'HDR' vs 'tone mapping'

[Shadowblade]

Just a quick rant - why do people insist on calling unrealistically tone-mapped photos generated from a single image 'HDR' or 'pseudo-HDR', or even associating unrealistically tone-mapped images generated from an HDR file with the HDR process, rather than the tone-mapping process?

HDR simply refers to combining several images, taken at different exposure settings but compositionally identical, to produce a single image with greater dynamic range than is possible with a single exposure. It is no more or less realistic in appearance than any other image, although, without further processing, it needs to be 'windowed' through different exposure values to see the highlights and shadows properly.

Every digital image - HDR or not - needs to be tone-mapped, otherwise it's just a string of 1s and 0s. It's in the curves setting of DPP or done automatically by the camera, and can be further modified in Photoshop using Curves, Levels and other tools, to give either realistic or unrealistic results, depending on the intent of the user, with either HDR or non-HDR images.

Associating unrealistic images with HDR (even when they're not HDR images) just gives HDR a bad name, particularly when it's such a good technique for producing completely-realistic images of scenes which could otherwise not be photographed in a single exposure, due to non-straight horizons and the like.

/rant

[wolfden]

I kinda raised something similar to this earlier:
Too Much HDR Confusion?
http://photography-on-the.net/forum/showthread.php?t=843177

HDR and Tone Mapped images have just been lumped together. I have a feeling if you made two seperate forums, the HDR forum, which would be true HDRs, would hardly have any posts while the Tone Mapped one would have plenty. On other forums I don't see it questioned as much as I do here. I understand the view point, so should they be seperate forums?

[dugcross]

HDR and Tone Mapped images have just been lumped together. I have a feeling if you made two seperate forums, the HDR forum, which would be true HDRs, would hardly have any posts while the Tone Mapped one would have plenty. On other forums I don't see it questioned as much as I do here. I understand the view point, so should they be seperate forums?

I agree with that. The problem is that too many photographers still think their one exposure image is a HDR image. They refuse to listen to anybody and think that just because at the very least they created 2 more images in lightroom or whatever from that single image and took it through Photomatix that the image is HDR when it is not. It is just a tone-mapped image! If you start out with a single image, no matter what you do to it, it will STILL not be a HDR image, it is tone-mapped. Tone-mapping an image does not magically make it HDR.

[wolfden]

I agree with that. The problem is that too many photographers still think their one exposure image is a HDR image. They refuse to listen to anybody and think that just because at the very least they created 2 more images in lightroom or whatever from that single image and took it through Photomatix that the image is HDR when it is not. It is just a tone-mapped image! If you start out with a single image, no matter what you do to it, it will STILL not be a HDR image, it is tone-mapped. Tone-mapping an image does not magically make it HDR.

I agree, some where along the line HDR and Tone Mapped just got lumped together and now we have the confusion to deal with. I don't know the solution but to educate that there is a difference.

[dmccabe]

Many people still do not understand "dynamic range"... and even with two images, are not getting HDR.

And on the other hand, I have seen many single images tone mapped that look fantastic.

But the bottom line is people need to stop worry about "post" techniques... and only judge the final image... not how it got there.

When people ask me what tools I use, I reply the "expensive" ones... and my prices are going up, so you better buy now.

[Gary McDuffie]

As several of you guys know, we've discussed this until we were blue in the face in other threads. You also know that I don't object to either process. What I do object to is the high number of people calling what are simply overprocessed tonemapped images HDR. Even more, I object to the fact that many of them don't even know any better. People just keep posting stuff and calling it HDR, which misleads the newbies that come in and read each thread and get it into their head that they are.

[bunyarra]

Being fair to single image processing. With cameras like the 5DII in RAW, you do have significant detail lost at the extremes - the DR captured is greater than can be shown by one image processed normally.

Creating 3 separate images from one original RAW and putting that through the "HDR" process will generate an image that can accurately be classified as derived from a scene's high dynamic range. Though, most will agree, it is High(ish) :)

[dugcross]

Creating 3 separate images from one original RAW and putting that through the "HDR" process will generate an image that can accurately be classified as derived from a scene's high dynamic range. Though, most will agree, it is High(ish) :)

The result of that process will be a tone-mapped image NOT a HDR image.

