LR4 & HDR

[jrp]

Interesting observation here: http://hdrlabs.com/news/index.php?id=1276328416487110799

" The RAW headroom data is automatically squeezed out, but from every image. Even those that barely have any. That waters down the resulting HDR, and all highlight details are mushed together."

The recommendation is to set Exporure to -1 and contrast to -40.

(This goes to the discussion about neutral/0 settings, which is ruffling some feathers.)

[Schewe]

The recommendation is to set Exporure to -1 and contrast to -40.

I think this is an edge case (and prolly dependent upon the sensor) and how certain HDR software has been designed. I think this has zero to do with single, non-HDR images. And since I've not tested this out, I really can't say whether or not the author is right or wrong. But in the case of the building shot that was shown as an example, I seriously question whether that images needed HDR in LR4. The scene contrast range sure doesn't look like HDR is needed...

[kencameron]

I think this is an edge case (and prolly dependent upon the sensor) and how certain HDR software has been designed. I think this has zero to do with single, non-HDR images. And since I've not tested this out, I really can't say whether or not the author is right or wrong. But in the case of the building shot that was shown as an example, I seriously question whether that images needed HDR in LR4. The scene contrast range sure doesn't look like HDR is needed...

The author explicitly states that his point has to do with how HDR software is designed. I agree that the building shot doesn't obviously require HDR, but his recommendation does seem to produce a better result when HDR is applied to it and he says he has tried it out on other image with the same results.  He is someone with credibility in the HDR world so I will be doing some experiments with my preferred HDR software (Oloneo) and reporting back on the results. It would be interesting if other HDR-users did the same (if there are any on LuLa).

BTW, the author says he loves LR4 and that to try it will be to buy it.

[jrsforums]

From my query to HDRSOFT...

Highlight recovery in itself is a good thing for HDR, as extracting more highlights from the original Raw files extends the dynamic range.

However, it is not a good thing for the HDR merge if the highlight recovery comes together with a compression of the highlights in a film-like roll off manner or to make the highlights better show on a low dynamic range image.  In that case, the 32-bit HDR image (i.e. the merged image before tone mapping is applied) will lack accuracy.  While this may not be too much an issue for tone mapping, this would definitely be an issue if you need to use the 32-bit HDR file for image based lighting rendering in 3D or visual effects programs.

So, in the same spirit as the advices in the "Important note" paragraph of the FAQ <http://www.hdrsoft.com/support/faq_photomatix.html#process_raw>, I would recommend to add the "Highlights", "Shadows" and "Whites" settings of Lightroom 4 to the list of setting that should be set to zero when the files are to be merged to HDR for tone mapping or for using the 32-bit HDR file itself.

However, when you are only interested in using Photomatix to process the files through Exposure Fusion, then the above advices to set to zero do not apply, and you will likely get more pleasant fused results by enhancing the source images in Lightroom prior to processing in Photomatix.

Sincerely,

Geraldine
Photomatix Engineering Team

In this case, unfortunately, as we know, the film-like roll off is automatic.

I am not saying there is a problem, just suspect there might be and it should be looked at.

[Blochi]

Hi,

I'm Christian Bloch and wrote the article you're talking about.

I completely agree with Schewe. It is an edge case and has nothing to do with non-HDR workflows. In fact, the new processing engine & defaults are miles ahead of LR3 for single RAW development. Not sure if that came out clear in the post.

The comparison image is actually just a small crop of the problem area. It would be a terribly framed photograph otherwise. It also includes a dark gate/tunnel (which you may spot on the thumbnail strip of the source images), so it actually is in desperate need for HDR methods. But that's actually not the point.

Don't take my word for it, try it for yourself on your next bracketing sequence.

Best,
Christian Bloch

[Schewe]

I'm Christian Bloch and wrote the article you're talking about.

I completely agree with Schewe. It is an edge case and has nothing to do with non-HDR workflows. In fact, the new processing engine & defaults are miles ahead of LR3 for single RAW development. Not sure if that came out clear in the post.

Christian, welcome to LuLa...thanks for the confirmation...I kinda thought so (as I indicated above).

It would be interesting to find out what in the LR4 pipeline is causing issues when doing HDR compositing...is it the highlight reduction (detail increasing/highlight recovery) or is it the masking that's being done?

Not sure anything could be done for the LR4 release (LR 4.0) but there may be adjustments in the future...

[Blochi]

Well, the root of the problem is that the HDR merger can't properly linearize the output from Lightroom. That's always the first step necessary to puzzle the luminance ranges of multiple exposures together to a 32-bit HDR file. And that doesn't work when each images gets optimized by itself.

I don't even think there would be anything in need for a change. They are simply opposite approaches. For RAW images you want it all optimized and enhanced. For HDR you want a predictable and uniform tonal distribution, because these are just intermediate files for the HDR merging.

It's kind of like in panorama stitching. You wouldn't crop each segment just so it has a nice framing by itself. It doesn't matter. You'd stitch the panorama first and then crop it nicely. Extending dynamic range with HDR is the same, just in the luminance dimension.

