Ad
Ad
Ad
Pages: « 1 ... 12 13 [14] 15 16 17 »   Bottom of Page
Print
Author Topic: Do you hate HDR too?  (Read 255527 times)
joofa
Sr. Member
****
Offline Offline

Posts: 485



« Reply #260 on: September 13, 2011, 02:31:17 PM »
ReplyReply

I admit I feel most comfortable with Andrew's definition, which seems eminently practical.

Yes, I think Andrew put it well.

Then we can argue about which name best fits each person's personal vision of "HDR."  

The way I see it the issue is that the definition of HDR is nebulous and variants of HDR definitions support both Andrew's and Guillermo's version. So in different contexts both of them are right - Andrew wants to apply HDR notion to signal acquisition and Guillermo to signal presentation.

So I suggest consulting with the experts

An important aspect of an "expert" is the ability to recognize the various contexts in the which ill-defined terms are used. I don't think there is any confusion in experts' mind that what exactly is HDR despite the variant definitions. The issue becomes pronounced when one group thinks that their definition should have more weight without sufficient persuasive logic.

Sincerely,

Joofa
« Last Edit: September 13, 2011, 02:39:25 PM by joofa » Logged

Joofa
http://www.djjoofa.com
Download Photoshop and After Effects plugins
LKaven
Sr. Member
****
Offline Offline

Posts: 787


« Reply #261 on: September 13, 2011, 03:00:44 PM »
ReplyReply

32 bits is 32 bits. If a 32-bit floating point number could represent more than 32 bits worth of information, it would be an information-theoretical break-through.

-h
But the degree of precision is not the same throughout the range of quantities represented.  You have to see here that the "lower" tones of an LDR capture are not of sufficient precision to support the variable blackpoint.  You can't amplify 3 bit quantities and expect to see subtle tonal variations.  But if you have bits allocated where you need them, as in a floating point representation, you have some resolution that is more or less good enough across the entire range of tones represented, enough to allow for a lot of subsequent processing.
Logged

joofa
Sr. Member
****
Offline Offline

Posts: 485



« Reply #262 on: September 13, 2011, 03:16:12 PM »
ReplyReply

32 bits is 32 bits. If a 32-bit floating point number could represent more than 32 bits worth of information, it would be an information-theoretical break-through.

-h
But the degree of precision is not the same throughout the range of quantities represented.  You have to see here that the "lower" tones of an LDR capture are not of sufficient precision to support the variable blackpoint.  You can't amplify 3 bit quantities and expect to see subtle tonal variations.  But if you have bits allocated where you need them, as in a floating point representation, you have some resolution that is more or less good enough across the entire range of tones represented, enough to allow for a lot of subsequent processing.

I think hjulenissen understands that the gap between floating point numbers is not uniform and I believe he has mentioned that before in different words, just that his articulation could have been better. I think he was trying to inform Bob here that both 32-bit integer and 32-bit floating point representations have the same number of "points (or colors)", just that set of these points or numbers are different between integers and floating points.
 
Logged

Joofa
http://www.djjoofa.com
Download Photoshop and After Effects plugins
LKaven
Sr. Member
****
Offline Offline

Posts: 787


« Reply #263 on: September 13, 2011, 03:56:59 PM »
ReplyReply

Maybe, Bob, the question is indeed about the attributes of a good definition.
(Scientifically) intuitively, I’m rejection your definition. It seems to me dependent on your individual experience and in particular dependent on the very current state of camera technology. Some lines above I had tried to contribute a more general definition, by describing a HDR-discriminator in post-processing, independent from how the data were collected. In your terminology, you would see it (all) as tone-mapping. But I would not exclude completely that it is just based on my experience…

I’m wondering, if there is any "definition theory"
to tackle the KPIs (key performance indicators) for the quality of a Definition.
Never came across this question.

Peter

Hold on for a moment, everyone.

Having spent years studying semantics as a grad student, I would press a point strongly here.  There are some interesting things being discussed here.  Unsurprising, given the bright bunch, some of whom are furthering our understanding of the subject.  But nothing proposed so far in the way of the semantics of HDR has been anything but a non-starter.  

