12
\$\begingroup\$

When I my camera takes RAW+JPEG with bracketed exposure (-1/0/+1 EV), it seems to me that the only difference that could logically exist between the RAW images is the shutter speed.

If I assume that the default 0 EV has reasonable exposure, does this mean that the -1 and +1 RAWs are more or less redundant? Or do they really have a significant amount of extra information that would be necessary if (say) I wanted to create HDR images down the road, or do something else nontrivial?

(Basically, I'm trying to see if I should really keep around all 3 RAW files, since they're so huge.)

\$\endgroup\$
3
  • 1
    \$\begingroup\$ It can be said that once you're done processing RAWs, there is no point in keeping them at all. \$\endgroup\$
    – Agent_L
    Aug 10, 2015 at 8:18
  • 4
    \$\begingroup\$ @Agent_L define "once you're done"... \$\endgroup\$
    – Chris H
    Aug 10, 2015 at 9:15
  • 2
    \$\begingroup\$ @ChrisH Exactly :) \$\endgroup\$
    – Agent_L
    Aug 10, 2015 at 11:52

8 Answers 8

15
\$\begingroup\$

The point of bracketing in the context you described is to come up with a higher dynamic range that the one you can get with a single picture because you want or need to do it.

Even if RAW has more bits per channel than JPEG, it still have a limited dynamic range. So depending of the scene you have in front of you, you will want or need to take pictures with different exposures. Moreover, it isn't all about RAW vs JPEG, each camera model has its own capabilities to capture a given dynamic range. Basically, if your histogram shows "burned" high light or completely black shadows, information will be missing with a single shot.

If you think that the 0EV shot has a reasonable exposure, maybe you don't need HDR in the first place (you can even get HDR like result from a single RAW).

Now obviously, the EV difference between two shots will be function of the dynamic range you want to capture, thus function of the scene. And yes, in this case some information will be redundant anyway, as what is in the middle range of picture A (say 0 EV) will also be present in B (say -1EV) and C (+1EV). To limit this, you have to set the bracketing parameter to adequate values.

Regarding the number of pictures you need and the EV difference between each shot, plenty of persons have plenty of advices. What works for me as a "standard" (say 1 2/3 or 2 EV difference with 7 pictures for a given scene) may be inadequate for someone else.

I would like to add that the shutter speed isn't the only variable. Some camera can change ISO or aperture to obtain bracketing (use aperture bracketing only if you know what it does).

Kartick Vaddadi provided an example to illustrated how information is indeed redundant, have a look : https://photo.stackexchange.com/a/67276/26456

\$\endgroup\$
3
  • 3
    \$\begingroup\$ Do not change aperture for HDR bracketing. You do not want different DOF between shots. \$\endgroup\$
    – ths
    Aug 9, 2015 at 15:12
  • \$\begingroup\$ You effectively don't want to change aperture for HDR bracketing. If you do, it's for DOF bracketing (photo.stackexchange.com/questions/12553/…) \$\endgroup\$
    – Olivier
    Aug 9, 2015 at 15:22
  • \$\begingroup\$ You don't bracket because you "want to come up with a higher dynamic range." You bracket because you want to make a scene with higher dynamic range fit in a display medium with lower dynamic range. Using HDR does not increase the DR of the display medium, it allows one to fit a wider DR scene into a more limited DR display medium. \$\endgroup\$
    – Michael C
    Jul 13, 2022 at 16:36
8
\$\begingroup\$

Bracketing is useful when the scene exceeds the dynamic-range of the camera. That is it. There is no answer that applies all the time because it depends on the scene.

\$\endgroup\$
0
6
\$\begingroup\$

If I assume that the default 0 EV has reasonable exposure, does this mean that the -1 and +1 RAWs are more or less redundant?

If the exposure is reasonable in both the shadows and highlights — that is, the scene has a dynamic range that easily fits within what the camera can capture — then yes, you're right. Bracketed exposure doesn't do much for you.

Even if the exposure is just one or two stops off at either end, you can often extract enough from a single raw to make a decent composite covering the full dynamic range.

However, if the scene has even more dynamic range than that, having more exposures will help. If there are both very bright highlights and very deep shadows, this may exceed the capability of a single RAW file. In that case, you probably want to bracket even wider than that, taking one central image, and then one to cover the highlights and another the shadows. You may even want to take multiples, getting both the extremes and more in the middle.

Before combining multiple images together for HDR became popular, the primary other reason to bracket was when you weren't sure of which exposure choice you want to make. That makes more sense with film — with digital, you'd normally just watch the histogram, or take a test shot and review.

