Talk:High dynamic range imaging

From Wikipedia, the free encyclopedia

You have new messages (last change).

Contents

[edit] History

This article is written as though HDR is new. It's not. Photographers were using similar methods to extend dynamic range even on film, and photographers have been manipulating digital images in this way ever since the birth of digital imaging. Some people working on articles like this must be very young indeed if they really and truly think there is anything new here besides an Official Name and Official Abbreviation.

Also, the examples provided in the article are quite poor. There is a tendency for newbies to this technique to overdo it. Agateller 15:52, 7 August 2006 (UTC)

I've just gone through and removed most overt references to photography, as well as the gallery at the end. I've also tried to explain where HDR images come frome (i.e rendering or from multiple exposures) and an intro to tone mapping.
Unfortunately, you're falling into the same trap as the others. HDR(I) is not a photographic method. What you're describing is merging of two or more exposures, and I think photographers have been doing that with film/paper long before digital photography came along.
HDR is simply about storing images in such a way as to allow a very large range of values to be stored, and for those values to represent real-world measurements instead of simply how bright a pixel should be on a monitor. Those pixel values can be produced from combining photos or they can be computed using some rendering algorithm. But here's the important part - those "HDR photos" are not HDR. The intermediate may have been, in which case they've been run through a horrible tone mapping algorithm trying to show too much dynamic range. But the photos we are looking at are not HDR.
--Imroy 17:16, 7 August 2006 (UTC)
The only utility to HDR is to compensate for limited dynamic range in image capture, which is really only a problem in photographic systems. It's just another TLA for something that has been around for ages. Sometimes geeks latch onto something that has been around for ages and then like to pretend that they've invented it themselves by giving it a name. Agateller 07:41, 8 August 2006 (UTC)
You're missing my point. Let me put it in bold: HDR is not necessarily about photography. Yes, there is a fad at the moment involving combining several photos, and they call it "HDR". But HDR(I) has been around since the mid-to-late 80's and has other applications. I'm also annoyed at all these horrible photos being labelled "HDR", but I'm not blaiming HDR. --Imroy 15:59, 8 August 2006 (UTC)
I'm all for historical accuracy, but we also have to be relevant today. Today, the term HDR is increasingly being applied differently. I claim that a large number, perhaps even a majority, of people today are looking for more information on the concept of taking multiple exposures in digital photographs for the purpose of bringing out a higher range of color. I think the revisions of this article after 15:49, 7 August 2006 fail to be effective for people who are looking for more information on that concept. I'm proposing, in an effort to be effective for more people, to have portions of old content (including the older photo of the church) moved into an article titled "High dynamic range (digital?) photography," with each article referencing each other appropriately.
The term "HDR" as it is used by many is actually a sub-set of HDR(I). For example, even though most people use digital point-and-shoot cameras nowadays, and refer to "taking photos" in everyday speech, the photography article deals with the history of photography and the full range of the art. Similarly, using Photoshop or any other package to "make a HDR" is just an application of HDR and this article should not restrict itself to the narrow definition that's come into recent use. While I accept that language is alive and changes, most people aren't experts, and they also use shortened or abbreviated terms in common usage. I have no problem of (re-)adding those sorts of photos into the article, as long as it is made clear that they are simply one application of HDR. Before my edits, the article (as I interpreted it) appeared to imply that those photos were HDR and that was it. A few (notably Zimberoff, see below) seemed convinced that HDR was only about photos and could be achieved by simply merging different exposures. Hopefully my edits and example images demonstrate that HDR images aren't just about photos and can be used for more interesting things (e.g lightprobes). --Imroy 07:46, 26 August 2006 (UTC)

[edit] Applications

I'm not sure I see the application for HDR. (I've had a look at HDRShop but wasn't much wiser.) Is it used to expand the dynamic range for print? I presume that typical PC displays would not show a high dynamic range. -DRB

A: No, LCD or CRT displays can not show HDR, but at least experimental HDR displays are already available (see this [link]). Photographers take HDR photographs of difficult scenes (like landscapes with huge luminance range), which they later manipulate in Photoshop using contrast, brightness, dodging and burning tools, to make them suitable for print. Tone mapping algorithms are supposed to do the same thing automatically.

