Tone Mapping

[kookeer]

Could someone explain to me what tone mapping is and how to do it?

[Colin Southern]

Could someone explain to me what tone mapping is and how to do it?

Hi Kookeer,

When we take a normal photograph and print or display it we usually have some (usually small) areas that are completely black, and some (usually small) that are completely white. It's not possible to display anything blacker or whiter than these two extremes.

Now normally the range of tones between these extremes of complete blackness and total brightness is adequate to display most "normal" scenes; for arguments sake, lets consider a value of 0 to be black and 100 to pure white / maximum brightness.

Now lets go shoot a scene that has dark shadows at one extreme, and extremely bright sunlight at the other ... if we assigned numbers to them we might still have 0 at the dark end, but the brightness of the other extreme might equate to, say, 500.

Since our cameras can't capture a scene that ranges in value from 0 to 500, we can take a number of shots that cover small portions of that range, say, 0-149 - 100-249, 200-349 etc - so in this case we might end up with 4 or 5 shots that between them cover the entire range we need.

Software like Photoshop (and many others) can take these images that overlap in tonal range and combine them internally into the full (in this case) 0 to 500 range; I say "internally" because the image can't exist outside of the computers internal representation of it (not even on the screen) ... our screen and paper are very limited in the range that they can display ... look at the sun and it'll burn your eyeballs ... look at a PICTURE of the sun on the screen and it's not a problem.

So on one hand our computer has this image tucked away inside that contains values from 0 to 500, but needs to turn them into something that can be displayed or printed (ie with values from 0 to 100) - so something has to give - and that's where tone mapping comes in.

Basically what the software does to overcome the limitations of the display device or media is translate (or "map") these values in the 0 to 500 range into the 0 to 100 range. Tonal range compression is essentially what's going on, but the range isn't compressed in a linear way - generally you want brighter things to remain brighter than darker objects, but the actual values used also depend on other factors like where the value occurs on the page, and what the values of other objects are around it (local contrast) - and that's where the fun starts.

Unfortunately, the vaguerancies of human vision are often difficult to put into a program, and what may "appear logical" to a computer doesn't always look good to the human eye, which is why we see a variety of programs trying to emulate human vision with different program models and give different results (along with other variables that the user selects) - and the final tone-mapped composite will also nearly always require additional post processing.

The limited HDR that I do is usually done with the intention of making the final result as photorealistic as possible - however - others like to process the images in a way that gives an altogether different look (more "arty") which, personally, I also find very appealing. The unfortunate side-effect though is that many people are now beginning to associate this "arty" look with HDR whereas it really should be associated with particular image techniques for processing, whether the source of the image is HDR or not.

Does this help?

[Dave Humphries]

VERY well put Colin

[Wirefox]

Colin

This is the best explanation I have read on the subject. Very clear and concise. Is there any way these various fundaments can be collated in a single post (I don't pretend to know how to even take the handbrake off when it comes to the mechanics of managing forums so I am not sure how that would be done). Perhaps it could be done as an introduction to digital photography etc. I am sure people would find it more than useful. Better still put it all into a manuscript, get a publisher and get paid for it.....I would buy it

Excellent work...

Steve

[Colin Southern]

Is there any way these various fundaments can be collated in a single post (I don't pretend to know how to even take the handbrake off when it comes to the mechanics of managing forums so I am not sure how that would be done). Perhaps it could be done as an introduction to digital photography etc. I am sure people would find it more than useful.

Thanks for that!

Antonio suggested the same thing to me in a PM earlier on today -- Possibly in the form of a locked & sticky thread. Something I'll give some more thought to, as I'm sure Dave and Sean will too.

Better still put it all into a manuscript, get a publisher and get paid for it.....I would buy it

Believe it or not, I've thought of doing that too (along the lines of field techniques as they apply primarily to landscape photography), but the bottom line is truely the bottom line, unfortunately - and that is that things like that take up a LOT of time, and the chances of making anything out of it are minimal. I enjoy helping and educating, but I think for me, this site probably helps more people than a book would.