[bunyarra]

The result of that process will be a tone-mapped image NOT a HDR image.

I did not even mention the tone mapping.

Image #1 = -1/1.5 EV
Image #2 = 0 EV
Image #3 = +1/1.5 EV

This is the same as exposing 3 images with -1/0/+1 compensation. No matter how it is derived, when your image prior to tone mapping was created from images with more dynamic range than a single exposure can accomplish, you have created an HDR photo.

I do not argue that the range is less than you would ideally need.

[Shadowblade]

I did not even mention the tone mapping.

Image #1 = -1/1.5 EV
Image #2 = 0 EV
Image #3 = +1/1.5 EV

This is the same as exposing 3 images with -1/0/+1 compensation. No matter how it is derived, when your image prior to tone mapping was created from images with more dynamic range than a single exposure can accomplish, you have created an HDR photo.

I do not argue that the range is less than you would ideally need.

It's not the same. You're only with the dynamic range available from a single exposure - just that one image has been pulled one stop and another has been pushed one stop. You can theoretically achieve exactly the same result with the curves function in Photoshop (or, since RAW is involved, by adjusting curves in DPP prior to RAW conversion) - you have mapped the shadows to be brighter than they previously were, and the highlights to be less so, but you have not actually created an image with greater dynamic range than that which was available in the original image. Essentially, it's just applying a custom curve to the image prior to RAW conversion, rather than a standard one.

[dugcross]

This is the same as exposing 3 images with -1/0/+1 compensation.

NO this is not the same at all. Creating 3 images from one exposure does not create an HDR image. I don't care how you do it, it's only a tone-mapped image. You're not going to create information that is not there to begin with, with one exposure. If you don't have at least 3 separate exposures to begin with you DO NOT have a HDR image. Shadowblade is correct with his post.

[bunyarra]

NO this is not the same at all. Creating 3 images from one exposure does not create an HDR image. I don't care how you do it, it's only a tone-mapped image. You're not going to create information that is not there to begin with, with one exposure. If you don't have at least 3 separate exposures to begin with you DO NOT have a HDR image. Shadowblade is correct with his post.
ACR or D

Sorry - I don't care what the theory is, you still need to use ACR or DPP to create one pulled and one pushed image to retrieve the "hidden" RAW data. Try it for yourself. I do this every week.

And, for the record, 2 images is the minimum needed by your rules not 3 :)

I don't want this to get heated or turn into a pissing contest. If you disagree with my definition of HDR, please do so. Don't presume I am wrong in my definition. My pre-tone-mapped images created from a single RAW file have a greater DR than you can create from PS/DPP/ACR in one go no matter how you manipulate curves, fill, recover.

I extract the absolute maximum from each end of the RAW file and need to use the 3 images in my workflow as the image degrades significantly away from the edges when you start significantly pushing/pulling.

Sorry you are militant in your HDR viewpoint - I'm not. I just want to create the best possible output for my customers :) I will defend what I do as creating high dynamic range images - you will disagree - and we all move on.

There is no right way to do anything in photography or post processing these days - the days of dictated methodologies has long gone, thank goodness.

If we are disagreeing over one term's interpretation, life is too short to bother :cool:

[_GUI_]

Creating 3 images from one exposure does not create an HDR image. I don't care how you do it, it's only a tone-mapped image. You're not going to create information that is not there to begin with, with one exposure. If you don't have at least 3 separate exposures to begin with you DO NOT have a HDR image. Shadowblade is correct with his post.

You are not going to create more information by doing several replicas at different exposure values of the same image or RAW file, but... who says a single image cannot contain _all_ the information of the scene? :D

BTW with just 2 exposures cleverly chosen you can capture noise free HDR scenes of 12 stops of dynamic range with any DSLR at ISO100, no 3 separate exposures needed. The number 3 just comes from the usual camera bracketing (typ. {-2,0,+2}), but it is not a definition nor requirement itself.

