tone mapping - before and after

[Guest WPalank]

Advertisement (gone after registration)

Far better off bracketing and reassembling things manually?

 

rob my friend,

 

Care to demonstrate (before and after)? Also, set a timer and tell us how long from start to finish.

I'll start you off, open the image as a "Smart Object" then "New Smart Object via Copy". Timer starts now.....

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[allamande]

William,

 

Thank you for these very helpful instructions. I made a note. I was questioning why the tones are sometimes altered (on the warm side) excessively once the image goes through TM. I'm trying to understand what goes on inside the TM black box in this regard.

 

I use some of the methods you suggest to adjust temp, but your note is so much more comprehensive than my usual ways that you taught me some new approaches as a result. I am grateful for that.

 

Ece

[Big T]

Here are two images taken with the M8.2 in DNG format, tone mapped and then converted to JPG.

 

Only a single image was used in each case and it was Tonemapped in Photomatix via the "Batch Single Files" process. You can only follow this process with RAW files (a single JPG will not work).

 

Try it with a Leica DNG file and you maybe surprised what it can do with the great dynamic range that DNG files have.

 

 

 

 

[pop]

Trevor - that's what I suspected a long while ago: that our cameras delivered a much greater useful depth than we normally could access.

[pop]

Yet another tidbit of theory. Do let me know when you're all fed up.

I've found yet another tool. It helps me visualize what the tonemapping procedure does. Indeed, it lets me analyze any change in the color space of an image at all. For those interested enough, here's what I did in order to get a grasp on the parameter for the gaussian blur which is rather central to the procedure.

This is my standard gray chart which I developed for the purpose. As you can see from the graph below it, each gray value is by the same amount lighter than the one to the left of it. Contrary to what your eyes might tell you, each gray field is uniformly gray.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

The following is the same chart after I applied the gaussian blur to it. I set the blur radius to 11 pixels which ought to be 2% of the image width. You can see that the middle regions of every gray field has still the same brightness as before. However, the edges have become round, which is to be expected of a well-behaved gaussian blur.


The following is the gray chart after applying the whole tone mapping procedure. The values I applied are 2-75-90-1 (in the order they appear in the GIMPs dialog window).


The next one is the same gray chart, again after applying the tone mapping procedure. This time, the values I applied are 10-75-90-1. In other words, I changed the blur radius and nothing else.


This is one more step towards a functional understanding of the tone mapping procedure, i.e. what its effects are on the image.

The first thing which jumps out is that the gray patches are not uniformly coloured, any more. Rather, the graph depicting the brightness of each patch is now sloped. This leads to increased contrasts on the interfaces between each pair of patches. The width and tilt of the slope is, of course, controlled by the blur radius.

[rob_x2004]

Philipp, most of the effect seems to be in hte middle and restrained a bit in the ends.

Does this extrapolate to most of the tone mapping occurs where you least need it, and that recovery of probelmatic highlights and shadows is not its forte? Also the 'shaping' of the middle seems consistent with flattening the overall contrast in an image? This is interesting stuff. Thanks.

[daveleo]

Advertisement (gone after registration)

. . . . . The first thing which jumps out is that the gray patches are not uniformly coloured, any more. Rather, the graph depicting the brightness of each patch is now sloped. This leads to increased contrasts on the interfaces between each pair of patches. The width and tilt of the slope is, of course, controlled by the blur radius.

[/schmal]

 

cool ! ! . . . . that is apparently how the image "maintains local contrast" while keeping (or even decreasing) the global contrast . . . it's the edge-of-the-patches effect.

 

it also explains why the images look "sharper" . . . in fact to me, this is the big problem with most of them . . . some surfaces (like skin) look too too sharp after tone-mapping.

 

A very nice piece of detective work, Philipp ! !

 

 

For the record, 99% of the time now i am working with just two layers as i described in my last image posting above. For one thing, it helps get around this over-sharped effect.

[jaapv]

Interesting thread. What happens if you go into LAB and apply the tonemapping to the channels?