\$\endgroup\$
0
6
\$\begingroup\$

I've found that bracketing by ±1 or ±2EV doesn't help, but bracketing by ±3EV does.

I did the following experiment:

First, I picked a high dynamic range scene, short of directly shooting the sun. Here, the sun was partially behind thin clouds, and it was uncomfortably bright to look at with the naked eye. In other words, it was much brighter than it looks in the photos below.

For these scene, I took three photos bracketed by 1EV, and did exposure fusion in Photomatix:

enter image description here

I then threw away the -1 and +1 EV originals, and synthesized them from the 0EV one in Lightroom. I then fused all three photos:

enter image description here

I don't notice any difference between them.

I repeated this experiment for ±2EV, first with bracketing:

enter image description here

And then without:

enter image description here

Again, no difference.

I then repeated the test with ±3EV, with bracketing:

enter image description here

And then without:

enter image description here

Please open the photos in separate tabs to view them at full size.

This time, there's a clear difference, both in the colors being off and, in the synthesized version, in a faint rainbow seen on the wall on the right, all the way from the top to the bottom.

The conclusion I drew from this exercise is that even for a high dynamic range scene, exposure bracketing by ±1 or ±2EV doesn't result in a better photo. All it does is waste time and add complexity, like keeping track of more photos, for no benefit. If I bracket, I'll bracket by ±3EV.

This is on the Sony NEX-5R; performance may vary on your camera.

\$\endgroup\$
5
  • \$\begingroup\$ +1 I love this! If you could post the other ones that'd be awesome. Thanks for taking the time to do this! \$\endgroup\$
    – user541686
    Aug 11, 2015 at 3:19
  • \$\begingroup\$ Done. I also rewrote the entire answer to make it clearer. \$\endgroup\$ Aug 11, 2015 at 13:55
  • 1
    \$\begingroup\$ Personal impression only - not a criticism of expressed opinion: A look at the +/- 1EV version at 100% crop shows there is substantially more noise in the single exposure version than in the 3 exposure version. This is visible across quite a lot of the image but especially so in the more shadowed areas eg on the base of the airconditioning unit. While this level of noise degradation would probably go unnoticed by most at eg typical screen resolutions I suspect it would start to show at 4K levels and in eg A4 prints. | This may sound like classic "pixel peeping", but for my own purposes .... \$\endgroup\$ Aug 13, 2015 at 23:57
  • \$\begingroup\$ .... I care much less about fine detail than whether I like the result and I think that in photos which were intended to be 'top quality' I'd notice that level of noise - even if I did not actually'care' about it in many cases. ie I think that those who seem to care quite a lot more than I often do about noise levels would find the differences significant. \$\endgroup\$ Aug 13, 2015 at 23:57
  • \$\begingroup\$ I could see the noise, but only after you pointed it out, and only at 100%. The criterion I use is whether I can see it full-screen on my 30-inch monitor (in fill mode, not fit, since the former zooms in a bit more), and I couldn't. I may change my view when I get a 4k monitor, but for now I'm happy with my opinion that exposure bracketing isn't needed with ±1 or ±2EV on the NEX-5R. In any case, thanks for your input. \$\endgroup\$ Aug 14, 2015 at 7:19
4
\$\begingroup\$

It depends. If the entire dynamic range of the scene can be captured in a single exposure, then it is redundant. If the entire dynamic range of the scene exceeds the capacity of a single exposure, and you wish to capture that entire dynamic range, then you need to bracket.

Be aware that a single 14-bit raw file may contain as much dynamic range information as a a -3, 0, +3 series of 8-bit JPEGs! So shooting a series of raw shots bracketed at -1, 0, +1 includes a lot of overlap, even with digital cameras that can only capture 10-11 stops of dynamic range. The best cameras today can go as high as 13-14 stops at base ISO. As you raise the ISO, any camera's DR will be reduced. For a very wide dynamic range scene you can still get all of the mid-tones without any gaps by shooting a -2, 0, +2 or even -3, 0, +3 series if you are shooting at fairly low ISO (around ISO 800 or less) and center your base exposure value properly.

