2
\$\begingroup\$

I shoot RAW because the Pros say "shoot RAW". I also shoot bracketed because in a lot of situations (landscape for instance) I can blend them and expose both the terrain and sky correctly.

But my question is: how many stop of expoure can I recover from a raw file without losing detail/sharpness etc? -/+ 1? -/+ 2?

\$\endgroup\$
1
  • \$\begingroup\$ It totally depends on what camera you are using. What camera are you using? \$\endgroup\$
    – Michael C
    Aug 24, 2015 at 21:02

4 Answers 4

4
\$\begingroup\$

First, it's not a issue of detail or sharpness, but one of signal to noise ratio. Second, instead of asking, why haven't you done the obvious thing and just tried it!?

Let's say you ultimately want a post-processed image with 8 bit per color resolution. In theory, that means any additional bits your camera converts color values to represent extra dynamic range you can use by grabbing only a part of the range and still get 8 bits. For example, if your camera makes 10 bit values, then you have 2 extra bits, so you could use 1/4 of the raw range and still get 8 bits.

However, it's nowhere that simple in practise. Those won't be 10 perfect bits, and the dark to light range usually doesn't extend to the ends of the sensor range. Also, just because numerically there is data there doesn't mean good values have actually been filled in.

There is always some noise in the digitized values. How much depends a lot on the quality of the camera and how well you exposed optimally for what the sensor can do. For example with the hypothetical 10 bit values, you could expose 2 f-stops down from ideal. The brightness values would then range from 0-255 (8 bits) instead of 0-1023 (10 bits). However, for any real camera those values will have significant noise on them. This is not a issue of resolution, but you get a sortof "grainy" looking image due to the noise.

In addition to all the above, rarely do you want the raw values to be mapped linearly to the output image. Any non-linear mapping that results in good 8 bits in the image must necessarily start with more than 8 bit values due to some codes being skipped.

All in all, you should be able to expose 1 f-stop, maybe 2, down from ideal and largely get away with it, but beyond that it depends greatly on the quality of your camera. The only way to really know is to try it and see what you get.

Also, if you're going to "underexpose", it's better to let the camera try and compensate than you doing it after the fact. You do this by setting the ISO value higher. For example, let's say your camera's sensor natively does 200 ISO. Instead of taking a picture at 200 ISO that is underexposed 2 f-stops, it is better to expose properly at 800 ISO. That requires the same f-stop and shutter speed, but with the higher ISO the camera adds gain before digitization, and may do some other processing that you can't do after the fact on the digitized raw values.

\$\endgroup\$
1
  • \$\begingroup\$ "rarely do you want the raw values to be mapped linearly to the output image.": Right. RAW data is usually linear, but a typical JPEG uses a gamma range hence 8 bits give you around 11 bits of dynamic range (with less details in the highlight). \$\endgroup\$ Aug 26, 2015 at 12:20
2
\$\begingroup\$

It depends on the camera. Cameras with more dynamic range will give you greater exposure latitude. With standard professional grade DSLRs you get 8-10 stops of latitude. With some cinematic cameras such as the Sony A7S you may be able to get 12 stops. As far as gathering detail from the shadows, a Canon 5D can get 2 maybe 3 stops out of the shadows. In the highlights it can get 2 stops. Remember this is when shooting RAW. To save space you could shoot in JPEG with a low contrast, low sharpening, picture profile. This can sometimes give you the same amount of dynamic range when editing as RAW.

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

It depends on the camera. For example, if your camera has 10 stops of maximum dynamic range at 100 ISO and you shoot at 400 ISO, you may not be able to get any additional stop when going RAW. Even more so at 800 ISO and up.

Check this article with practical examples about it: http://www.clarkvision.com/articles/raw.versus.jpeg1/

Modern cameras also process JPEGs in a specific way to get the most out of RAWs and already compress the original dynamic range (the 10 stops from above) into the final 8 bits of JPEG. Moreover, JPEG is non linear and can already contain a wider dynamic range than what you expect.

Some cameras also have additional features that reduce the difference between JPEG and RAW: for example, "ADR Auto" on Nikons.

Conclusion: use RAW if you want the freedom to change the way the image is processed, doing it for the additional stops is ok, but probably not as much worth as you expect.

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

In addition to the other answers: for sure, it depends on the camera, but it also depends a lot on the picture you are taking. If your picture does not have a lot of contrast (i.e. if the difference between the darkest point and the lightest one is small), then you have much more freedom to over/under expose it. Basically, the histogram of your picture is narrow, you can put it anywhere within the range of acceptable values.

It also depends on the ISO setting of your camera. The higher the ISO, the lower the dynamic range. See for example this chart (tab "Dynamic Range"). Roughly, around 1000 ISO with a typical DSLR, your dynamic range starts being smaller than the one of the JPEG, and there's not much you can expect (but RAW gives you more options for stuff other than exposure compensation, like noise reduction).

\$\endgroup\$

Your Answer

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

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