Lagoon Nebula comparison

[sjastro]

While working on processing the Trifid Nebula in PixInsight I came up with a new pixel mapping function.
It is a mathematical function based on a dampened sine wave which only maps pixels in the shadow and mid tone range.

I tried it out on a luminance image of the Lagoon Nebula and the results look encouraging. Compared to an unmapped image its looks like a much deeper exposure without the star bloat and excessive noise if the image were simply over stretched.

The unmapped image is here.
http://members.iinet.net.au/~sjastro/lagoonHaNormal.jpg

The mapped image is here.
http://members.iinet.net.au/~sjastro..._Pixelmath.jpg

Regards

Steven

[multiweb (Marc)]

A very nice M8 with sharp details. Nice stretch too.

[LightningNZ (Cam)]

I think the original is far superior sorry (as Marc said, very nice M8). The core looks all burnt out to me in the pixel mapped version and you can clearly see the background noise now coming through - at least in the JPEG anyway. Perhaps you could use it at half-strength or something so it wasn't quite so pronounced?

Cheers,
Cam

[sjastro]

Originally Posted by LightningNZ

I think the original is far superior sorry (as Marc said, very nice M8). The core looks all burnt out to me in the pixel mapped version and you can clearly see the background noise now coming through - at least in the JPEG anyway. Perhaps you could use it at half-strength or something so it wasn't quite so pronounced?

Cheers,
Cam

Thanks for your feedback Cam.
Here is the application of the maths at varoius iterations.

Original image.
http://members.iinet.net.au/~sjastro...h/lagoonHa.jpg

One iteration.
http://members.iinet.net.au/~sjastro...agoonHaIt1.jpg

Two iterations.
http://members.iinet.net.au/~sjastro...agoonHaIt2.jpg

Three iterations.
http://members.iinet.net.au/~sjastro...agoonHaIt3.jpg

Regards

Steven

[White Rabbit]

Very nice, the result looks similar to the HDR multiscale Transform module in the Pixinsight.

I like the way it makes the dark nebula pop out more.

Nice work.

[LightningNZ (Cam)]

Thanks Steven. That's a really interesting progression. Is there a way to set the sine cutoff point? The core of the lagoon and two of the "walls" of gas are quite bright compared to everything else in the image (other than the stars) and I've found the burn out very easily - as is seen with this progression.

Anything to preserve the dynamic range in the top part of the histogram.

[PRejto (Peter)]

Nice image! I think 1 iteration is quite enough - and makes the image pop just the right amount without it looking overcooked.

Peter

[strongmanmike (Michael)]

They all still look like Hommer Simpson though

Great bit of tinkering Steve

Mike

[Bassnut (Fred)]

1 iteration is more than fine. After that it just blows out to hell. Of the 1st 2 pics, the original is far better, the "mapped " version albiet deeper, is awefull. Im not sure this fancy mapping gear is that flash really, is it that special?.

[sjastro]

Originally Posted by Bassnut

1 iteration is more than fine. After that it just blows out to hell. Of the 1st 2 pics, the original is far better, the "mapped " version albiet deeper, is awefull. Im not sure this fancy mapping gear is that flash really, is it that special?.

Fred it's the challenge of doing something different.
I've never heard of type of function being used in processing and there is always the scope for improvement.

Originally Posted by strongmanmike

They all still look like Hommer Simpson though

Great bit of tinkering Steve

Mike

Very imaginative Mike. Homer is very apparent.
Do you see Homer in ink blot tests.

Originally Posted by PRejto

Nice image! I think 1 iteration is quite enough - and makes the image pop just the right amount without it looking overcooked.

Peter

Thanks Peter.

Originally Posted by LightningNZ

Thanks Steven. That's a really interesting progression. Is there a way to set the sine cutoff point? The core of the lagoon and two of the "walls" of gas are quite bright compared to everything else in the image (other than the stars) and I've found the burn out very easily - as is seen with this progression.

Anything to preserve the dynamic range in the top part of the histogram.

Unfortunately working on the maths function won't improve the dynamic range as I can only use a narrow range of values by either by changing the multiplier of the angle in the sine term and/or altering the exponential term. The function needs to map each individual pixel to either zero or a positive value. If the mapped value is negative PixInsight will invert the pixel. When I was developing the function I had some bizarre cases where parts of the image were inverted.

There is a solution however. I have a developed a function where the sine term has been replaced by a -tan function. The image needs to be subjected to an even number of iterations otherwise the image is inverted. The resultant mapped image is a low contrast version that retains the information of the unmapped image.

Merging 6 iterations using this function with 3 iterations of the overcooked version with the original function, and using PixelInsights Local Histogram Equalization seems to do the trick.
The faint nebula in the background is revealed without burning out the Lagoon.

http://members.iinet.net.au/~sjastro...HaS_Tmerge.jpg

Originally Posted by White Rabbit

Very nice, the result looks similar to the HDR multiscale Transform module in the Pixinsight.

I like the way it makes the dark nebula pop out more.

Nice work.

Thanks for your comment.

A very nice M8 with sharp details. Nice stretch too

I must admit as others have mentioned Marc, the number of iterations was a bit excessive.

Clear skies

Steven

[White Rabbit]

The new one you've posted is really good.

[White Rabbit]

What was them nm of the Ha filter you used?

[sjastro]

Originally Posted by White Rabbit

What was them nm of the Ha filter you used?

I used a 7 nm Ha filter.

Regards

Steven

[sjastro]

If anyone is interested here is the summary of events.

(1) Process image normally in PixelInsight.
(2) Create a clone image.
(3) Apply 3 iterations of exp(-0.15)*sin(pi/3)x to original image.
This produces a high contrast overcooked image.
(4) Apply 6 iterations of exp(-0.15)*-tan(pi/3)x to the clone image. An even number of iterations is required otherwise the image is inverted.
This produces a very low contrast undercooked image.
(5) Merge the high and low contrast images.
(6) Apply Local Histogram Equalization (LHE) to the merged image.

The original and final image are attached.

Regards

Steven

[LightningNZ (Cam)]

Your new method is quite impressive. I think you've convinced me to buy a copy of PixelInsight!

Cheers,
Cam

[sjastro]

Originally Posted by LightningNZ

Your new method is quite impressive. I think you've convinced me to buy a copy of PixelInsight!

Cheers,
Cam

Thanks Cam.

PixInsight is good enough even without the pixel maths options.

Regards

Steven

[A23649 (Nathan)]

Steven,

What is the actual formula you insert into Pixelmath? Is it where you have an x that you include either the normal or cloned image? And how do you set up the iteration amount? I can only really invert the clone image and then bring it back to normal. Looks like a very interesting method and I wanted to give it a go but so far I have had no success with it. If you could point me in the right direction I would really appreciate it.

Cheers,
Nathan

[sjastro]

Originally Posted by A23649

Steven,

What is the actual formula you insert into Pixelmath? Is it where you have an x that you include either the normal or cloned image? And how do you set up the iteration amount? I can only really invert the clone image and then bring it back to normal. Looks like a very interesting method and I wanted to give it a go but so far I have had no success with it. If you could point me in the right direction I would really appreciate it.

Cheers,
Nathan

Hello Nathan,

I've come up with an improved formula which no longer requires the use of clone images nor two formulas.

Cut and paste the following formula into the RGB/K section of the pixel math dialogue box.

$target*exp(-0.2*$target).

$target allows you to apply pixel maths to any image opened in PixInsight.

The number of iterations is the number of times you apply the triangle icon in the dialogue box to the image.
When finished apply LHE (Local Histogram Equalization).

The following example is applying 4 iterations of the new formula followed by LHE to a previously stretched image of the Lagoon.

Regards

Steven

[LewisM]

I tried your pixel math equation in PI on my reprocessed M8, and I do like the result.

[sjastro]

Good to see it works for you Lewis.

Regards

Steven