\$\endgroup\$
1
  • \$\begingroup\$ +1 great points about DR and also how ISO affects it, thanks! \$\endgroup\$
    – user541686
    Aug 9, 2015 at 19:51
4
\$\begingroup\$

Let me add the issue of noise to the answers already given. Suppose that the 0 EV exposure is good enough, the shadows are not underexposed, the bright areas are not overexposed, then it's still the case that using the +1 EV exposure (for the parts that are not overexposed) will yield a lower noise picture. By exposing for longer and then adjusting the brightness, to the desired level, you are eliminating noise by averaging. Also, in a HDR picture compiled from different exposures, it's not the case that each pixel originates from only one of the exposures; typically all the exposures will contribute to some degree. So, there is then also an averaging involved with the 3 different exposures, this will then further reduce the noise.

\$\endgroup\$
9
  • \$\begingroup\$ +1 although I would imagine the +1 EV is therefore more blurry too, though, right? i.e. not necessarily strictly better even if none of the pictures has the shadows and highlights over- or under-exposed? \$\endgroup\$
    – user541686
    Aug 9, 2015 at 20:35
  • 1
    \$\begingroup\$ Why would the +1 exposure necessarily be more blurry? \$\endgroup\$
    – Michael C
    Aug 9, 2015 at 20:39
  • 3
    \$\begingroup\$ @AbdulNQuraishi: Noise is mainly a problem in the darker parts of a picture. Reducing exposure by one f-stop will double noise, and reducing by two f-stops will quadruple it. On the other hand, if one takes a -2/0/+2 sequence of pictures, one would normally mask out the darker parts of the -2 exposure, so the amount of noise in those parts won't matter. The only portions of the -2 picture one would likely use would be those which are exposed to at least 10% of maximum, and those are the parts which will have the least noise. \$\endgroup\$
    – supercat
    Aug 9, 2015 at 21:40
  • 1
    \$\begingroup\$ @AbdulNQuraishi: When using a linear scale, the difference between 4% and 2% brightness will generally be far more visible than the difference between 97% brightness and 99% brightness. If one part of a picture is supposed to be at 3% brightness and another part at 98%, and both have the same +/- 1% worth of noise, that noise will cause some pixels within the "3%" area to be twice as bright as others, but within the "98%" area the brightest pixels will be less than 2.07% brighter than the dimmest. \$\endgroup\$
    – supercat
    Aug 10, 2015 at 17:26
  • 1
    \$\begingroup\$ @AbdulNQuraishi: Another way to look at things is to say that if you need to boost apparent exposure after a picture is taken, that will increase any noise captured with it; if you reduce the exposure after taking a picture, that will reduce noise, but it won't rescue any highlights that were blown out when the picture was taken. The purpose of the lightest and darkest pictures isn't to be usable by themselves, but instead to ensure that, if nothing moved between the three pictures, it will be possible to produce a picture where nothing is terribly noisy and... \$\endgroup\$
    – supercat
    Aug 10, 2015 at 17:32
1
\$\begingroup\$

RAW images do not contain all information that could be retrieved from a scene. They simply contain more information than JPEG images. When a photographer chooses to bracket photographs and combine them later, it is usually because the bright and dark areas of their scene are sufficiently different that the camera is not capable of detecting the differences in the dark areas, while also detecting the differences in the bright areas. To detect the differences in he dark areas, the bright areas would saturate the photosites. By bracketing the photographs the camera detects either of the detail in the dark areas, or more of the detail in the bright areas, thus the information contained within the three photographs is different, sometimes drastically. These bracketed images can be combined to create one image that is a reasonable representation of the scene as it is seen by the human eye.

If you were to try to create the same image from JPEGs, it is possible that more of them would be required due to their reduced dynamic range.

\$\endgroup\$
0
\$\begingroup\$

You may find you want to take either the +1 or the -1 from the RAW, and you won't know until you sit down to work on it. This is true even if you have bracketed exposures.

If something moves between exposures you can have strange effects with proper HDR. Even if you think everything is still the slightest breeze can do this, or a solid tripod on a less-than-solid floor, someone shifts their weight. Even a subpixel movement can show up. Or how about a bug flying across the scene?

Don't forget you can take your +1 raw and get more shadow detail than you could from the +1 jpg, or the raw imported using defaults.

Finally raws aren't that big, not really. They're 10s of MB when drives are 100s of GB. You could always zip them when you're finished editing as well (I used to do this with the 12-bit files from a scientific camera to good effect, I haven't tried with my .cr2s). I work on the principle of "only delete rubbish".

\$\endgroup\$
1
  • \$\begingroup\$ I agree; brackets aren't always for merging. And it can help in culling to be able to look at a bracket and see whether there's anything in that shadow to bother with, without having to process the RAW to see it. \$\endgroup\$
    – 211Oakland
    Aug 23, 2015 at 15:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.