Perhaps when I retire

[kookeer]

Thank you so much for the explanation. I recently updated PS7 to PS CS3. This is all new to me now and I will have to reread it over and over and try to absorb it. I thought I was getting better with understanding photography but I still don't understand a lot of the tutorials. A work in progress.
Thanks again.
Kookeer

No worries

CS3 has a lot in it, but it's probably not the best choice for doing HDR work. If you want a good book for doing general photo processing in CS3, check out this book from Scott Kelby (I've shown it to two people who both ordered copies!)

[Hntr]

Thanks for the explanation Colin!
I just started working with HDR a few weeks ago and still don't really know what I'm doing. This helped a great deal.

[arith]

look at the sun and it'll burn your eyeballs ...

Sure does Colin you should my eyeballs; great explanation though.

[Colin Southern]

No worries

Just remember that 9 times out of 10 the problem isn't that the camera can't capture the ranges we need - it's more of the fact that we can't print or display the dynamic range that is captured.

[Sahil]

Boy, what an explanation. Even I had the same doubts, but never found something as good as this, to make me understand.

I was wondering that most of the members here are professional photographers or they take photography as just their hobby?

I am working with financial consultants, as a Credit Analyst, and photography is my newly discovered hobby. And I am loving it.

[atlantean]

Just remember that 9 times out of 10 the problem isn't that the camera can't capture the ranges we need - it's more of the fact that we can't print or display the dynamic range that is captured.

Would i get a Nobel Prize if i successfully invented a printing or displaying medium that can show everything with their human-sight-level fidelity (including the eyeball-bursting sun, for that matter) ?

[GiantTristan]

Would i get a Nobel Prize if i successfully invented a printing or displaying medium that can show everything with their human-sight-level fidelity (including the eyeball-bursting sun, for that matter) ?

Just to put things in perspective - The human eye can discern differences in brightness of about 1:10 000 or about 14 stops, while a high end digital camera like the D700 has a dynamic range of about 1:4000 or 12 stops. This is sufficient for most situations. Therefore, a single raw file contains in most cases all the information that is needed. My NEC monitor has a contrast ratio of 1:500 or 9 stops. I think the highest contrast monitors are around 1:3000, though I don't know how good they are.

It seems that the Nobel Prize is within reach...

[Colin Southern]

My NEC monitor has a contrast ratio of 1:500 or 9 stops. I think the highest contrast monitors are around 1:3000, though I don't know how good they are.

The problem with monitor contrast ratios are that they're more about theory than practice; eg if you calibrated and profiled a typical screen and then put a circle with a level of "2" inside a square with a level of "1", you wouldn't see it - so in reality monitors work in the range of around 5 to 6 stops of contrast. Past that and you really have to turn to monitors like Brightside with localised backlighting to get you up to around 11 stops.

Interestingly even a top-of-the-line Eizo is still only around 500:1 contrast ratio.

Regarding the Nobel ... if Al Gore can get one then there's hope for all of us

[rogerb]

Interestingly even a top-of-the-line Eizo is still only around 500:1 contrast ratio.

So even high end monitors struggle to get a "reasonable" dynamic range. Now, I expect that prints would have much less dynamic range than that. Would you say it's in the range of 4 - 5 stops?

That being the case, we're taking photos with cameras capable of 10 - 12 stops of dynamic range and then having to squeeze the tonal range down considerably in the printed image. Are we gradually generating images that are more and more "flat" looking? I know there is a lot of discussion about making sure we can see details in all the shadows, but shadows are there for a reason. Are we loosing overall visual impact by this "tone-mapping/tone-flattening" process?

Roger

[Colin Southern]

So even high end monitors struggle to get a "reasonable" dynamic range.

