2023-01-19: APP 2.0.0-beta13 has been released !
!!! Big performance increase due to optimizations in integration !!!
and upgraded development platform to GraalVM 22.3 based on openJDK19
We are very close now to releasing APP 2.0.0 stable with a complete printable manual...
How to get true colors for bright stars?
I am new to APP - I just downloaded it. I am mostly a landscape astrophotographer. I mostly use Canon DSLR, including a full spectrum modified 5D Mark IV, but I also recently acquired a QHY 128c chilled color camera.
I would like to make wide angle starscapes that are close to "true color" for the stars, even the bright ones.
Some stars deserve to be white. PixInsight has a photometric calibration feature where it finds stars or galaxies that are within your frame, and which are close in spectrum to our sun and thus serve as a white reference.
Is there a feature like that in APP?
Many bright stars have strong colors. However, all of the bright stars in my photos are massively overexposed. It's easy to see why - a DSLR and decent aperture lens with stacked exposures can record stars down to 11th or 12th magnitude, but that is far wider dynamic range than the 14 bits of range that the sensor records.
In normal photography terms, stars are intrinsically a high dynamic range subject. So, I have tried HDR techniques but there are several problems:
An exposure short enough to capture very bright stars without saturation has few stars in the photo so it can be hard to align. But worse, when you do align it you find that the small, correctly exposed star must fill a huge gap. That is because over exposed stars create a spot much larger than unsaturated stars. That's because the point spread function for the stars drops off with distance, but intrinsically has much larger radius as the exposure goes up.
So, even if you correctly align the exposures, you have a large diameter saturated hole to fill. A conventional HDR merge leaves the overexposed ring and halo around the smaller, correctly exposed core.
One could imagine techniques to get around this, but so far I have given up on trying to make all stars the same size. As it stands, we have all gotten used to astrophotos with bright stars that are huge - a bit like the convention on maps where big cities are shown with a dot size proportional to their population.
But I still want the stars - whatever their size - to be the correct color.
A separate short exposure is one way to sample the correct color, but there might be another way.
The slopes or wings of the point spread function often record the color of the star as a fringe. The core is saturated in one or more RGB channels, but as one moves radially away from the core you typically reach an area where the pixels are not saturated. This raises the possibility that one could do a single-exposure HDR, where one reconstructs the color from the fringe
It could be tricky because the fringe or edge might interact poorly with Bayer array.
So this leaves two ways to make true-color bright stars:
1. Use short exposures to capture the color, then use that to color the over-exposed star image. This is using a capture at different exposures, like HDR, but it really isn't HDR.
2. Use the unsaturated fringe around bright stars to sample their true color, then propagate that to the saturated portions.
I am interested in whether anybody has experience doing either one of these things in APP ? Or a different way to solve the problem?
HDR is not possible at the moment in APP, but the star color correction tool is (in my humble opinion) a bit more reliable than in PI. It actually takes a population of stars and calibrates them according to a black-body model. This is a good way of doing that as we know that the main sequence stars are basically “black bodies”, so they have a predictable radiation on which we model the color.
OK, but I would still need to do something to take the residual star color from the fringes of overexposed stars and use that to color the star.
Yes, that would probably require HDR processing like you mentioned. This on our to-do list, but not yet possible.
What about recovering the star color from the fringe? That does not require HDR - it only requires some fancy masking and manipulations - I think.
Maybe, but to implement something like that very nicely is tricky I think. Something like masking and other features will be in the future of APP, but we're not there yet unfortunately. The basics are now growing up and from there we will make step-wise progress. We do intend to have these tools to be best in class, just like the calibration which is unique and the mosaic feature which is still unique in its ease of use and result.