Big Pink Stars after ArcsinhStretch

[John_Gill]

Hi,

I have recently tried the ArcsinhStretch and have had some success.  I am now re-processing some DSLR images from last year (due to clouds and rain).  Anyway, I do LocalNormalization, DynamicCrop, PhotometricColorCalibration, Automatic or DynamicBackground Extraction.  I then run Extract CIE L* Component, Stretch and apply as a mask and Invert the mask. Then MultiscaleLinearTransform for noise reduction.  After I apply ArcsinhStretch, the big and bright white stars have big pink cores.  But if I stretch the image with Histogram Transformation and ScreenTransferFunction then no pink stars ....

Any ideas or suggestions cos I am tired of seeing pink stars ....
John

[John_Gill]

Hi,

I recently subscribed to www.ip4ap.com to improve my processing skills and did a search on the web site for a tutorial on ArcsinhStretch.  Warren Keller has already done a tutorial on how to fix the pink star issue.  The issue is caused by the the over protection of highlights during the MLT noise reduction.  Anyway the solution is as follows:

Load the problem image
From the Scrips menu select Utilities and then Repaired HSV Separation
Select V - No repairs
Adjust repair level (my repair worked on 0.1)
Adjust the Max Repair Radius (I used 24) and OK
This generated 4 output files.

Now select Process ---> ChannelCombination
Select HSV Color Space
In the Chanels / Source Images select the H, Sv and V files
(Note that the V file is the Unrepaired_V file)
Apply this to the problem image file
Now run ArcsinhStretch
and no more pink stars

If there are still stars with pink centers, then create a mask for the large stars
Invert the mask and then Convolution will fix the last of the pink stars

Looking up with no pink stars ...
John

[pfile]

i'm not entirely sure that the problem is caused by MLT, i think it might be an artifact related to 14-bit cameras. you see this all the time in DSLR images of M42, where the core comes out pink. i think it could be because the image is saturated in the red channel only, and then when stretched it comes out pink since in 16-bit space the saturated red channel is not actually saturated. it's as though it's an image of something strongly colored red which becomes washed out to pink as it is stretched.

it's possible using rescale on each channel before color calibration would help with this problem, i am not sure.

rob

[John_Gill]

Thanks Rob, I will try what you are suggesting.  The image in question is of the Prawn Nebula, IC 4628, which has a large amount of red nebula.  Thinking about it now, when using ArcsinhStretch on other images, there has only been a little bit of red, mostly blue, so maybe you are right.  I will have another fiddle and see what happens.

Look up
John

[sharkmelley]

The pink star cores are easily explained - it's caused by the bright stars having saturated cores in the raw data.  This means the R,G,B channels have identical maximum values.  When white balance is applied, the green channel must be scaled down (or the R&B channels scaled up).  The resulting colour is pink.  This pink colour is preserved by ArcsinhStretch because ArcsinhStretch deliberately preserves colour ratios in every image pixel.

I use the HistogramTransformation and adjust the highlights slider down until the saturated cores have equal levels of R,G,B once again and are therefore white.  Do this at some stage in your workflow before using ArcsinhStretch.

Mark

[pfile]

ok, that makes sense. i was going off what happened to me while processing DSLR eclipse images - the red channel was totally blown out while the blue and green were not, leading to some funny stuff happening.

rob