To be honest Roger, I tend to think of even 500:1 as being perfectly "reasonable" or "perfectly adequate". The crux of the issue is that LCD monitors are all lit by a flourescent tube - so if the crystals are not attenuating the light at all then the white that you get is - in essense - as bright as it's ever going to get - so if you have that set to something fairly bright (I like 200cd/m2) then the only way to increase the contrast ratio is at the "dark end" where - frankly - the difference between a deep deep deep gray and a deep deep deep deep gray just isn't significant (although the manufacturer of the latter will gloat that his product has TWICE the contrast ratio of the former). So in other words once you get past about 500:1 (in my opinion anyway) anything better is simply lumped into the same category: "more than adequate". If 500:1 is good enough for Eizo, then it's good enough for me

Now, I expect that prints would have much less dynamic range than that. Would you say it's in the range of 4 - 5 stops?

Yes - about 4 stops normally. In reality we tend to (technically) misuse the term "black"; if somethig were truely black then we wouldn't be able to see it (even when shining a powerful light on it) - so in reality were just dealing with shades of gray, with blacks as we know it still reflecting quite a lot of light. That's why a bride in a white dress standing next to a groom in a black suit is easy to capture from a technical point of view as it only requires around 4 stops of DR from a camera that can typically grab 12 (assuming no backlighting or specular highlights) - displaying them usually isn't too difficult on a calibrated and profiled monitor - but you have to be a lot more precise in printing because the DR is so limited.

That being the case, we're taking photos with cameras capable of 10 - 12 stops of dynamic range and then having to squeeze the tonal range down considerably in the printed image. Are we gradually generating images that are more and more "flat" looking? I know there is a lot of discussion about making sure we can see details in all the shadows, but shadows are there for a reason. Are we loosing overall visual impact by this "tone-mapping/tone-flattening" process?

At the end of the day, 12 doesn't go into 4, so something has to give. Tone mapping gives us a technical solution, but local contrast plays a BIG part in making the image believeable (hence you can have "naked sun" in the top 1/2 of the image (pure white) and pure white in the bottom of an image from a reflective object, and the image is visually acceptable even though in reality metered light from the sun would be many many times brighter than the light from the reflected object. Unfortunately many HDR programs struggle with this part. It's my belief that it's the job of us humans to them take the auto-processed HDR image and then apply manual corrections to make the image believeable; unfortunately the current generation seems to be accepting what comes out of the likes of Photomatix as "gospel" because it's a different look (which they come to mis-associate with being the "HDR" look).

[Terry Tedor]

Definitely a very informative thread. Colin, thanks for taking the time to explain it.

Unfortunately many HDR programs struggle with this part. It's my belief that it's the job of us humans to them take the auto-processed HDR image and then apply manual corrections to make the image believeable; unfortunately the current generation seems to be accepting what comes out of the likes of Photomatix as "gospel" because it's a different look (which they come to mis-associate with being the "HDR" look).

I take it then you don't have a very high regard for Photomatx? I dabbled with it a time or two, but not knowing what I was doing I was extremely disappointed with the results. I would like to experiment with HDR. Can you make a recommendation or steer a newbie in the right direction?

I was wondering that most of the members here are professional photographers or they take photography as just their hobby?

Just a hobby for me for about the last 40 years or so.

[Colin Southern]

Definitely a very informative thread. Colin, thanks for taking the time to explain it.

No worries Terry

I take it then you don't have a very high regard for Photomatx? I dabbled with it a time or two, but not knowing what I was doing I was extremely disappointed with the results. I would like to experiment with HDR. Can you make a recommendation or steer a newbie in the right direction?

To be honest Terry, I was beginning to think that it was a bit of a "one trick pony", but now that I've seen that it's capable to photorealistic results, I've concluded that the "over-processed" look must either be by choice, or due to lack of final processing.

Personally, I only ever work with photorealistic HDR, and in that regard, Photoshop works just fine for me. Having just said that, I typically compress the DR with GND filters a lot more than I rely on software.