Regards

[Shadowblade]

You are not going to create more information by doing several replicas at different exposure values of the same image or RAW file, but... who says a single image cannot contain _all_ the information of the scene? :D

BTW with just 2 exposures cleverly chosen you can capture noise free HDR scenes of 12 stops of dynamic range with any DSLR at ISO100, no 3 separate exposures needed. The number 3 just comes from the usual camera bracketing (typ. {-2,0,+2}), but it is not a definition nor requirement itself.

Regards

If it all comes from a single original, exposure, it's not HDR. If you're creating new 'exposures' by pushing and pulling in DPP/ACR, all you're really doing is adjusting the tone curves separately for each image, then combining them into a single, more complex tone curve - nothing you can't do by manually adjusting the curve for the RAW file in the first place.

HDR involves taking combining information available in one exposure but not the second, with information available in the second exposure but not the first, to create a new image containing information from both exposures. If you already had all the information in one exposure, you're not actually increasing the image's dynamic range - merely brightening the shadows and dimming the highlights.

[dugcross]

I don't want this to get heated or turn into a pissing contest. If you disagree with my definition of HDR, please do so. Don't presume I am wrong in my definition. My pre-tone-mapped images created from a single RAW file have a greater DR than you can create from PS/DPP/ACR in one go no matter how you manipulate curves, fill, recover.

I extract the absolute maximum from each end of the RAW file and need to use the 3 images in my workflow as the image degrades significantly away from the edges when you start significantly pushing/pulling.

Sorry you are militant in your HDR viewpoint - I'm not. I just want to create the best possible output for my customers :) I will defend what I do as creating high dynamic range images - you will disagree - and we all move on.




I'm not making an argument over this either but simply put this is not my opinion but fact. A lot of people think you can pull more information out of one exposure and make multiple exposures of it thinking you're getting more information out of it. But the whole point of HDR is taking more exposures to get more information than you can get out of one exposures.

"High dynamic range (HDR) images enable photographers to record a greater range of tonal detail than a given camera could capture in a single photo. This opens up a whole new set of lighting possibilities which one might have previously avoided—for purely technical reasons. The new "merge to HDR" feature of Photoshop allows the photographer to combine a series of bracketed exposures into a single image which encompasses the tonal detail of the entire series." from http://www.cambridgeincolour.com/tutorials/high-dynamic-range.htm

"High-dynamic-range photographs are generally achieved by capturing multiple standard photographs, often using exposure bracketing, and then merging them into an HDR image. Digital photographs are often encoded in a camera's raw image format, because 8 bit JPEG encoding doesn't offer enough values to allow fine transitions (and also introduces undesirable effects due to the lossy compression)." http://en.wikipedia.org

"HDR Photography allows photographers to capture a greater range of tonal detail than any camera could capture thru a single photo." http://photocritic.org/hdr-photography-how-to/

I'm not saying that you can't get an HDR "look" from a single image as you are doing but the fact is the result is a faux HDR image or tone-mapped image.


HDR involves taking combining information available in one exposure but not the second, with information available in the second exposure but not the first, to create a new image containing information from both exposures. If you already had all the information in one exposure, you're not actually increasing the image's dynamic range - merely brightening the shadows and dimming the highlights.

Once again, I totally agree with you Shadowblade

[_GUI_]

If you already had all the information in one exposure, you're not actually increasing the image's dynamic range - merely brightening the shadows and dimming the highlights.

If you already had all the information in one exposure, you don't need more exposures, and you can't increase the captured dynamic range because you already have it all. Full stop.

The reason for doing several shots in HDR imaging is because in most high dynamic range situations, with present digital cameras we cannot capture all the information in one exposure: to get non-blown highlights, shadows become too noisy, or to get well exposed shadows, the highlights get clipped. But if we could, just one shot would be fine. For example a HDR camera like the Fuji Super CCD (http://translate.google.com/translate?u=http%3A%2F%2Fwww.guillermoluijk.com%2F article%2Fsuperccd%2Findex.htm&langpair=es%7Cen&hl=EN&ie=UTF-8), or the use of a dense GND filter (http://en.wikipedia.org/wiki/Graduated_neutral_density_filter) in a sunset will allow us to capture a high dynamic range in a single shot.

HDR is about capturing all the information (noisefree shadows plus non-clipped highlights) of a high dynamic range scene, and tone mapping it so that it becomes well rendered in the final image. The number of shots needed to achieve it (one, or one million) doesn't play any role in the definition of HDR.