If one is looking for a nominal essence here, something that might be used for a definition, I think it will be very hard, perhaps impossible, to find.  Is there a single necessary set of conditions for being "HDR" or for saying that one is "doing HDR"?  I suspect not.  I think you will more likely find clusters of conditions.  This is why I speak of the "commonplaces".  

In the end, I think the theoretical significance of "HDR" /per se/ in a technical theory of photography is slight.  But the things that we refer to when we say we are "doing HDR" have some very practical value, and help point the way to new theory.

There is nothing essential about bracketing, or "supersampling" of any kind.  However, the practice of supersampling leads into some very useful techniques.  By supersampling, you gain an increase in precision.  Thereby you allocate more bits, and perhaps use floating point to even out the allocation of bits along the entire range of numerical values represented.  With this increased precision, you have the opportunity to process image data having dynamic range that exceeds the dynamic range of your output media.  Thereby you gain the opportunity to have variable white and black points, virtual re-lighting.  All of this is facilitated by the gain in precision and the allocation of bits evenly across the range of values.  The fidelity of the "low tones" is improved.  

This describes widespread practices, but in no way suggests at anything essential or necessary in the semantics of "HDR".

The idea of "tonemapping" may turn out to have more theoretical significance, since as far as I can see, it just refers to the idea of mapping one set of tones onto another set of tones by whatever means.  But we use this generally, and not just when "doing HDR."  Sometimes we map tones having a greater dynamic range onto a set of tones having a lesser dynamic range.  Sometimes not.  
Logged

Peter_DL
Sr. Member
****
Offline Offline

Posts: 421


« Reply #264 on: September 14, 2011, 11:25:03 AM »
ReplyReply

Having spent years studying semantics as a grad student, I would press a point strongly here...  But nothing proposed so far in the way of the semantics of HDR has been anything but a non-starter.  

If one is looking for a nominal essence here, something that might be used for a definition, I think it will be very hard, perhaps impossible, to find.  Is there a single necessary set of conditions for being "HDR" or for saying that one is "doing HDR"?  ...

The idea of "tonemapping" may turn out to have more theoretical significance, since as far as I can see, it just refers to the idea of mapping one set of tones onto another set of tones by whatever means.  But we use this generally, and not just when "doing HDR."  Sometimes we map tones having a greater dynamic range onto a set of tones having a lesser dynamic range.  Sometimes not.

OK - I'll try to explain.

Aside from the basic requirement that Scene DR > Output DR
(and we may discuss about the influence of absolute luminance in cd/m2 separately)
the clear HDR differentiator for me is:  that any global tone mapping function i.e. the "tone curve" fails to provide a pleasing rendition showing all relevant details in the highlights and shadows, so that pixel-selection-based tone mapping techniques are needed.

Whether it is Blending of different camera Exposures or different exposure variants from one shot, or skillful equipment of the tone curve with a luminosity mask – it is irrelevant, in my opinion.

Such Definition is independent from camera DR & technology
and fwiw I’m not aware about any reported alternatives in post-processing (I'll be happy to stand corrected)
- although there is certainly no full disclosure about current software mechanics.



Peter

--
« Last Edit: September 14, 2011, 11:30:32 AM by Peter_DL » Logged
LKaven
Sr. Member
****
Offline Offline

Posts: 787


« Reply #265 on: September 14, 2011, 12:35:20 PM »
ReplyReply

[...]
the clear HDR differentiator for me is:  that any global tone mapping function i.e. the "tone curve" fails to provide a pleasing rendition showing all relevant details in the highlights and shadows, so that pixel-selection-based tone mapping techniques are needed.

Whether it is Blending of different camera Exposures or different exposure variants from one shot, or skillful equipment of the tone curve with a luminosity mask – it is irrelevant, in my opinion.

This appears to rule out simple "compression" techniques which do involve global tonemapping functions, but which one would include as a part of "doing HDR" in some cases.  [You probably meant "e.g." here instead of "i.e." but I understood the point.]  Also, the "pleasingness" condition is rather subjective here, bringing in the entirety of aesthetics in such judgments. 
Logged

Peter_DL
Sr. Member
****
Offline Offline

Posts: 421


« Reply #266 on: September 14, 2011, 01:07:03 PM »
ReplyReply

This appears to rule out simple "compression" techniques which do involve global tonemapping functions...