HDR image formats are just an application of the general principal that you should keep information as long as possible. It's very easy to generate images with enormous dynamic range (especially with artificial rendering); a HDR output format allows you to keep an exact representation such an image for later post-processing. This is different from traditional image formats, which are optimized for the limitations of display devices, and can't represent many images exactly -- e.g. if you discover after your 5 day ray-trace is done that you misjudged the lighting, you can edit a HDR output image to adjust it, without losing anything, but if you used a more limited output format, you may have to simply redo the rendering. HDR images also make it much more practical to do post-processing of images.

HDRI images are used extensively in 3d modeling as they can be used as realistic light sources. This is especially important when integrating CG effects into live action cinema as the CG model can be lit by the real scene rather than resorting to modeling the scene with approximation of the main light sources. As mentioned below its is also being introduced into gaming.--Pypex 23:39, 9 November 2005 (UTC)

[edit] HL2

As I recall Valve released a demo called "Half Life 2: The Lost Coast" which showed off HDR. Shouldnt this be included in the article? (I dont know the detsils but Im interested)

[edit] Merge "High dynamic range rendering" to this article

It seems both articles have an almost similar title but different contents. Given High dynamic range imaging refers to broader range of fields, I found it would be more appropriate to merge contents from High dynamic range rendering to this page. Comments? ╫ 25 ring-a-ding 19:51, 1 October 2005 (UTC) ╫

Yeah, they should probably be merged. Thunderbrand 13:16, 6 October 2005 (UTC)
I for once agree with the merge, the HDRR is rather incoherently written IMHO, and HDRI is by far the more important aspect of the subject.--Pypex 23:44, 9 November 2005 (UTC)
OPPOSE: I am pretty sure that high dynamic range imaging and high dynamic range rendering are different; mostly related to the differences in ray-tracing and real-time polygonal rendering.
HDRI is a non-real time render of HDR, while HDRR is a real time render. But I think the objectives between HDRI and HDRR are different. HDRI attempts to preserve and record lighting data from photographs using say another 32-bits for lighting information from different exposures of the camera (since we don't really have an HDR camera). HDRR attempts to simulate the highs and lows of any particular scene by adjusting the brightness of any given object by the monitor's brightness capabilities. Lighting information isn't exactly accurate with the 24-bit color model used today, since there's only 256 shades of any given color, meaning the contrast ratio is stuck at 256:1. Both HDRI and HDRR technically should benefit from an HDR display though. Given this case, I think it's appropriate to say that HDRI is a photography related subject while HDRR is a computer graphic related subject.
OPPOSE: It seems to be clear HDR itself contains, or includes, both HDR imaging (HDRI) and HDR rendering (HDRR). Even though they are related, imaging is about caputring, processing, viewing and printing of HDR, while HDRR is focusses on rendering virtual scenes by using more than 8 bits of color definition. The goals are very different, only the means are similar, and so also the content is very different.
But there is at least two elements which needed fixing: The HDRI article talksed for a bit more than a whole paragraph about HDRR. This detracts because it belongs to HDRR. HDRR should be seen as a specialized sub-article of HDR or HDRI. I moved these to HDRR and added a short explanation to the link to the HDR rendering article.
Clearly, don't merge, but specialize. Eventually, a common parent article on HDRI and HDRR named just HDR may be created to which topics which might concern both HDR photo imaging and virtual scene rendering, which is the basic explanation of highy dynamic range and a deeper discussion about displaying a higher dynamic range. I think this is the ultimate way to go, because both articles will grow over time and things need to be carefully placed then: Common things on a common place, different things in different articles.
Lastly, the Link list of this article needs to be cleanup or better, structuring of the external links list. It was a huge linear list without any classification, so it looked at this link list and the one at High dynamic range rendering, and removed double links, but still, two links need to be reviewed and categoried or deleted --NoSoftwarePatents 07:46, 20 January 2006 (UTC)

[edit] Link removal

I'm removing a few of the external links at the end of this article. In case anyone wants to resurect any, here are the links with my notes.

