2

Is "exposure fusion" distinct from HDR compositing techniques, or is it essentially a marketing term for a specific HDR process. Or is it just a synonym for HDR?

Or, to put this in "meta" terms, should exposure fusion be treated as a distinct tag from HDR, or should it get lumped in with ?

Michael C
  • 175,039
  • 10
  • 209
  • 561
feetwet
  • 3,603
  • 2
  • 26
  • 60

4 Answers4

1

"Exposure Fusion" just refers to a certain type of the tone mapping of an HDR image. It doesn't include all HDR tone mapping algorithms, and is rightly tagged as "HDR".

JenSCDC
  • 2,339
  • 14
  • 23
  • 2
    This sounds right to me. I have this impression that there's a formal logic that would show that all HDR processes are transformations of a set of aligned images into a single image of the same color depth, and one can show that "exposure fusion" is just a special case of such suitably characterized transformations. – feetwet Sep 07 '14 at 02:39
  • Bingo. You said better than I did. – JenSCDC Sep 07 '14 at 08:00
0

HDR and exposure fusion (as done by a tool like enfuse) are two different ways of attacking the same problem, but exposure fusion is NOT an HDR technique and does not require HDR file formats or tone mapping.

Enfuse's main algorithm does not involve remapping tones along a larger dynamic range, but selects and weights values from the original pixel data of member images, based upon three criteria. From the enfuse website:

The basic idea is that pixels in the input images are weighted according to qualities such as proper exposure, good contrast, and high saturation. These weights determine how much a given pixel will contribute to the final image. A Burt & Adelson multiresolution spline blender is used to combine the images according to the weights. The multiresolution blending ensures that transitions between regions where different images contribute more heavily are difficult to see.

Instead of mapping out all the tones in the member images along a high dynamic range, and then tonemapping back down to an LDR format, weighted averaging is done on the original member pixel values to create the final image pixel values. Essentially every value in the final image is going to be within the range of the original member image values, not necessarily the result you get with HDR and tonemapping, which can manipulate the luminance and saturation of pixels outside the original range.

In more manual methods of exposure fusion, simple masking allows for pixel selection from member images into the final image--again, no mapping to an HDR file values or tonemapping is required. You can also think of enfuse as masking and selecting at the individual pixel level.

See also:

inkista
  • 52,065
  • 10
  • 88
  • 161
  • 1
    Your definition of High Dynamic Range Imaging is awful narrow. HDR imaging goes all the way back to the 1850s using a variety of techniques to expand dynamic range beyond the conventional capability of the time. In my view both exposure fusion and 32-bit floating point files are slightly different forms of HDR. – Michael C Sep 07 '14 at 06:54
  • 1
    Yes, it's clear we've got some contention of terminology going on here. But I think it's also clear we have two different algorithms and that while broadening the definition of HDR could include both, that does not make them the same algorithm, or the same thing by two different names. By the broadness of definition others are insisting on, one could also call reverse-S Curves adjustments or fill flash HDR. – inkista Sep 07 '14 at 08:14
  • @MichaelClark if you refer to this http://en.m.wikipedia.org/wiki/High-dynamic-range_imaging#Mid-nineteenth_century , this is exactly what we are talking about. it is exposure fusion, not HDR. you confusion may come from the fact that it is mentioned in the history of HDR section, as precursor of HDR. – ths Sep 07 '14 at 13:29
  • 1
    Why does HDR depend on the existence of a file with a larger than normal dynamic range? Isn't HDR a process where one takes multiple images with different exposures and turns them into a LDR? – JenSCDC Sep 07 '14 at 15:51
  • Because to digitally represent an HDR, you can't do it with an LDR file format--the possible numerical value range is too small. That's why you need a new file format with a larger bit-depth. JPEGs, for example, typically use 8-bit or 16-bit depth per channel. Some HDR file formats use 32-bit floating point values. – inkista Sep 07 '14 at 16:05
  • @inkista which are then tone mapped back down into 8-bits. So by your definition a single 14-bit RAW file mapped into 8-bits with the right curves is an HDR image? – Michael C Sep 08 '14 at 02:52
  • The confusion with the separate articles at wikipedia is precisely the same confusion here: Some who want to restrict the definition of a term that has been around much longer than digital imaging has been around because they haven't been around longer than digital imaging and others who have been around before digital imaging in its current form existed and already used the term High Dynamic Range Imaging to describe techniques that have been used and developed for 150 years! The former group prevailed in the Exposure Fusion article, the later group prevailed in HDR Imaging. – Michael C Sep 08 '14 at 02:56
  • @inkista Re: reverse S-curves. YES, that is a form of HDR imaging! Re: fill flash. NO. That is a way of reducing the dynamic range of a scene before capturing it. – Michael C Sep 08 '14 at 02:59
  • there is no difference between those two articles. the history of hdr section describes techniques that where used before hdr became possible. if you insist on using that word for any method of reducing an images dynamic range, i urge you to think about the literal meaning of the letters HDR. you can't possibly believe that an image with lowered contrast earns the name High Dynamic Range? if you want, you may call exposure fusion a tone mapping algorithm, inexact as it may be. – ths Sep 08 '14 at 21:11
  • @Speising The term High Dynamic Range Imaging has been around long before digital imaging. It was and continues to be used to refer to imaging of scenes that have a wide dynamic range that can't be conventionally reproduced. Just because many first heard the term in reference to your narrower understanding does not mean it is the only usage the term has ever had. – Michael C Sep 08 '14 at 22:36
  • @speising You are reading your own preconception into the wikipedia article on HDRI. Read the end of the first sentence: "...using standard digital imaging or photographic techniques." Read the lead sentence of the fourth paragraph: "The two primary types of HDR images are computer renderings and images resulting from merging multiple low-dynamic-range (LDR)[5] or standard-dynamic-range (SDR)[6] photographs." Paragraphs 2 and 3 of the introduction can be applied equally to both digital and no-digital forms of photography! (bold emphasis added) – Michael C Sep 08 '14 at 22:44
  • As to what you refer to as "...techniques that where (sic) used before hdr became possible." - The entire section is under the heading History of HDR Photography, not a heading such as Precursors to HDR Photography! – Michael C Sep 08 '14 at 22:45
  • With regard to your insistence that exposure fusion is not true HDR because it does not create a large floating point file, read the first two paragraphs of the Late-twentieth century heading. The HDR video image referenced in the first paragraph (as HDR) has less in common with the Modern HDR Imaging that uses a "completely different approach" of the second paragraph than contemporary exposure fusion does. Yet both are called HDR! So if what is described the first paragraph is HDR, then at the very least so is the approach referred to as Exposure Fusion. – Michael C Sep 08 '14 at 22:51
  • you seem to be under the misconception that HDR video doe not have an extended dynamic range. it does. HDR video does not mean "tone mapped video". you need tone mapping to produce something viewable on ordinary output devices. yes, you can merge LDR or SDR images, scanned analog or digital to HDR. but HDR still means "a higher dynamic range than possible with standard methods". a tone mapped image does not have a particularly high dynamic range, i don't know how that could be confusing. – ths Sep 09 '14 at 19:46
  • You seem to suffer under the misconception that High Dynamic Range is a term that refers to the form of an intermediate stage of an HDR image. It is not. It is a term that refers to the *Scene. An HDR* image, as a finished image viewed on a standard monitor or as a print, has no higher dynamic range than any other photograph. What makes it an HDR image is that it manages to represent (with the limited DR available) more highlights and shadows from a scene that has much higher DR than can be conventionally represented on a monitor or in a photo. – Michael C Sep 10 '14 at 06:02
  • You also still seem to think HDR Imaging can only be done digitally. While digital is one way to do it, you can also do it with analog film in a darkroom without ever having to scan anything. It simply means you are using a technique that allows you to represent a scene with extended DR using a medium that can not display the same wide DR, and you are doing so in a way that allows details, both in shadows and highlights, from the HDR scene to be displayed in the LDR medium. – Michael C Sep 10 '14 at 06:08
0

Whether Exposure Fusion is a form of High Dynamic Range Imaging (HDR) or not depends on how you define HDR.

If you have a broad definition of HDR Imaging that includes techniques that have been around since the 1850s when Gustave Le Gray first used parts of two differently exposed images to create photos of seascapes, then Exposure Fusion is a form of HDR. If you include the tone mapping done in the darkroom using dodging and burning when printing from negatives that Ansel Adams raised to an high art form in the mid-20th century, then Exposure Fusion is a form of HDR.

If you choose to restrict the term High Dynamic Range Imaging (HDR) to the technique developed in the late 20th and early 21st centuries to deal with the limited dynamic range of digital cameras compared to existing films at the time by producing a 32-bit floating point image file that is then tone mapped into an 8-bit image then Exposure Fusion is not a from of HDR.

See http://en.wikipedia.org/wiki/High-dynamic-range_imaging for more on the history of High Dynamic Range Imaging.