[Terry Tedor]

No worries Terry


To be honest Terry, I was beginning to think that it was a bit of a "one trick pony", but now that I've seen that it's capable to photorealistic results, I've concluded that the "over-processed" look must either be by choice, or due to lack of final processing.

Personally, I only ever work with photorealistic HDR, and in that regard, Photoshop works just fine for me. Having just said that, I typically compress the DR with GND filters a lot more than I rely on software.

I think a lot of it is indeed by choice. It seems to have become the latest fad. Any tips on using Photoshop CS4 to do photorealistic HDR? I'm still crawling along with it as I just upgraded from the original CS to CS4 over the holidays.

[GiantTristan]

To be honest Roger, I tend to think of even 500:1 as being perfectly "reasonable" or "perfectly adequate". The crux of the issue is that LCD monitors are all lit by a flourescent tube - so if the crystals are not attenuating the light at all then the white that you get is - in essense - as bright as it's ever going to get - so if you have that set to something fairly bright (I like 200cd/m2) then the only way to increase the contrast ratio is at the "dark end" where - frankly - the difference between a deep deep deep gray and a deep deep deep deep gray just isn't significant (although the manufacturer of the latter will gloat that his product has TWICE the contrast ratio of the former). So in other words once you get past about 500:1 (in my opinion anyway) anything better is simply lumped into the same category: "more than adequate". If 500:1 is good enough for Eizo, then it's good enough for me



Yes - about 4 stops normally. In reality we tend to (technically) misuse the term "black"; if somethig were truely black then we wouldn't be able to see it (even when shining a powerful light on it) - so in reality were just dealing with shades of gray, with blacks as we know it still reflecting quite a lot of light. That's why a bride in a white dress standing next to a groom in a black suit is easy to capture from a technical point of view as it only requires around 4 stops of DR from a camera that can typically grab 12 (assuming no backlighting or specular highlights) - displaying them usually isn't too difficult on a calibrated and profiled monitor - but you have to be a lot more precise in printing because the DR is so limited.



At the end of the day, 12 doesn't go into 4, so something has to give. Tone mapping gives us a technical solution, but local contrast plays a BIG part in making the image believeable (hence you can have "naked sun" in the top 1/2 of the image (pure white) and pure white in the bottom of an image from a reflective object, and the image is visually acceptable even though in reality metered light from the sun would be many many times brighter than the light from the reflected object. Unfortunately many HDR programs struggle with this part. It's my belief that it's the job of us humans to them take the auto-processed HDR image and then apply manual corrections to make the image believeable; unfortunately the current generation seems to be accepting what comes out of the likes of Photomatix as "gospel" because it's a different look (which they come to mis-associate with being the "HDR" look).

This is a very informative post. I would just like to add the following:

The newer high-end monitors have LED instead of fluorescent back lighting. This should help with the "white end".

A good approximation to a perfect "black body" consists of a large box painted black on the inside. If you look at a small hole in this box you see a "deep black" in other words the hole is a total black body that absorbs all incident radiation and reflects none. To my knowledge, no monitor is capable of displaying a perfect black.

[Colin Southern]

The newer high-end monitors have LED instead of fluorescent back lighting. This should help with the "white end".

It'll be interesting to see where this leads us. A company called brightside developed a monitor where they would vard the backlight LEDs to give up to around 11 stops of DR, but I haven't heard anymore of them recently, so not sure what the latest is.

A good approximation to a perfect "black body" consists of a large box painted black on the inside. If you look at a small hole in this box you see a "deep black" in other words the hole is a total black body that absorbs all incident radiation and reflects none. To my knowledge, no monitor is capable of displaying a perfect black.

I'm interested to see if/how LED monitors might fare in this respect; in theory if the LED isn't emitting at all then it should be black, with the only light coming from reflected room lighting.

The other issue one could face though is having monitors with TOO much DR - it would then make it very hard to translate an image on the screen to one on paper.