I also gave a name to http://www.rendermania.com/HDRI/. I remember finding it myself a while ago and it is reasonably informative.

Imroy 16:43, 3 March 2006 (UTC)

[edit] Added more example exposures

I've added some example exposures of the HDRI. I can't get the positioning looking nice though. If someone else wants to have a go then be my guest. —The preceding unsigned comment was added by Deanpemberton (talk • contribs).

[edit] Flickr HDR group?

I've just removed links to the Flickr "HDR" group for the third time. I'm just wondering what other people think of this external link (http://www.flickr.com/groups/hdr/).

My objection is that the photos presented are not actually HDR images. They're created from a set of bracketed exposures, apparently all with PhotoShop CS2, and then badly tone-mapped into JPEG's. Yes, a HDR image is produced at some point, but I've signed up for Flickr and still nowhere is there an option to download the photos in a HDR format.

On top of that, I really, really don't like the tone mapping used. For example, User:Deanpemberton's recent addition of Image:Old saint pauls 2.jpg. It's just so flat! Take the image into an editor like The GIMP and look at the histogram. It's centred around 50% and decreases out from there. That's not natural. Tone mapping was originally meant to simulate a photo-like "exposure" from a HDR render e.g Radiance. The tone mapping employed here is something else entirely. It's trying to compress the whole dynamic range of an image into an LDR JPEG. The result is flat and almost cartoon-like.

I guess what I'm trying to explain is that the Flickr group does not represent HDR imaging, and perhaps even leads to a misunderstanding of that "HDR" means. This can already be seen over at High dynamic range rendering, which appears to be mostly written by gamers. To many there, "HDR" involves blooming or other effects that can be faked without HDR. Please don't mislead any more people about what HDR is. Imroy 02:48, 23 March 2006 (UTC)

[edit] So what is HDR?

I think you're right in general that the flickr HDR group is misleading about what HDR is, but I disagree with the two reasons you gave.

Except for just a couple of very expensive pro cameras, taking bracketed exposures *is* currently the way to capture HDR images! Debevec published _the_ paper on how to do this.

As far as tone mapping goes, I fail to see why bad tone mapping makes something not HDR. The images on flickr are HDR images with mostly bad tone mapping. I don't mind keeping the flickr link off this page- it's certainly not defining or leading HDR development in any way. But you might want to consider using the flickr link as motivation for why people are actively researching tone mapping-- because it actually takes some effort and it's fairly easy to get it wrong even with Photoshop.

Maybe there should be a section of external links that show commodity application of different HDR techniques and approximations- flickr and Photoshop could certainly go there, along with Half Life and whatever other junk people want linked.

In general it would help to be clear about what HDR is, rather than asserting what it isn't. Both bracketed exposures and tone mapping have historically been an integral part of the subject of High Dynamic Range Imaging. If we divorce those things completely, all we're left with to talk about is "more than 8 bits per channel". Yay. —The preceding unsigned comment was added by 63.166.224.67 (talk • contribs).