Michael C
  • 175,039
  • 10
  • 209
  • 561
  • 1
    that's a great answer - often, I tend to hyper focus on the little post processing world on my computer screen. Your thoughtful answer helps to remind me of the long heritage of photography – B Shaw Sep 07 '14 at 08:34
  • Yes, thanks for the excellent answer! Would you care to address the "meta" question of whether, in current practical use, "HDR" should be expected to include "exposure fusion"? Or do photographers now generally expect that "HDR", including in-camera HDR, specifically refers to the process of compositing a higher-depth image and then tone-mapping it back to an 8-bit image? – feetwet Sep 07 '14 at 15:14
  • I disagree about Ansel Adams. Tone mapping is not the same thing as HDR. – JenSCDC Sep 07 '14 at 16:00
  • @AndyBlankertz It is if you define HDR as High Dynamic Range Imaging. Dodging and burning was all about getting the additional dynamic range the negative could contain into prints on paper that couldn't contain the same dynamic range. Read Adams' trilogy The Camera, The Negative, and The Print. The zone system he developed was entirely about squeezing the wider dynamic range in the scenes he photographed into the narrower dynamic range of the photographic paper on which he printed. – Michael C Sep 08 '14 at 02:43
-1

Exposure fusion means blending of differently exposed images. This resulting image will not have an expanded dynamic range, but is basically a composite of the dark parts of the overexposed image, and the light parts of the underexposed one.
HDR on the other hand is an image which was created by combining all the data of the source images, resulting in an image with expanded ("high") dynamic range, potentially far more than can be displayed by any devices or papers.
Those HDR images can then be tone mapped to a narrower range so it can actually be viewed; this tonemapping result is not an HDR image anymore, although it is often erroneously called so.

So, no, they are not the same, but different techniques with similar applications.

ths
  • 7,161
  • 1
  • 11
  • 25
  • 1
    I'm sorry, but you're wrong. You can start out with the same sequence of bracketed exposures, combine them into one HDR image, and then tone map it as you like- and "Exposure Fusion" is one of the available algorithms. Do you understand the difference between a technique and an algorithm? You don't seem to. – JenSCDC Sep 06 '14 at 23:59
  • 1
    i most certainly do. maybe you should take a look at the relevant wikipedia entry: http://en.m.wikipedia.org/wiki/Exposure_Fusion just because some tone mapping program chooses to call one of its options so the much older technique doesn't disappear. – ths Sep 07 '14 at 00:05
  • 2
    Nonsense, as has already been pointed out. – user28116 Sep 07 '14 at 00:17
  • 1
    If an algorithm takes a high dynamic range image and maps its value to a LDR image, it is by definition a tone mapping algorithm. – JenSCDC Sep 07 '14 at 00:53
  • 2
    @speising: Are you just making a semantic argument? Of course, 16- or 32-bit images literally have a higher dynamic range than traditional 8-bit images. And the product of HDR processing is an 8-bit image with no more theoretical color depth than an unenhanced image. But when photographers talk about "HDR" are they not talking about images that have been processed to incorporate more dynamic range than a sensor can capture, regardless of whether they explicitly pass through a stage in which higher color depth is stored? – feetwet Sep 07 '14 at 01:16
  • 1
    Somehow I don't think we're going to get through to him. – JenSCDC Sep 07 '14 at 01:45
  • @speising The "older technique* to which you refer was developed a full 150 years after the advent of the first HDR Imaging techniques in the 1850s. It doesn't have to be a 32-bit floating point file tone mapped into 8-bits to be an HDR image. – Michael C Sep 07 '14 at 07:11
  • @MichaelClark citation needed. also, those who think they know better, it would be good of you to correct the wikipedia article, including references, of course. – ths Sep 07 '14 at 12:42
  • Source: the following Wikipedia article. Section: History of HDR Photography, Paragraphs: Mid-nineteenth Century and Mid-twentieth century.

    http://en.wikipedia.org/wiki/High-dynamic-range_imaging

     – Michael C Sep 08 '14 at 02:47
  • yes. now apply reading comprehension. you might find mention of pieces of string in an article about the history of computing, that doesn't mean pieces of string are computers. – ths Sep 08 '14 at 05:16
  • That depends on whom you ask. An historian of mathematics will certainly refer to an ancient abacus as an early analog computer and may well also refer to devises used to compute sums, differences, products, and quotients as such. You miss the mark if you insist that only a digital computer can be defined as a computer when the word itself was around and had uses prior to the advent of digital computing. Notice the earliest recorded use of the word in English was in 1646 and referred to a person who did calculations. http://www.merriam-webster.com/dictionary/computer – Michael C Sep 08 '14 at 23:02
  • And http://www.oxforddictionaries.com/definition/english/computer – Michael C Sep 08 '14 at 23:05
  • From http://en.wikipedia.org/wiki/Computer- "The first use of the word “computer” was recorded in 1613 in a book called “The yong mans gleanings” by English writer Richard Braithwait I haue read the truest computer of Times, and the best Arithmetician that euer breathed, and he reduceth thy dayes into a short number. It referred to a person who carried out calculations, or computations, and the word continued with the same meaning until the middle of the 20th century. From the end of the 19th century the word began to take on its more familiar meaning, a machine that carries out computations." – Michael C Sep 08 '14 at 23:09
  • exactly. that's what i meant. – ths Sep 09 '14 at 19:39