No, it's not your neighbor's kid and his garage band keeping you
awake late at night. It's an entirely different problem and it can
still cause some sleeplessness.
Many things in our daily lives reflect a signal-to-noise ratio. If
you live in an urban environment you'll be familiar with trying to
hear the person on the other end of your phone when you're in a room
full of other people talking. The person you wish to listen to is the
"signal," all the other voices you're trying to filter out is the
"noise." If you're using a cell phone and it drops a call it's also
trying to deal with too much noise compared to the radio signal it
needs to work with. Do you try to shield the sun from your eyes when
you are trying to see something? This is close to the topic to the
topic this article will cover.
Signal-to-Noise ratios (SNR) can be described in many different ways
and affects many different things we all experience daily. I'm not
going to explain more about that though, you can learn plenty about
it online if you're interested.
I'm going to spend a little bit of time discussing a very specific
area of SNR and how it can affect certain aspects of digital imaging.
Again, I'm not even going to spend too much time on the whole
digital-imaging process because there are plenty of other venues to
learn more about that too.
What I am going to describe is specifically related to
POST-PROCESSING (PP for short) a digital image file, in this case one
generated by a digital camera, and, even more specifically, a file
produced by a digital camera called a RAW file. Low end cameras don't
generally produce raw files. Digltal SLRs and many high-end compact
cameras do, however.
If you are already familiar with a raw file then read on. If not, again, there are other resources out there to learn about raw files too. If you're an avid or aspiring photographer who wants to get the most out of your camera, then you will likely want to learn what raw files can help you achieve. Raw files are very much more versatile and malleable than jpg image files. You can do a lot with them, especially if your creative needs have you adjusting your photographs to look a different way using PP software and techniques.
Photographing a scene with a large dynamic range is a little like
wearing sunglasses to comfortably see a bright snowy landscape but
now it's difficult to make out a black bear standing in the shadow of
some trees. Ok, bad analogy because the bear should probably be
hibernating but let's go with it.
You control the brightest part of the scene so that it can be seen or
recorded but that tends to push the already dark parts of the scene
even darker. Our eyes can see in a very wide range of brightness,
from starlight at night to a sunlit snowy landscape. But our eyes
can't cover more than a part of that very wide range at once and
neither can cameras. We perceive a wide range of bright to dark in a
given scene because our eye quickly adapts to the brightness of the
area we focus on, a form of biological auto exposure. The only way we
can accomplish something like that with a camera is to be able to
record a very wide range of bright to dark areas with good fidelity
and then compress this range so that it all stays within a
comfortable viewing (or printing or displaying) range.
Camera and photo-printing systems already do some dynamic range compression and non-linear response in order to produce a more natural looking image to us. Sometimes we want to have more control over this part of the process for creative means.
Look at the two images below;
Image notes:
Canon 40D, 100 ISO, 1/1000s, f/5.0, Aperture
priority, +2/3 EV compensation, evaluative (matrix) metering.
If I am to believe the luminace histogram for the raw file, this
exposure covered a range of nearly -4EV to nearly +3EV, O being
midtone for the camera.
About 2/3 EV of headroom remained unused even with the +2/3 EV
exposure compensation.
The top is how the camera interprets and records the scene.
The image on the bottom has been PP'd to better convey how I
perceived the same scene and therefore how I want it to look in print
or display.
Both images are based on the same information contained in the same
raw file.
In this case software was used to embellish many details of the image to obtain the final result but the most striking difference is in the improved balance of tone between the bright sky and the dark green pasture. Post processing effected a further compression of the overall brightness of the scene by "pushing" the dark areas about 2 to 3 EV brighter.
This was possible because the camera was still able to record the darker areas of the image with enough fidelity that they could be enhanced without showing any noise or other disagreeable artifacts. This camera still has a good SNR at those lower signal levels (dark areas) so when we amplify the signal, the also amplified noise is still negligible and does not appear to detract from the final image.
This technique is sometime referred to by the long phrase, "Expose
for the highlights and push the shadows in post."
It also qualfies for the ETTR or, Expose To The Right method of
maximizing your image quality when PP is going to be done to it. ETTR
is another thing you can learn more about elsewhere, it's a very
useful method when properly applied.
This can also be considered a form of HDR photography based on a single image rather than using multiple exposure-bracketed images. This single-shot type of HDR is feasible when the range of bright to dark areas being recorded is not too great for the camera to be able to record and the camera's SNR at dark levels is negligible. That latter point is not always the case, however, bringing us to the crux of this article.
For a lot of people, this just does not matter or they choose not
to do it because they are content with the image as originally
rendered.
For those of us who DO use this technique, for whatever our artistic
reasons, this is an important point to know so that we may get the
most out of our camera or possibly even use this as a basis for
selecting a specific camera for certain kinds of work. The above
mentioned methods of ETTR and-or expose for the highlights have been
in use for a while and, to some extent, help to maximize the quality
of the image you can extract from a file.
Being a hardware geek by nature and profession, I wanted to see what various camera sensors recorded for dark noise using a standard "exposure" setting when shooting.. nothing. Cameras were tested with the lens cap on, in a dark room, viewfinder port covered, cameras LCD display off, all noise reduction methods disabled, all tone curve adjustments disabled and with manual selection of DayLight White Balance and 1/200s shutter speed with f/16 aperture. Test shots were then taken at base and every full EV ISO setting (100, 200, 400, etc.)
Prompted by some problems I recently encountered working on some
of my images, I've gathered "dark shots" from numerous different
cameras in raw file formats during February and March of 2012.
I then processed those images by performing a +4 EV adjustment. This
causes the dark levels of the image to be brought up by about 4EV so
any noise present will become visible. This was done for most or all
typical ISO settings for the tested cameras.
Why? Well, images I shot with a fancy new camera in 2011 turned out
not very workable in the PP phase. This was not just disappointing
but almost infuriating to discover that many of my shots of high DR
scenes taken with this camera, were not going to meet my standards if
I pushed the shadow areas up where I wanted them. I wish I would have
discovered this much earlier, I may have decided to return the camera
or at least change how I used it to mitigate this problem. This
discovery promted a large diversion into learning more about software
options for PP and even a look over the fence at cameras produced by
a different manufacturer.
More to come on this page:
- testing procedure and post processing details for how this was done.
- random vs pattern noise (hiss vs tone analogy), which is worse and why