OK, firstly - the *whole idea* of tone-mapping is to make something "not HDR". It's to convert from HDR to LDR for display. And if we're going to have tone-mapped images, I would prefer to show the effects of different algorithms on the same image - to show that the LDR images are just representations of the same HDR.
The problem with trying to show HDR is that no web browser (to my knowledge) can currently display HDR images. So we're forced to make LDR JPEG's that do display in web browsers. And I think this is where it gets tricky. I think instead of using a tone-mapping algorithm to compress a huge dynamic range into a flat, lifeless image, we should try to show that the HDR images have a wide range of light levels (dynamic range). One way is by showing mutliple exposures of the same image. But then if it's a photo, people will just say "you took different exposures, big deal!". We need some way of demonstrating that the HDR image has all of these "exposures" in one image, and why it's so useful.
Perhaps one way is with a rendering using a light probe, like Debevec's "rendering with natural light". Demonstrate (somehow) that the light probe (a HDR environment map) is actually lighting the scene, with no other sources of light. Or is that too technical? I know it's been used for several SFX shots in hollywood movies over the last few years. In fact, it's probably quite common now.
Another demonstration could show how various effects (e.g motion blur) are more realistic when done in a high-dynamic colour space. Any other ideas?
Imroy 00:12, 18 April 2006 (UTC)
> We need some way of demonstrating that the HDR image has all of these "exposures" in one image, and why it's so useful.
Well, yeah on the web we can show bracketed exposures before and bracketed exposures after, or we can show bracketed exposures before and a tone-mapped image after. For wikipedia's purposes it seems easier to motivate HDR for the lay-person using the latter, IMO. Also, you could always link to a java HDR viewer (such as http://webuser.hs-furtwangen.de/~dersch/hdr/hdrtest1.html). But that's just interactive tone-mapping, so doesn't exactly solve your issue. But tone mapping is integral to HDR's history, and aside from that you don't need to have an actual elephant in the Wikipedia in order to write an article about one.
Tone mapping came before the HDR file format, and the original reason for inventing an HDR file format was to be able to apply different tone mappings without re-rendering, and to research different tone mapping techniques. You're right in a strict technical sense that tone-mapping is there to make something not HDR, but saying that doesn't exactly paint an accurate portrait of HDR techniques or the historical reasons for having them, IMO.
Your suggestion for demonstrating multiple tone mappings is a good one- either this article or the tone mapping one should do that. Motion blur is a worthy example as well, possibly most appropriate in the HDRR article...
Anyway, as far as your original question goes, don't worry too much about flickr, I think it's just fine to keep that link off this article.
63.166.224.67 20:32, 18 April 2006 (UTC)

Maybe a section in the article about possible "artistic abuses" of the HDR concept would solve the problem? It is true that the original purpose of HDR formats was to capture and represent a wide range of physically accurate radiative data. But the current trend in certain circles happens to be to use it as just another effect filter. There's not really a point in ignoring this development. It will be better to name it for what it is, mention a few notable(!) examples (possibly even by pointing to the Flickr group), and get over it. --Latebird 13:08, 25 April 2006 (UTC)

[edit] Photographic related links?

Is this article solely intended to be about HDRi in CG/Rendering environment or may it be designated to add information about photographic concerns with HDR and tone mapping? -- I would like to add http://www.hdrsoft.com/resources/dri.html as external link, beacuse it is a FAQ about HDR and photographic concerns. What du you think about this approach?

That link looks good. It's not trying to sell a product (that I can see) and provides some real infomation. Of course, it would be best if that sort of information could be incorporated in the Wikipedia article (don't violate WP:COPY), but I wouldn't have any problem with that link being added. Imroy 18:13, 11 April 2006 (UTC)
"It's not trying to sell a product" - and that "purchase" button on top of the page? If there are links to sites offering Photomatix and HDRShop I would like to see here also links to easyHDR website (http://www.astrofoto.pl/easyHDR) and FDRTools (http://www.fdrtools.com/front_e.php). [not a Wikipedia editor] 30 July 2006
On March 30th I added an external link referring to the homepage of FDRTools. Imroy responded with:
"Please do not add commercial links (or links to your own private websites) to Wikipedia. Wikipedia is not a vehicle for advertising or a mere collection of external links. You are, however, encouraged to add content instead of links to the encyclopedia. See the welcome page to learn more. Thanks. Imroy 10:09, 30 March 2006 (UTC)"
I am heavily surprised to see now links to the homepage of Photomatix. May be Photomatix is now open source software without the need to pay some fee? FDRTools 15:33, 07 September 2006 (UTC)

[edit] the example

shouldn't the heading state that those are tone mapped aswell?