A nice "flat conversion" development preset would be nice to ship with lightroom. That's all the change it needs. In the long run, it would of course be great if it could read and understand OpenEXR files, could be treated as alternative to DNG.

[BlackSmith]

A simple solution, if the new process version doesn't conform to a specific work flow, is to simply revert back to the LR3 process version. Using the old process versions doesn't need to been seen as "only there for backward compatibility". If the LR3 process version has advantages for you, use it.

[kencameron]

A simple solution, if the new process version doesn't conform to a specific work flow, is to simply revert back to the LR3 process version.

+1 to this. I am finding that as an alternative to the HDR-specific develop settings in PV2012 recommended above by Christian Bloch, switching back to PV2010 works well. LR isn't doing quite such a good job with the highlights but the results work better for HDR. And if you prefer the PV2012 controls, when you bring the HDR image back into Lightroom you will find that LR has defaulted back to PV2012.

[AveryRagan]

Hi,

 In fact, the new processing engine & defaults are miles ahead of LR3 for single RAW development. Not sure if that came out clear in the post.

For me your comment about the new processing engine may be the understatement of the year...so far. I have an of imags of the French Quarter that I have not been capable of making look natural using single image HDR or LR3. With LR4 my problems seem to have disappeared. This image alone made the upgrade cheap. I thought at first that I wouldn't be able to live without Recovery but now that I am getting the hang of Highlights and Shadows I don't think I will miss it at all. Kudos to the Adobe Lightroom team.

[AveryRagan]

But in the case of the building shot that was shown as an example.

I personally think that this is a much easier image than mine to manage based on the what I see in the image. I personally think, and I don't claim to be an expert, that you wouldn't have a problem handling this image in LR4.

[madmanchan]

If you wish to generate scene-referred linear files so that other software can do appropriate HDR merges, I would suggest using PV 2010.

[Schewe]

If you wish to generate scene-referred linear files so that other software can do appropriate HDR merges, I would suggest using PV 2010.

Or using DNG Profile Editor to come up with a linear contrast curve for use in PV 2012. See this post...

The only thing I would add is to become really familiar with what you can get out of PV 2012 with regards to the tone curve. Many images that may have required HDR in the past may now longer need that...YMMV.

[kencameron]

The only thing I would add is to become really familiar with what you can get out of PV 2012 with regards to the tone curve. Many images that may have required HDR in the past may now longer need that...YMMV.

The other benefit is that you don't get the loss of sharpness as a result of small-scale misalignment that is pretty much unavoidable with multi-shot HDR unless you are photographing immobile objects and using a large tripod.

[dreed]

Having become aware of how LR now maps tone images, I have revisited some of my images that have blown highlights.

In some cases, it would appear that the new behaviour is less useful.

For example, in one image with 2010, a large section of the sky was a blown highlight (i.e. red.)

With 2012, a large section of the area that was red is now just an off-white colour.

In this scenario I'm not sure that the new behaviour is actually better - if anything, it is actually deceiving.

That said, in other images the new behaviour is better. Probably there are more of these images than those where it is creating misleading results.

The attachments show import with 2010, converting that to 2012 and then applying an auto tone to 2012.

I've also added 2010+auto.

What is really striking here is that 2012+auto has thrown a lot of shadow highlights around the tree on the right of the image away and whilst it has achieved less blown highlights, the end result is of questionable value and possibly worse (the really sudden dithering from light blue into off white.)

To be fair, this isn't an image that I marked as a "keeper", it's marked for garbage collection and just hasn't been thrown out yet. However it does serve to highlight the changes in tone mapping from 2010 to 2012 with respect to highlights and shadows.

[kencameron]

Having become aware of how LR now maps tone images, I have revisited some of my images that have blown highlights.

Interesting comparison. To me your 2012 with auto-tone looks, overall,  a little better in the sky and a little worse in the shadows, but surely the test wouldn't be what auto-tone gives you but rather the best you can do with your own adjustments. I have usually been happy with the results of revisiting old images but have noticed something like the "sudden dithering" problem that you mention.

[BobFisher]

Or using DNG Profile Editor to come up with a linear contrast curve for use in PV 2012. See this post...

Is this really any different from PV 2010?  In PV 2010, if you zero'd out all the sliders and set the Curve to Linear, you still had the tone curve embedded in the camera profile that got applied.  PV 2012 is no different except that all the sliders are already zero'd out and the Curve is already set to Linear.  You still have the tone curve embedded in the camera profile being applied.  It seems, then, that what might have changed is that the tone curve in the camera profiles may be different.  Maybe?

I've also done some experimentation with HDR and both the 2010 and 2012 PV.  There are subtle differences.  With the 2010, I had a Develop Preset in LR called No Adj that zero'd out the sliders and set the Curve to linear so from that aspect both were starting at the same point.  The difference then should come down to the tone curve in the camera profile (same profile name used in all comparisons).

[elied]

With the 2010, I had a Develop Preset in LR called No Adj that zero'd out the sliders and set the Curve to linear so from that aspect both were starting at the same point.  The difference then should come down to the tone curve in the camera profile (same profile name used in all comparisons).

Unless, of course, zero means something entirely different in PV 2012.