HDR Tame: Default Value 'Gamma Highlights (Tame)'

[decay]

Hi Ivo & all,

for quite some time I have been thinking about the default value 1.25 of the 'Gamma Highlights (Tame)' parameter of the HDR module.

I have the impression that this default setting might be too aggressive and might lead to unexpected results - at least for some objects. Of course, if you know what's going on, it's just a simple click to reduce the value, but for new users with less experience, it might take some time to figure it out. This was at least the case for me

Best regards, Dietmar.

m27_hdr_tame_ani.gif (86.03 KiB) Viewed 5516 times

m31_hdr-donut.gif (192.8 KiB) Viewed 5516 times

[decay]

I have tried a little more and it seems that this somehow goes together with context size. I tried context size 5x5 (instead default 50x50) and then it's even stronger (please see below).

Up to now I thought that small context sizes could be used to work out small structures. But here we can see a huge impact on a very large area? Maybe 'Gamma Highlights' is somehow independent of the context size?

Possibly there is still something I do not understand? Really, I'm not sure about all this. If I am the only one struggling with this we should maybe forget about it ...

Thanks & best regards, Dietmar.

m31_2_hdr.gif (246.09 KiB) Viewed 5457 times

[hixx]

Hi Dietmar,
no You're surely not the only one....
I'm sure it's all good math performed on the data, but to connect that with visible results is the challenge we are facing
I am also reducing this default to below gamma 1.1 (usually between 1.03 and 1.07). But please note, those effects like demo'ed by You below are just visible in direct comparision (Before/After & Pre/Post Tweak). Looking at the gamma 1-25 pic as such it does not look flawed
Also I found the Gamma Smoothen parameter extremely helpful to cover the "slope" between Gamma processed and linear sections. This gives me more headroom for increasing gamma values without making these visible.
Furthermore I use to reduce think the detail enhancement to 20% or 25% to avoid over processed look.
I rarely use context sizes below 30 px - 5 pixel's context size will work on noise or stars at best. There should be some "Context" HDR can hold onto I think...
Clear Skies,
Jochen

[decay]

Thanks for your reply, Jochen

Yes, I read your and Guy's addendums and usernotes (thank you both!) and most times I get reasonable results after some tries. What bothers me a bit that this is still some kind of trial and error processing for me, despite to the fact, that I thought I would understand how these sliders work. But especially with context size setting and it's consequences for the other sliders I'm not sure any more. Maybe this is something that might be worth to overthink for one of the later following versions of ST ...

What I had in mind, is that one of Ivo's ideas for ST is the ease of use with well-performing default workflows and settings. (And for sure, it's already pretty good!) But I think, this 'Gamma Highlights' default setting often leads to unexpected results - and is not only visible with direct comparison (Before/After & Pre/Post Tweak).

So my intension was, if this could be mitigated (at least a bit) by changing this default value, this is something what could be included in upcoming version 1.9.

Best regards, Dietmar.

[decay]

But please note, those effects like demo'ed by You below are just visible in direct comparision (Before/After & Pre/Post Tweak). Looking at the gamma 1-25 pic as such it does not look flawed

and is not only visible with direct comparison

Sorry, Jochen ! Of course, this is your impression/opinion/meaning and therefore true by definition! I just wanted to say, that this is often not the case for me. And as said, I may depend on the concrete object and dataset.

(I remember for example, that Martin lately wrote, HDR shows an impact like shattered glass on his datasets and therefore I assumed, this could be a more common problem. But of course, I don't know ...)

Best regards, Dietmar.

[hixx]

Hi Dietmar,
I could reproduce both Tame and shattered glass effects. I agree some HDR results are a bit personal taste, it is so powerful it's easy to overprocess.
I assume HDR will perform very differently, depending on focal length / pixel pitch ratio and/or subject and also on global stretch prior HDR

The default settings that work best for my gear (F=553mm, DSLR w / 5,9mµ pixels) are: Context size: 30-50 px, Gamma smoothen 30-40 px, Highlight Gamma: 1.03-1.06, Shadow Gamma:1.05 -1.07). I guess these will be different using 8" to 16" Newtonians or even Hubble images. Let alone using Astro Cams with larger or smaller pixel pitch.

You are right, maybe the way to go is using various default settings - COMPOSE now has SCENE modes which will chose default settings in various modules. A potential way forward might be to expand on that and provide ST more information on capturing equipment so more default set accordingly - Until then it'll be a bit trial and error to understand which parameters work, which is the actual fun part to me
Clear Skies,
Jochen

[Startrek]

I may have already mentioned HDR 1.8 previously but you certainly have to tread lightly with it especially using Reveal and Tame. I too, found defaults way too aggressive and only use it sparingly otherwise your mid to large structures become pronounced to the point of looking like shattered glass
I believe DSO core is being re introduced into V1.9 which really worked well in 1.7
In fact I still prefer HDR 1.7 over HDR 1.8 to a certain degree

Cheers
Martin