[edit] Creation/Composition

This article needs more information on how these images are created. For example, does it require that you have all six shots, or can the effect be accomplished with fewer exposure samples? --DDG 21:37, 2 June 2006 (UTC)

[edit] two different subjects

there should be two articles, one for photoshop hdr/tonemapped image processing (taking several exposures and combining them into one image) and one for cg related hdr (using an hdr image dynamically for lighting, etc). they are related, but different enough to warrant seperating. 70.30.194.144 12:12, 21 June 2006 (UTC)

Photoshop may be a special case, more targeted at artistic uses. But other than that there's no inherent difference between photographic-composite and computed images. In fact, the two approaches are often combined. Composite images are used for environment mapping within computed images, and computed images are merged with composite images for presentation purposes (architectural mock-ups). As long as we're talking about scientifically "correct" uses, the method of creation is determined by practical requirements, not because there are different principles involved. --Latebird 16:52, 21 June 2006 (UTC)

[edit] Comment in article

It is inconceivable that any one person should be credited with pioneering the use of so-called "high dynamic range imaging." Images using this technique have been created by analog means for decades, quite easily at that, by means of a process called image "sandwiching." Sandwiching essentially achieves the same results as HDRI by, well, sandwiching and rephotographing several film transparencies captured at various exposures to combine the effects of various exposures on a new transparency. The final result is the new "original." It is supposed to more closely resemble how the brain perceives what the eyes actually see. The transposition of this technique to digital capture does not represent a new paradigm, any more so than do the techniques of electronically dodging or burning an image as compared to performing these techniques in the darkroom.--Zimberoff 00:48, 4 August 2006 (UTC)--Zimberoff 00:48, 4 August 2006 (UTC)

This was in the article. I'm removing it and pasting it here. --64.24.181.159 02:27, 4 August 2006 (UTC)

[edit] Rearrangement

Seeing Zimberoff's misguided contribution being removed finally made me do something for this article. I rendered a simple scene, manipulated the image, uploaded them, and added them. I hope people find the images and text useful.

Now, for a while I've been meaning to rearrange this and a number of related articles. Here's my plan:

  • Merge the common info into this HDRI article. Make it a more general 'overview' type article, with links to the more specific and related articles.
  • Add some more example images and explainations.
  • Rename the High dynamic range rendering article to something like e.g Realtime HDRI rendering. Make it less game-specific.
  • Move some content to the tone mapping article, in particular the ghastly Photoshop CS2 tone-mapped photos.

What do people think of these ideas? --Imroy 16:35, 4 August 2006 (UTC)

[edit] Displays

Are there displays that are capable of showing HDR pictures - at least in a better manner than current standard consumer displays do? --Abdull 17:08, 11 August 2006 (UTC)


[edit] Dynamic Range vs. Radiance/Tone

I would suggest that this article confuses "dynamic range" with the concept of "radiance vs. tone". Dynamic range refers to the precision and range used to store values, regardless of whether the values represent radiance information or tone information. In contrast, radiance measures amount of light and can go from zero to infinity while tone measures amount of pigment or light emited from a screen, and goes from zero to a finite maximum (say one, for simplicity). Radiance can come from rendering light in computer graphics, directly reading a digital camera's imager, or cleverly combining several photographs taken at different exposures. Tone is what you see in a photographic print or on a screen.

Tone mapping is just the process of converting radiance information to tone information, regardless of dynamic range or precision. Even low dynamic range radiance information must be logarithmically tone mapped to simulate film's respone to light and to give a natural looking photograph (if that is the desired result). Thus, to say "a high dynamic range image is usually tone mapped" is confusing, because all radiance information is usually tone mapped.

Perhaps this article should focus only on the fact that HDRI permits the storing of precise information, be it radiance or tone, across a large range from very dark to very light without loss of information. References to tone mapping could be restated as "HDRI permits precise radiance values to captured/calculated and used in innovative tone mapping algorithms that reveal detail in highlights and shadows, detail that is often lost in simpler tone mapping algorithms. HDRI also permits storing and manipulating precise tone values, which is important in digital image manipulation to prevent gradual degradation of the image as operations are performed."


[edit] Poor example image

There is an example iamge near the top of this entry. It is a nice picture and someone even though of making it a featured wiki image. I has nothing to do with HDRI since it is a jpg. This could be edited and replaced with any image and be just as valid.

HDR images cannot be displayed on a regular screen, or print. They need to be downsampled. I will reword the captions. HighInBC (Need help? Ask me) 01:36, 3 November 2006 (UTC)
Retrieved from "http://en.wikipedia.org../../../h/i/g/Talk%7EHigh_dynamic_range_imaging_6360.html"