[pop]

Philipp, most of the effect seems to be in hte middle and restrained a bit in the ends.

Does this extrapolate to most of the tone mapping occurs where you least need it, and that recovery of probelmatic highlights and shadows is not its forte? Also the 'shaping' of the middle seems consistent with flattening the overall contrast in an image? This is interesting stuff. Thanks.

 

A closer look reveals that the "real" contrast is greatest at both ends of the range.

 

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

 

You can clearly see that the step from the "very black" to the "merely black" square is much taller than the step between to squares near the middle of the range. Anyway, there are four parameters to that single tonemapping procedure, and I've changed but one.

 

As to the applicability of the procedure: look at the examples in this thread. Some have been processed using this procedure. Some have gained in the process, some - a bit less.

[pop]

cool ! ! . . . . that is apparently how the image "maintains local contrast" while keeping (or even decreasing) the global contrast . . . it's the edge-of-the-patches effect.

 

That's exactly how it is done. It exploits an idiosyncrasy of the brains PP which you can perceive above in the unprocessed gray scale. Each patch appears darker next to a brighter one and brighter next to a darker one. The procedure mainly adjusts the brightness of the area within the patch to exaggerate the effect.

 

it also explains why the images look "sharper" . . . in fact to me, this is the big problem with most of them . . . some surfaces (like skin) look too sharp after tone-mapping.

 

That's because the sharpening procedures work pretty much the same way. As a matter of fact, most "sharpening" procedures do not actually sharpen but they merely raise the contrast in the vicinity of edges. If I find the time, I will do some illustrations on sharpening.

[pop]

Interesting thread. What happens if you go into LAB and apply the tonemapping to the channels?

 

The procedure I am exercising here actually works on color images. I am using a colorless color image here in order to reduce some of the complexity. This leads, of course, to some questions not being answered, most notably the question why some of the images processed this way turn out to have colors like candy.

 

I wouldn't advise applying the procedure to individual channels, though. I think you might get some startling colors if you manipulate the constituents separately.

[pop]

If I find the time, I will do some illustrations on sharpening.

 

Here:

 

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

 

I made this incredibly complex image (part 1) with the GIMP. From left to right:

- Original box drawn with a pencil and then filled

- blurred with gaussian blur

- blurred with gaussian blur then sharpened a bit with unsharp mask

- original box sharpened a bit with unsharp mask.

 

Both blurs and unsharp masks have been done at the same time, therefore with the same value.

 

I then applied the Image Analyzer to draw the by now nearly well-known graphs.

 

I'm sure you will spot the similarity of the edges in this example to those in the tone mapped one.

[Jamie Roberts]

Interesting thread. What happens if you go into LAB and apply the tonemapping to the channels?

 

Yeah, except instead of "tonemapping" and HDR, I'd call it "curves, dodging and burning"

 

Oh, and layer modes Most of the saturation effect, as you know, could be handled in LAB in a couple of clicks; the L channel is too tweaky perhaps to gain the shadow detail, but it certainly could be done.

 

Actually, you can do a lot of the dodging and burning directly from the RAW file with curves (and individual channel curves) in C1... I'm sure at some point it will have a layers interface as well, so you won't get the colour torquing and saturation that's typical of "HDR" style processed images. LR probably does this already...

 

But I'm glad people are seeing how much detail there is in a digital M shot by essentially applying some fancy curves, local contrast and sharpening Dave--if you want to follow up, you should read "The Canyon Conundrum" by Dan Margulis... it's a great resource for understanding how to get these effects (and many more) without mystery... well, without much, anyway

[pop]

.., I'd call it "curves, dodging and burning"

 

which happens to be tone mapping, technically speaking.

 

Actually, you can do a lot of the dodging and burning directly from the RAW file with curves (and individual channel curves) in C1... I'm sure at some point it will have a layers interface as well,

 

Quite. There's a few post near the beginning of this thread which show the individual steps to use for one of the more frequently used tone mapping procedures with the GIMP.

 

so you won't get the colour torquing and saturation that's typical of "HDR" style processed images. LR probably does this already...