Find here one single RAW file (http://www.guillermoluijk.com/download/rawvirtual.dng) containing the 12 stops of information of a HDR scene. No more files needed because that single RAW file summarizes (http://www.guillermoluijk.com/article/virtualraw/index_en.htm) the content of two RAW files shot 4EV apart, that served well to capture all the dynamic range of the real scene. What you can achieve with that single file, is exactly the same to what you could achieve with the original 2 RAW files, no more, no less.

Regards

[bsmotril]

If you start with a RAW file that has recoverable detail in highlights and shadows, and use an HDR processor like Dynamic Photo HDR, then you can get an HDR from a single image. It will not have as large of a dynamic range as a multi exposure based HDR, but it will have more dynamic range than a psuedo tone mapped image from a single exposure. The key is the way the raw file is processed and how good your original exposure was. Shadow Noise certainly becomes a factor though at anything shot much above ISO400.

I agree with that. The problem is that too many photographers still think their one exposure image is a HDR image. They refuse to listen to anybody and think that just because at the very least they created 2 more images in lightroom or whatever from that single image and took it through Photomatix that the image is HDR when it is not. It is just a tone-mapped image! If you start out with a single image, no matter what you do to it, it will STILL not be a HDR image, it is tone-mapped. Tone-mapping an image does not magically make it HDR.

[Shadowblade]

If you start with a RAW file that has recoverable detail in highlights and shadows, and use an HDR processor like Dynamic Photo HDR, then you can get an HDR from a single image. It will not have as large of a dynamic range as a multi exposure based HDR, but it will have more dynamic range than a psuedo tone mapped image from a single exposure. The key is the way the raw file is processed and how good your original exposure was. Shadow Noise certainly becomes a factor though at anything shot much above ISO400.

There's no such thing as 'recoverable detail'. Detail is either there or it's not - it just may not be apparent in the default tone curves applied in RAW conversion. You can make the detail visible by applying a custom curve to brighten the shadows and darken the highlights - which is essentially what you're doing by using HDR software to combine several conversions of a single RAW file which have been pushed and pulled to different extents, only that you're doing it in another program other than the RAW converter. You haven't changed the image's dynamic range at all, merely the windowing. The extra 'detail' which you're seeing is really a consequence of local contrast enhancement in darker and lighter areas, which, if badly done, can result in 'haloing', as often seen in overblown tone-mapping.

[dugcross]

There's no such thing as 'recoverable detail'. Detail is either there or it's not - it just may not be apparent in the default tone curves applied in RAW conversion. You can make the detail visible by applying a custom curve to brighten the shadows and darken the highlights - which is essentially what you're doing by using HDR software to combine several conversions of a single RAW file which have been pushed and pulled to different extents, only that you're doing it in another program other than the RAW converter. You haven't changed the image's dynamic range at all, merely the windowing. The extra 'detail' which you're seeing is really a consequence of local contrast enhancement in darker and lighter areas, which, if badly done, can result in 'haloing', as often seen in overblown tone-mapping.

Agreed +1

[_GUI_]

There's no such thing as 'recoverable detail'. Detail is either there or it's not.

In the highlights this is fairly true: you kept the information with maximum detail, or you totally ruined it. We could still talk about partial saturation though (when just one or two of the RAW channels got clipped).

But in the deep shadows of a digital camera, being so taxative (detail is, or it is not) is not correct. When lifting the shadows noise appears, and wheter noise will prevent us to obtain detail from those areas or not will depend on several factors:

Scene: if there was detailed texture in the dark areas or not, and how much it was affected by noise
Application: the size of the final print copy or resized digital image (noise reduces when downsizing an image)
User: how demanding we are about visible noiseFor some scene/application/user, an image could be fine, while for another scene/application/user could have a too low quality.

Regards

[Shadowblade]

In the highlights this is fairly true: you kept the information with maximum detail, or you totally ruined it. We could still talk about partial saturation though (when just one or two of the RAW channels got clipped).

In which case, creating multiple images would not provide any more detail than adjusting curves (either using the sliders, or manually) during RAW conversion. The detail still isn't recoverable - it's either there (in the non-clipped channels) or not (clipped).


