HDR processing - total integration time for shorter subs

[cuivenion]

I'm wanting to try out HDR processing to get the best out of stars and galaxy cores etc. Should total integration time for the shorter subs equal that of the longer ones? Are there any good guides out there for HDR imaging?

[vlaiv]

Stacking images is in fact HDR imaging even if single exposure duration is uses.

But above observation does not answer your question.

We are satisfied with image once we have high enough SNR. For bright stars and bright central core of galaxy, single short exposure will result in rather high base SNR hence you need just a hand full of them to make them with enough SNR to go into the image.

One approach that will work good - take a few short exposures (like 10-15 seconds is enough) and stack those using average method (no fancy stacking needed there). Scale pixel values to match those from long exposure (multiply with ratio of exposures - if base exposure is 5 minutes and short exposure is 10s - multiply with 10s/300s = 1/30 - or divide with 30 same thing).

Now that you have matching photon flux it is the simple matter of blending in two stacks and you can do it with pixel math (or similar tools):

for pixel math it would be something like : resulting pixel value is equal to long stack value if value is below threshold (set it to something like 90% of brightest pixel in long stack) - else use pixel value from short stack.

for other kind of blending - make selection of all pixels in long stack that have higher value than 90% of highest value in the image. Copy pixels using same selection (long stack and short stack need to be registered / aligned) from short stack and paste them onto that selection in long stack.

[cuivenion]

Thanks Vlaiv. I was going to use the HDR combination tool in Pixinsight, it's seems like that will be easier to get my head around than the pixel math method. I have the commitment but not the brains unfortunately. I'm unused to messing around with pixel values.

[vlaiv]

20 minutes ago, cuivenion said:

Thanks Vlaiv. I was going to use the HDR combination tool in Pixinsight, it's seems like that will be easier to get my head around than the pixel math method. I have the commitment but not the brains unfortunately. I'm unused to messing around with pixel values.

It looks like HDR combination tool will do exactly what I described - so you should use that, here is article about it:

https://www.lightvortexastronomy.com/tutorial-producing-an-hdr-image.html

Mind you - it advocates use of 64bit precision - but I don't think its necessary. In fact I was earlier concerned if 32bit can give enough precision when doing large number of subs stacking - but I neglected the fact that it is floating point precision, so 32bit float point is generally sufficient for all operations on the image.

If you read thru that tutorial, you will find that we agree on some points like this one:

Quote

Binarizing threshold is perhaps the most important parameter that can be tweaked. The default value of 0.8000 works very well in most cases. This parameter helps define the areas of pixels from a longer exposure image that are replaced by a shorter exposure image (in terms of their relative brightness).

It is very much equivalent to mine: replace pixels with 90% or higher value ... (I used 0.9 value in my example, but 0.8 is perfectly valid as well - will not make much difference).

[cuivenion]

Thanks Vlaiv. I just need some clear skies now.

[ollypenrice]

As Vlaiv says, there is no need to match integration time. What you're after in short subs is the right exposure for a small part of the target and you'll only be including the bright signal into the final image anyway, so noise won't be an issue.

For this reason I don't shoot luminance for the short subs either. The purpose of luminance is to get more signal and here we want less.

If you feel like a change from PI then this is a great Photoshop tutorial. http://www.astropix.com/html/j_digit/laymask.html This is how I did my own M42, using LRGB for the long subs and then just RGB for those of 50 and 11 seconds for the core.

Olly

[cuivenion]

Thanks Olly, I'll check it out. Thats an amazing M42 by the way.