Yes.

Also, the "pleasingness" condition is rather subjective here, bringing in the entirety of aesthetics in such judgments.

Valid concern,
unless we could define "(HDR) pleasingness" by numbers, i.e. to reproduce relevant highlight & shadows details in a perceivable way.
Seems to me not impossible.

OT – however it may support the point:
http://www.cis.rit.edu/people/faculty/johnson/pub/hue_CGIV04_sm.pdf

Peter

--
Logged
RFPhotography
Guest
« Reply #267 on: September 15, 2011, 03:04:55 PM »
ReplyReply

I think trying to come to some agreement on a set of numbers that represent a 'pleasing' image would be about as difficult as coming up with a cut-off point to determine how many stops mean HDR.  Roll Eyes
Logged
Peter_DL
Sr. Member
****
Offline Offline

Posts: 421


« Reply #268 on: September 16, 2011, 11:03:35 AM »
ReplyReply

 
Just try the inverse approach, Bob:
Take all your bracketed HDR shots and glue them together in any digital way preferred,
but use a global (continuous) function only for tone mapping. Happy with the results  Shocked?

Peter

--
« Last Edit: September 16, 2011, 11:12:08 AM by Peter_DL » Logged
RFPhotography
Guest
« Reply #269 on: September 16, 2011, 11:29:42 AM »
ReplyReply

In point of fact, I have done that.  It's one of the things that makes Photoshop a wonderful tool for HDR.  I've used the regular editing tools in PS to tonemap many HDR images and not run them through the local tonemap operators of HDR Pro or any other HDR software.  No masking, no selections.  It's a different result but it can certainly be one that works.  Unless you consider Levels/Curves/Color Balance/HSB to be local adjustments.

The attached image was merged in HDR Pro, adjusted with Exposure & Gamma, then further adjustments were made using only global tools in PS.  No selections, no masking, no layering of other parts of other images.  This is referred to as 'soft tonemapping'.
Logged
Guillermo Luijk
Sr. Member
****
Offline Offline

Posts: 1274



WWW
« Reply #270 on: September 16, 2011, 12:30:04 PM »
ReplyReply

The attached image was merged in HDR Pro, adjusted with Exposure & Gamma, then further adjustments were made using only global tools in PS.  No selections, no masking, no layering of other parts of other images.  This is referred to as 'soft tonemapping'.

There are only two ways in which you can obtain a pleasant output image just by using global tools:
- The adjustments you made in HDR Pro were far from being global tools (they lifted the shadows for you, at the same time as they preserved the highlights and kept a good amount of local contrast), so this cannot be considered a case of using 'just global tools'.
- Your scene was mid to low DR, so no special tone mapping strategies were needed.

In any other case, global tools won't suffice the process for HDR images, and this is what Peter meant. The reason? DR gap between the input (scene) and the ouput (monitor).

Regards
Logged

hjulenissen
Sr. Member
****
Offline Offline

Posts: 1666


« Reply #271 on: September 16, 2011, 12:46:51 PM »
ReplyReply

In any other case, global tools won't suffice the process for HDR images, and this is what Peter meant. The reason? DR gap between the input (scene) and the ouput (monitor).
I think it makes sense to describe black/white clipping (what any LDR camera will do) as a special kind of global tone-mapping operator. So a scene may have its dynamic range, the final output medium may have its dynamic range, and in-between we have the choice of:
1. Global clipping of blacks and whites
2. Global smooth mapping (curves, levels,...)
3. Local tonemapping

If you only consider input (scene) and output (paper/display), it does not really matter how you got there (stacked images, sony-sensor, tonemapping, in-camera JPEG, whatever). A scene may have more or less dynamic range, but the output will (for now) be low dynamic range. No matter what we do, there will some operator mapping scene to output that usually will be non-linear.

Kind of unusual to think of people snapping the grand canyon with their cell-phone as applying an abrupt global tonemapping to a HDR scene in order to fit into the LDR lcd...


