mask was used the inverted  Luminanz for these 4 steps.

Hope this helps

Gerald

[cdavid]

Great Thanks
carlos

[joelshort]

Well Gerald, I tried to follow the steps you outlined and I did see some improvement, but not like you were able to achieve.  I think my problem has to do with the LUM mask.  I'll keep experimenting.

What I'm wondering now is if it would be better to attack the problem at the source, and take care of this in the linear stage with more careful DBE.  Any tips on how to work with DBE on an image like this that is dominated by signal with little background? 

[oldwexi]

Hi Joel!
Agree that the linear stage generally should experience a better DBE. However,
with such images full of nebula  it is very  difficult. In such cases i end up with trial and error and trial.

To use ABE on the stretched image was only an idea, based that the "fog" on the left side was
clearly visible und ABE created a nice large scale background image which was perfect for
use as a mask in the following step.

Gerald