 

But then, you can get them with any kind of PP software, as you can get very "decent" and unobtrusive results even with the full monty of HDR.

 

But I'm glad people are seeing how much detail there is in a digital M shot by essentially applying some fancy curves, local contrast and sharpening

 

The tone mapper's arsenal actually contains more tools than just those. But then, all of those tools are the standard staple of practically all PP programs.

[daveleo]

Philipp, more great science from you !

 

James, going back to the start of this thread, we backed into the topic and have been "reverse-engineering" what tone-mapping is and what it does. Of course anything that a canned program can do, you can do manually, once you figure out what the program is doing. That's pretty much where this exploration is leading.

 

Right now, I manually use 2 or 3 layers and dodge, burn, multiply, divide,whatever, and tweak layer opacities until I'm happy . . . then I toss it out and start all over . . .

 

I haven't had a successful "tone-mapped" image in several weeks now.

[Jamie Roberts]

(curves) which happens to be tone mapping, technically speaking.

{snipped}

The tone mapper's arsenal actually contains more tools than just those. But then, all of those tools are the standard staple of practically all PP programs.

 

These are my points precisely; any digital post process includes "remapping" tones and pixels from one source to another (optical post processing does the same thing without, of course, the pixels ). Some techniques and tools are better than others for achieving certain ends.

 

Personally, I like doing a lot of the initial curves and contrast adjustment in a RAW program, where you have the most amount of information to work with. I don't know if the GIMP works on RAW files or not (I gave it up when I gave up LINUX --and it certainly didn't read DNGs at that point)

 

I did read the whole thread, too... I just haven't seen anything from the output that couldn't be easily reproduced with standard RAW converter or PS layer / image mode and curves tools... that's all. With a lot less "mystery" as well (though I admit what PS does in LAB mode is very often counter-intuitive.... useful though it is).

 

None of which says I didn't find the reading interesting. I think figuring out how a plugin does its stuff is interesting, if not ultimately different from the many other ways to skin that particular cat.

[pop]

I just haven't seen anything from the output that couldn't be easily reproduced with standard RAW converter or PS layer / image mode and curves tools... that's all. With a lot less "mystery" as well

 

Not having much experience with PP, the technique employed to apply variable contrast enhancements depending on the brightness of the environment was quite novel to me. In that sense, analyzing what the "original" procedure did was an eye opener for me.

 

And, no, the GIMP does not process raw images and keeps the color depth to 8 bits per channel.

[thighslapper]

All this sounds like unnecessarily complicated hard work to me....

 

I just set my 'HDR' profile of 3 or 5 bracketed images which the M9 rattles off instantly.... fire up Oloneo Photengone (free in beta), load the images, press a button, fiddle with a few sliders and save (takes 20 secs max) and export as TIFF/JPEG. Output is realistic, sharp, artefact free and needs no other adjustment.

 

This is Gdansk last weekend taken at sundown into the sun with the sun behind the buildings.

 

Second is Trinity College Chapel Cambridge from the weekend before.

 

 

HDR is basically software infill flash.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[thighslapper]

I mentioned Oloneo on another HDR thread a while ago.

 

Beats all the other HDR progs by a mile.

 

This is Gdansk again (open late for the annual amber fair), and my local church a few weeks ago. Just to show the others aren't flukes...

 

All hand held or with the camera resting or propped up against something. I'm usually too lazy to use/carry a tripod.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[jaapv]

All this sounds like unnecessarily complicated hard work to me....

 

I just set my 'HDR' profile of 3 or 5 bracketed images which the M9 rattles off instantly.... fire up Oloneo Photengone (free in beta), load the images, press a button, fiddle with a few sliders and save (takes 20 secs max) and export as TIFF/JPEG. Output is realistic, sharp, artefact free and needs no other adjustment.

 

This is Gdansk last weekend taken at sundown into the sun with the sun behind the buildings.

 

Second is Trinity College Chapel Cambridge from the weekend before.

 

 

HDR is basically software infill flash.

For static subjects - yes. But what do you do when something moves and you get ghosting?