Anyways, I don't see how you can claim that global tonemapping never will suffice for HDR images. I see that local operators often will produce results that you and I and most people may judge as subjectively superior to global operators, but not that this will be the case for all scenes, all of the people and all of the time?

-h
« Last Edit: September 16, 2011, 12:49:01 PM by hjulenissen » Logged
JohnKoerner
Guest
« Reply #272 on: September 16, 2011, 02:56:42 PM »
ReplyReply

The one thing that is clear to me at this point is that several different definitions are needed to keep everybody happy (I admit I feel most comfortable with Andrew's definition, which seems eminently practical).
So I suggest consulting with the experts at Epson's Paper-Naming division to come up with appropriate names for the different varieties of "HDR." We might then have "Archival DR" (or "ADR"), "Enhanced DR (EDR)", "Super Premium DR (SPDR)", etc.
Then we can argue about which name best fits each person's personal vision of "HDR."  
 Huh


I am out of my league on the subject of HDR, but logically-speaking, it must be exactly the opposite: in order to understand "HDR" the term needs to be universally defined.

Based on what little I read, since the term "High-Dynamic-Range" is in reference to a final image, then the image has to be the deciding factor as to whether it qualifies a "HDR" image or not, rather than "how" it was created.

If 5 shots with a 10D can't get me the same DR result as 1 shot with a K5 can, it would be logically-absurd to call the composite image produced from 5 shots with the 10D "HDR" ... when in point of fact the one shot from the K5 has greater DR. It seems pretty obvious that the actual range of the final image is the deciding factor as to whether it is HDR, not the methods used to get it there.

So until someone (or some entity) actually comes up with a concrete definition to go by, it will be impossible for anyone to agree on anything without such a reference point.

Jack


.
Logged
JohnKoerner
Guest
« Reply #273 on: September 16, 2011, 03:00:48 PM »
ReplyReply

In the end, Andrew it's unlikely we'll ever get that level of agreement.  Some, like me, define HDR (the process) more narrowly.  And while I can accept that HDR isn't necessary to produce higher dynamic range scenes, I'm not as willing as others to use the term to describe all methods.  I've said it before and I'll say it again; I prefer Caponigro's XDR terminology to describe all processes of producing higher/extended dynamic range which include software like Enfuse/Tufuse, SNS HDR (which isn't technically an HDRI software app), HDR software apps and manual blending.
 
I absolutely agree that tonemapping isn't an HDR-only term.  Again, I've said it before and I'll say it once more, tonemapping is just a fancy word for editing ...

Agreed. If that were even close to possible, we’d also have Canon and Nikon saving DNG, we’d have a true raw histogram on the camera, people would never mix up DPI and PPI, and never confuse dynamic range an bit depth. But some can hope for such clarity some day <g>.


I am sorry, but I find this discussion quite entertaining

Jack



.
Logged
Guillermo Luijk
Sr. Member
****
Offline Offline

Posts: 1274



WWW
« Reply #274 on: September 16, 2011, 03:02:43 PM »
ReplyReply

Anyways, I don't see how you can claim that global tonemapping never will suffice for HDR images. I see that local operators often will produce results that you and I and most people may judge as subjectively superior to global operators, but not that this will be the case for all scenes, all of the people and all of the time?

If you only use global tools (by global I mean operator affecting the entire image, like curves without layer masks), you can arrange the input DR onto the output DR (a simple curve will suffice), but the result will not be satisfactory if the source DR was very high due to the lack of local contrast. Local processing is needed.

Regards
Logged

RFPhotography
Guest
« Reply #275 on: September 16, 2011, 03:25:49 PM »
ReplyReply

The scene was not of mid to low contrast.  Note the highlights in the flowing water and the deep shadow areas of the rock.  Holding the highlights in the water and generating sufficient shadow detail wasn't possible with a single shot.  

Levels/Curves are global operators.  As I noted, the image was merged in HDR Pro then the Exposure & Gamma tool was applied.  This too is a global operator.  From there the only other tools used were Levels/Curves (to be honest I don't recall which and as I was doing this for a test I didn't save the working TIFF file) and Hue/Sat.  

