[APh] HDR merge tone mapping

[NormanPCN]

The HDR merge works pretty well but we do not have any control over the white and blacks of the tone map/compression. The algorithm very much seems to want a certain amount of tones at the boundaries of the range. Shadow/highlight on tone map both harm contrast. Very much so with shadow. So these are not the best options to adjust these boundary tones.

 

If we run the tone map persona a second time to get "more tone compression", well that does not help because the algorithm seems to want a certain amount of tones on the boundaries. Only the stuff inside those boundary tones gets the second dose of compression.

 

I would suggest that we have some mechanism to control the algorithm needs for tones pushed to the boundaries. Something like whites and blacks controls.

 

Also It could not hurt to have the tone compression control provide a little greater range/power. A double dose of this has issues without control over the tone pushes to the boundaries by the map.

 

Some publicly available HDR bracket examples to demo my comments. Also note my relational comparison for Affinity HDR merge is against SNS-HDR and some Photomatix.

 

"chain bridge" Note how the tone map algorithm really wants to blow out the lights on top of the bridge. Even -100% on highlights is not enough. A second dose of tone "compression" just blows the lights even more.

 

"marble hall" an interior shot. The windows in the APh tone map are really bright and hard to control in relation to the darkness of the rest of the tone map. We lose contrast beyond a certain amount of highlight slider control.

 

"sunset". Here we have two items. The blown brights of the sun and the super dark linear band of plants on the left side. Raising shadows for that dark band turns everything into mush. No contrast. The personal desire on situations like this is to have detail visible but dark. The brights/sun is reasonably controllable with highlights but with some contrast issues.

 

But part of the point is, the tone map is pushing some % of tones to the boundaries and then we needs something else to attempt to compensate for the light/dark push of the tone map. It should be better to just let us control the white/black boundaries in the tone map controls and fine tune with the "standard" image editing controls.

 

Also, it is easier to us to push tones down/up, darker/lighter than the reverse. The reverse can have contrast consequences. The tone map algorithm seems to do a very good job with tone compression while preserving contrast as best as possible. We only need a touch more/better control to be golden.

 

"chain bridge" and "marble hall" HDR samples

http://farbspiel-photo.com/learn/hdr-pics-to-play-with

 

"sunset" HDR samples

http://captainkimo.com/hdr-software-review-comparison/

 

Thanks for listening

[kirkt]

Try using all of the tone mapping tools, especially the amount of compression, the exposure and most importantly for what you are complaining about, Curves.  The Curve dialog will help you move specific tonal regions around.  It would be nice if, in the Tone Mapping persona, there was a 32bit color readout to help identify where the unbounded pixel values are on the tone curve so it is easier to isolate them on the Curves adjustment.

 

I find that the black point and contrast sliders in the Tone Mapping persona are way too sensitive and crush the shadow values with very little adjustment away from zero.

 

Another approach is to do your merge and DO NOT automatically go to the Tone Mapping persona.  This way you can make adjustments to the 32bit data (with Curves or any other 32bit tool) and manually compress and remap the values before committing them to the Tone Mapping persona.

 

kirk

[kirkt]

Also note that the Chain Bridge and Marble Hall image sets contain data was artificially created in Adobe Camera Raw from source raw files - that is, the exposure for a source raw was adjusted in ACR to make a new, artificial exposure at a higher or lower EV level. I do not think this is advisable, especially if you use ACR's PV2012, where all sorts of automagic hanky-panky is going on without your ability to control it - mostly in the highlights.  This can lead to all sorts of problems in the areas where the data are not what the merger and tone mapper expect (based on the EXIF, or an estimation of the exposure and gamma curve for each source image).

 

Use your own source raw files and try using the full set of Tone Mapping persona tools.

 

That said, there are better tone mappers out there, mostly in dedicated HDR applications.  However, AP is somewhat unique in that  most of its set of editing tools is 32bit compatible, so you can manually tonemap your 32bit data in AP where other image editors might not be able to handle 32bit operations.

 

kirk

[NormanPCN]

Thanks for the reply.

 

It did not occur to me to lower the tone "compression" slider when the default full setting is overblowing the highs. Seems the word "compression" is potentially misleading. At least to me. In my mind I was thinking that less "compression" would leave more highs. Thanks for making me look at that more. Anyway, lowering that does reduce the highs getting overblown with the AP tonemap but then introduces other changes in the tonemap that need to be compensated for, IMO. 

 

I choose those examples simply because when I was looking around for HDR software I came across those. In my view they are tough test cases. So I saved them to disk for future reference. What very little I have is way too easy to be a good test to evaluate the ability of an HDR merge and tone mapper to tweak tone range. Yes some of those TIFF examples referenced use the technique of generating extra bracket photos with the raw developer by deliberate over/under exposure. Not necessarily ideal practice but it works and frankly other HDR software handles those with aplomb.

 

Anyway I have SNS-HDR. I have no need for Affinity HDR. I am just commenting about difficulties I have with the Affinity HDR offering on those tough examples. Comments they could maybe use to improve tone map/control capability in the future. With SNS I find it is easy to trivial to handle those test cases.