Processing HDR fused image

Hi guys! this seems to be a lovely place to arrange the discussion.
Since there’s not much on the web about fused HDR processing with Darktable, we might have it comprehensively discussed here :slight_smile: )

So I have an image which I got by fusing three shots of a landscape at different exposure (-1, 0, +1). For that I used DT’s HDR function which provides fusion of RAW files (unlike the conventional HDR masks blending or .tiff output fusion). I’m pretty happy with the dynamic range I got after that. Then, after adjusting the exposure and white-balance controls, I have obtained acceptable average lightness and colors. I believe, those are basic steps to be done for the HDR to average lightness and whitebalance of the fused images.
Then I’m completely lost with my following step.
Once all of the above was done, I started tweaking the tonal group to bring out the contrast, since, after the fusion, the image lacks contrast and has washed-out look. Though, nothing helped. Each tool could not provide any acceptable result - it increased the contrast while lost a lot of original colors. For each tool I referenced the manual, but with no significant success. Also I could not find specific advice for HDR development. The Tone Mapping provided the best contrast, but, unfortunately, it sacrificed a lot of colors initially present in the HDR. The image came out contrasty, but dropped a lot of saturation and looked grey-ish.
It looks there is a large potential in this image with its high dynamic range, but I can’t get it further processed correctly. Did anyone experience the same problems with the contrast?

Here’s what I got after basic processing steps http://i.imgur.com/vwhW9S0.png
Any advice how to get rid of that greyish look while retaining the colors?

1 Like

Do you fancy posting the darktable-generated HDR so we can have a play?

yep, just in a while

Cool. Oh, and welcome, by the way. :slight_smile:

Hi, here’s the link Dropbox - File Deleted
about 40 Mb
Hence I use custom camera Basic Curve for my Nikon J1 and calibrated Input Profile, so the colors on your desktop will differ from those on mine. though, it won’t make a big difference and hopefully you will be able to reproduce the effect. For the basic curve the closest is the Neutral and for the Input Profile you might use the “standart color matrix”.

So, basically it’s all about haze :slight_smile: immortal enemy of all photographers, but this time it’s a bit more sophisticated :slight_smile: I’ve got to figure out how to treat it properly with the Lab curve of Darktable (because it’s easier, personally for me, to use RGB curve and there are tons of documentation on RGB curve) and figure out comprehensive algorythm for handling HDR fused stacks that might be useful for all of us aspiring amateur photographers
you might say use GIMP, but in that case the file would have to be converted into TIFF format thus losing capabilities of RAW DNG
Why HDR - because fused HDR is a poor man’s technique to boost the dynamic range

Hi @mosaster could you also share the source raw files?

Sure,

Maybe Process HDR images using darktable. | darktable contains some hints?

Thanks Houz for referencing that. I read this article, and it is really helpful for processing HDR EXR files. If I’m not mistaken, EXRs contain masks, unlike exposure fused DNG file. However, of course, regardless of the differences, after I failed to save the colors while bringing up the contrast effcetively with the Tone Curve, my initial guess was to try the Global Tonemap (genuinely HDR tool), but it produced poor results taking all the histogram chart to the right, with the adjustments having no significant effect. Then I tried the Tone Mapping module which gave, to my taste, better results, thouigh still it failed to keep the colors.
I might be doing something wrong or lack knowledge of proper modules application, but, while there is a vacuum in this topic, please forgive any of my faults should you find them.

Also, thanks for the DT Blog updates, there is also a good article on using the Lab curves Mastering color with Lab tone curves | darktable
that brings a light onto adjustments within the Lab colorspace

Apples and oranges. Conventional HDR does what DT’s HDR does, it just does it using demosaiced source images.

The lovely photo shows a low dynamic range scene. The histogram proves it. One gains nothing by putting it in a high dynamic range container. The dynamic range you got out is the same one you put in. With such a low dynamic range scene, you might as well save yourself time by using only the brightest of the three images.

One is made from the “HDR”, the other is made from only the brightest of the three:

The only thing making a HDR in this case accomplishes is that it kills the horse who wouldn’t stand still.

1 Like

Nice, can I see your modules history for the DNG?

I tweaked around with the masks and reduced the haze to more pleasing level while retaining the colors
Here’s how I did it http://i.imgur.com/jYdQIS3.png

So, what workflow procedure would you recommend? It is more interesting to figure out right sequence of modules than just getting this very image processed. I used this sample image just as an example of contrast and color issues one might face working with DT modules

And, of course, thanks for the advice on HDR procesing essentials :wink: )

Processing photos costs me enormous mental effort because I’m faced with countless possibilities and I’m incapable of choosing just one :grin: :joy: :sob: :head_bandage:
I love it, but it’s also exhausting.

I download files straight to /tmp (RAM), so when I closed my RT laptop I also lost the download and the sidecar file. When I’m back on that laptop I’ll see if it’s in the cache; if it is then I can share it with you. The absolutely only parameter difference between the “HDR” and the normal image is that I had to adjust the exposure in one to match the other.

The haze. I used to fight such things, now I just go with the flow. Maybe one could get rid of it in post-proc (Retinex in RawTherapee and luminosity masks in GIMP come to mind), maybe it could be even done to look realistic, but it was there, it gives the photo its look, so embrace it.
…Though now I am tempted to try removing it, just for kicks :slight_smile:

Dealing with real high dynamic range scenes is not simple. Endless possibilities. Sometimes curves alone work great, sometimes they’re not enough. Sometimes tone-mapping works, sometimes not. Sometimes you have to dodge and burn and use luminosity masks. There is no one-size-fits-all. Maybe with another 30 years of experience I’ll be able to offer some advice, but not now.

Well, yes and no. EXR files can contain a lot of things, like masks, depth data, other color channels, reflectance data, … However, most often EXR is just used as a HDR capable container when used for stacked image sequences so none of those extra features are used.

Sure, combating haze is very selective thing to do in photography. I bet the haze would can never be eliminated with a tool that operates in one click. What I see from posts of commercial software users about about recent De-Haze plugins, they are not happy as the tool either overkills or underperforms. So, as one can see, it is unworthy to evaluate all cases algorithm.
Likewise, solely the Tonal Mapping does not do the job. It kills colors while brings the contrast up.
For HDR stacks development and mostly artifacts that, usually, follow them, such as low contrast in shadows or highlights due to wrong exposure or faulty exposure selection, I believe it is viable to find a solution. It would be great to evaluate a sequence of actions to take control of trimming down the unwanted tonal range without crushing the colors.

Here’s what I managed to attain with the Levels+Masks VS the best I could with the Tone Mapping.


The left bright side is the Levels+Masks adjustment VS the right dull and washed out Tone Mapping result. Look at the colors, can you see the difference ? Wouldn’t be great to figure out a better workflow ? :smile: )

P.S. Hence with the Levels+Masks way I yet made no contrast tweaks, just slided the margins to the histogram.

That’s what I ultimately got following this peculiar way :slight_smile:

2 Likes

Sorry to resurrect an old thread but I’ve just found the Haze module in my darktable 2.3.0 ~git647, not sure how long its been there! Just been playing around with it at normal settings. Its quite impressive, especially on reflections and sky. Going to look for some landscape shots to try it out (I don’t normally shoot landscapes). Well done darktable!