To be perfectly frank, GL I really don't care whether you believe that's all I did or not or whether you believe the scene could have been properly rendered in a single shot.  We don't agree.  We won't agree.  Likely on anything.  And I'm just fine with that.

Logged
RFPhotography
Guest
« Reply #276 on: September 16, 2011, 03:28:20 PM »
ReplyReply


I am sorry, but I find this discussion quite entertaining

Jack



.

It's a very nuanced discussion, isn't it Jack?  Grin

The issue with using the final image as the determination of whether an image is HDR or not is that the final image isn't HDR either.  Take a camera like the K5 or D7000 that can capture around 12 stops of brightness.  Most monitors can't reproduce that nor can any printing media.  So the final image, the one that appears 'proper' on screen or is printable doesn't have all the DR that the camera was able to capture.  It's been edited or tonemapped down into something narrower that can be used.  Think of it like a Zone System image where the exposure was increased to generate desired highlight detail.  If the scene DR was, say, 8 stops and the shadow detail was lifted 2 stops, you're most likely going to have blown out highlights without adjusted development of the film.  But if the film undergoes contracted development, those highlights can be brought back into range.  That's one form of tonemapping.  The contracted development maps highlight tones from higher to lower zones.  Then, in printing, additional techniques would be used to render a pleasing print taking into account the fact that the paper couldn't reproduce all 8 stops of brightness.  That's another form of tonemapping.  But that final printed image doesn't contain all of the brightness range that was captured initially.  So if we use the final image as the determination, until technology changes no image would be considered HDR.
« Last Edit: September 16, 2011, 03:38:44 PM by BobFisher » Logged
Guillermo Luijk
Sr. Member
****
Offline Offline

Posts: 1274



WWW
« Reply #277 on: September 16, 2011, 03:50:03 PM »
ReplyReply

To be perfectly frank, GL I really don't care whether you believe that's all I did or not or whether you believe the scene could have been properly rendered in a single shot.  We don't agree.  We won't agree.  Likely on anything.  And I'm just fine with that.

Of course I believe that's all you did, the wrong part is thinking what HDR Pro did for you was a global operation. In any case, calm down man, you will survive without agreeing with me. I promise.

Logged

hjulenissen
Sr. Member
****
Offline Offline

Posts: 1666


« Reply #278 on: September 16, 2011, 03:50:56 PM »
ReplyReply

So if we use the final image as the determination, until technology changes no image would be considered HDR.
"If the final image contains visible details that spanned significantly more than a 100:1 (or 1000:1?) range in the original scene"?

High-DR scene -> clipping -> jpeg: LDR
High-DR scene -> Bracketing -> tonemapping -> jpeg: HDR
High-DR scene -> Pentax K5  -> tonemapping -> print: HDR
Low-DR scene -> clipping -> jpeg: LDR
Low-DR scene -> Bracketing -> tonemapping -> jpeg: LDR
Low-DR scene -> Pentax K5  -> tonemapping -> print: LDR
Logged
RFPhotography
Guest
« Reply #279 on: September 16, 2011, 05:13:29 PM »
ReplyReply

Oh, I know I'll survive.  No calming down needed.  Merging to 32 bit then dropping to 16 bit using only Exposure & Gamma, leaving the default values drops to a lower bit depth with no alterations.

H, not entirely sure what you're trying to get at.  If the final image contains a range of brightness that monitors can't reproduce and that printing media can't reproduce then sure, if the desire is to leave it in that state as the final image then yes, if the final image is the ultimte determining factor; and ignoring that it doesn't fit my definition of HDR (nor I'd suggest pretty much anyone else's who'd be considered an authority on HDR), then yes that would be HDR.  I was writing on the basis that the 'final' image is something that could generally be 'used'; as in viewed properly on screen or rendered in a print.  I'm not going to go back and read everything again so I don't know where the italicised quote comes from.  

As it is, it's Friday night and I definitely have far better things to do.  Wink
Logged
Pages: « 1 ... 12 13 [14] 15 16 17 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad