Cepheus to Cygnus, 57 panel mosaic.

[symmetal]

7 hours ago, ollypenrice said:

Interesting question. The stars have been given an absolutely bog-standard log stretch using the mid point slider in Ps Levels, so the size/brightness range should be perfectly normal and typical of most astrophotos. (Indeed, since there has been no star reduction*, it should be a classic AP stretch.)  However, the stretch itself was much lighter than that given to the background. If  you use one stretch for all you end up with this:

Now some people might like or prefer this and, as an image demonstrating the richness of the MW starfield, it's fine. It's just not what our image is about. Modern processing allows a telescopic look to be extracted from a lens image, meaning smaller stars and more visible nebulosity.

A consequence of the separate stellar stretch is that small/faint stars will not reach the level of the widespread nebulosity and will remain invisible, so diminishing the range of the stars we can see. This is exaggerated when faint nebulosity is lifted well clear of the background sky but, again, that is the whole point of an image like this. Perhaps this accounts for what you are seeing?

Olly

*By star reduction I mean the reverse-processing of a star of a given size to make it smaller. Our stars have simply never been stretched far enough in the first place to need it.

 

Thanks Olly. It's possible that SXT is not preserving all the individual stars when the density is very high and may be combining or omitting stars when creating the star layer. In the 'normal' stretch above the stars above Deneb are much denser than those in the dark patch below it but in your separately star stretched composite image the density is very similar. it would be interesting to see what it looks like if you gave the separated stars image the same stretch as the background and added it back in. Does it look like the image above? 🤔

Regarding the background stars all having a similar brightness, at short focal lengths several faint stars will fall on the same pixel and register as one brighter star Certain image scales may show this effect more than others.

When rescaling to 80% the algorithm used will affect the star shape/size too along with jpeg compression of course. 

Alan

 

[ollypenrice]

On 29/09/2023 at 16:55, symmetal said:

Thanks Olly. It's possible that SXT is not preserving all the individual stars when the density is very high and may be combining or omitting stars when creating the star layer. In the 'normal' stretch above the stars above Deneb are much denser than those in the dark patch below it but in your separately star stretched composite image the density is very similar. it would be interesting to see what it looks like if you gave the separated stars image the same stretch as the background and added it back in. Does it look like the image above? 🤔

Regarding the background stars all having a similar brightness, at short focal lengths several faint stars will fall on the same pixel and register as one brighter star Certain image scales may show this effect more than others.

When rescaling to 80% the algorithm used will affect the star shape/size too along with jpeg compression of course. 

Alan

 

Good points. However, I wouldn't say that SXT has any role in preserving or discarding stars. In order to create the starless layer it obviously removes them. However, when they are replaced, they are replaced by stars from the original image, untouched by SXT. There are different ways of doing this but the simplest (which I no longer use) just involves putting them as a top layer in blend mode lighten. If Photoshop is doing what it says it is, these stars will come into view as the stretch renders them brighter than the background. This means that the imager is likely to halt the stretch before the faintest stars have become visible. I'll see if this method gives a greater range of stellar brightness.

One thing I have done is experiment with ways of varying the star stretch according to the brightness of the background around them. While this is not true to the data, it may be more true to the visual impression since stars of a given brightness otherwise look way brighter against a dark than a light background. I didn't do that on this image.

Olly

[symmetal]

6 hours ago, ollypenrice said:

However, when they are replaced, they are replaced by stars from the original image, untouched by SXT

I thought you were using the SXT separate star layer to add back in but that's not the case. Maybe using curves to stretch the stars rather than the levels centre slider you can get more control on what stars are brightened and end up with more variation.

Alan

[powerlord]

9 hours ago, ollypenrice said:

Good points. However, I wouldn't say that SXT has any role in preserving or discarding stars. In order to create the starless layer it obviously removes them. However, when they are replaced, they are replaced by stars from the original image, untouched by SXT. There are different ways of doing this but the simplest (which I no longer use) just involves putting them as a top layer in blend mode lighten. If Photoshop is doing what it says it is, these stars will come into view as the stretch renders them brighter than the background. This means that the imager is likely to halt the stretch before the faintest stars have become visible. I'll see if this method gives a greater range of stellar brightness.

One thing I have done is experiment with ways of varying the star stretch according to the brightness of the background around them. While this is not true to the data, it may be more true to the visual impression since stars of a given brightness otherwise look way brighter against a dark than a light background. I didn't do that on this image.

Olly

Don't use ps, but in affinity photo it would be an Add blend layer. The opposite of how you get the star layer in the first place (by flipping the starless layer to subtract, set above the original image and merging the visible layer into a new 'stars' layer). Screen can give a milder result, but Add is the 'as it started' way.

What I tend to do is start with either it set to Add or Screen as I feel best suits, then apply curves, levels and saturation to that layer as appropriate. I usually do end up 'normalising' a little by reducing the bright stars.

What's the equivalent process in Ps ? Or I suppose rather, what do you do in Ps ?

Oh, and how long did starx take on the whole mosaic ?? How many mp is it in ps ? Must be Giga pixels ?

Stu

[ollypenrice]

25 minutes ago, powerlord said:

Don't use ps, but in affinity photo it would be an Add blend layer. The opposite of how you get the star layer in the first place (by flipping the starless layer to subtract, set above the original image and merging the visible layer into a new 'stars' layer). Screen can give a milder result, but Add is the 'as it started' way.

What I tend to do is start with either it set to Add or Screen as I feel best suits, then apply curves, levels and saturation to that layer as appropriate. I usually do end up 'normalising' a little by reducing the bright stars.

What's the equivalent process in Ps ? Or I suppose rather, what do you do in Ps ?

Oh, and how long did starx take on the whole mosaic ?? How many mp is it in ps ? Must be Giga pixels ?

Stu

I use a mathematically more complex system in Ps but I don't do the maths myself! Essentially it's invert and divide.

1 Log stretch the image to about 80% of full stretch and set the black point not too dark. Save as Stretch1.

2 Run StarX. Save as Starless. Process Starless as you see fit. I use Noise Xterminator as a bottom layer and erase the bright, sharp bits which don’t need it. I also fix the background and stretch a little more above that, using Curves. Do all contrast enhancement and sharpening to the starless image. Save.

3 Paste Starless over Stretch I. From here on I have made an Action:

4 Ctrl I to invert both layers.

5 Top layer active, set blend mode to Divide.

6 Stamp down. (Alt Ctrl E) This gives a new top layer.

7 Ctrl I to invert that layer.

8 Flatten image. (Do this under the layers palette from the top toolbar. Ctrl E does not work for me.)

9 Save as Stars. End action.

10 Paste Stars onto the processed starless, blend mode screen.

11 Use Levels mid-point slider to reduce stars.  Small stars benefit from contrast reduction. Large soft stars benefit from contrast increase.

We were working with an image of 552MB after stacking in super-pixel at 80%. This is much smaller than huge telescopic mosaics in which I've participated.

StarX worked reasonably quickly. I wandered off and came back, but 20 minutes would cover it, I'm sure. No big deal.

Olly

[Clarkey]

That is a truly outstanding image.

I was trying to count up how long it would take me to complete a 57 panel mosaic in Cumbria. I ran out of fingers🤣

[JonHigh]

Just came across this post and wow, that is awesome and inspirational! Just puts things into perspective!  A massive well done 👍🏻. I’m now looking up at cloudy UK skies yet again.