But in the deep shadows of a digital camera, being so taxative (detail is, or it is not) is not correct. When lifting the shadows noise appears, and wheter noise will prevent us to obtain detail from those areas or not will depend on several factors:
Scene: if there was detailed texture in the dark areas or not, and how much it was affected by noise
Application: the size of the final print copy or resized digital image (noise reduces when downsizing an image)
User: how demanding we are about visible noiseFor some scene/application/user, an image could be fine, while for another scene/application/user could have a too low quality.

Regards

That still doesn't make the detail 'recoverable' or 'non-recoverable' - it's either there (above the noise threshold) or not (indistinguishable from noise). The amount of detail visible above the noise threshold may vary, but what is there is there, and what isn't there cannot be recovered, since it isn't there in the first place.

[pcr1968]

I vote not having any info about the shot, and let the experts tell us wich ones are HDR and wich ones are not.

[_GUI_]

The detail still isn't recoverable - it's either there (in the non-clipped channels) or not (clipped).
I said I agree with your point of clipped vs non-clipped in the highlights.

That still doesn't make the detail 'recoverable' or 'non-recoverable' - it's either there (above the noise threshold) or not (indistinguishable from noise). The amount of detail visible above the noise threshold may vary, but what is there is there, and what isn't there cannot be recovered, since it isn't there in the first place.
I never talked about recovering, just pointed that in the shadows, the border between 'detail is there' and 'detail is not there' is not clear like in the highlights.

For example if you shoot a scene and plan to make a print 2m high, and also publish the image on your site at 800px, for the 2m high print the detail could 'not be there' because the noise was excessive to be able to recognise detailed textures at that size. But the noise reduction when downsizing to 800px for the web could end in an image where the detail 'is there' (of course all the detail you can expect in a 800px image).

In digital sensors, information in the highlights is mostly boolean (we captured it 100%, or we clipped it, no mean term), while information in the shadows is relative (it will depend on several factors such as scene/application/user to consider it valid or not).


I vote not having any info about the shot, and let the experts tell us wich ones are HDR and wich ones are not.

The only way to find out if an image is HDR is to know how much dynamic range the original scene had. To help on this, experience is a must. An indoor room with a window open outside to a sunny day will be HDR, with at least 12 stops of DR. A landscape with no sun in the frame and no strong shadows, will never be HDR.

This IS HDR (not very well tone mapped though, since the floor is more brilliant than the window that produced the reflection):

http://img66.imageshack.us/img66/2452/f1gv0.jpg


This is NOT HDR (but it is a specially ugly picture):

http://img413.imageshack.us/img413/4640/mierdimatixla5.jpg

[pcr1968]

I said I agree with your point of clipped vs non-clipped in the highlights.


I never talked about recovering, just pointed that in the shadows, the border between 'detail is there' and 'detail is not there' is not clear like in the highlights.

For example if you shoot a scene and plan to make a print 2m high, and also publish the image on your site at 800px, for the 2m high print the detail could 'not be there' because the noise was excessive to be able to recognise detailed textures at that size. But the noise reduction when downsizing to 800px for the web could end in an image where the detail 'is there' (of course all the detail you can expect in a 800px image).

In digital sensors, information in the highlights is mostly boolean (we captured it 100%, or we clipped it, no mean term), while information in the shadows is relative (it will depend on several factors such as scene/application/user to consider it valid or not).




The only way to find out if an image is HDR is to know how much dynamic range the original scene had. To help on this, experience is a must. An indoor room with a window open outside to a sunny day will be HDR, with at least 12 stops of DR. A landscape with no sun in the frame and no strong shadows, will never be HDR.

This IS HDR (not very well tone mapped though, since the floor is more brilliant than the window that produced the reflection):

http://img66.imageshack.us/img66/2452/f1gv0.jpg


This is NOT HDR (but it is a specially ugly picture):


http://img413.imageshack.us/img413/4640/mierdimatixla5.jpg

Sorry Mate Mine comment was sarcasm, I coudn't care less how many pics people take. If the final result is a nice pic, who cares. I stopped posting to HDR forums due to how much whinging and whining there is. Not just on this site but most other forums as well.