When people see a post-processed picture after using software such as Adobe
Photoshop, or an HDR picture where tone mapping [1] has been used, a
frequent comment is that they do not think the picture looks realistic - this
begs the question of what people mean by realistic in this context. I believe
that when people see a heavily post-processed picture they consider to be
unrealistic, what they really mean is it does not look like a normal
photograph.
Before the invention of photography, the only time people would see representations
of scenes was in paintings and drawings. Landscape paintings by artists such as
Constable do not look much like modern photographs of similar scenes.
Interestingly when people see a tone-mapped high dynamic range picture they
often say that it looks more like a painting than a photograph.
Superficially the human eye is similar to a camera in that it has a lens
that focuses on a sensor, and it has aperture control and automatic focusing.
However there are major differences: essentially a camera takes a snapshot with
one focus setting and one aperture, and is a freeze of the picture taken in a fraction
of a second. Conversely the brain/eye combination is more like a movie camera
in that the eye is continuously scanning the image and the brain is composing a
picture, so over a short period the aperture and focus are changing. Another
significant difference is that we have two eyes, so depth perception is much
better because of 3-D vision. Although 3-D cameras are available they are
unlikely to become mainstream any time soon because a special viewer is need to
see 3-D pictures.
When people have taken photographs in bright light conditions with dark
shadows, they are often disappointed with the results, as often they do not
think they represents what they actually saw at the time. The main reason for
this is that human vision has a much greater dynamic range that that available
on most modern cameras.
Dynamic range (or contrast ratio) is often measured by ratios such as
1:100,000. That is the brightest part of the picture is 100,000 times brighter
than the darkest part of the picture. The darkest part is defined as the
darkest colour above pure black that can be represented. These ratios can be
difficult to understand, and I think it's better to use EV (Exposure Value)
terminology, which is commonly used in photography. An increase of one EV is a
doubling of the light received. Modern cameras can capture up to about 8 EVs if
the picture is stored as a JPEG. In raw mode (usually only available on SLRs) more can theoretically be captured, but in practice most cameras cannot do
more than about 12 EVs. Note that the dynamic range is more a function of the sensor than the bit depth. How human vision actually works is not well understood, but
it is believed that the eye can capture at least 14 EVs. The fact that modern
consumer technology cannot achieve the dynamic range of the eye is one of the
reasons for many disappointing photographs. A modern display screen/display adapter
can represent about 8 EVs, although in practice it is usually lower than that
on LCD displays because of the backlight, which tends to turn black into dark
grey. It is possible to get displays that represent about 17 EVs, but they are
extremely expensive (around $50,000), so these are only really used in
specialist applications like medical imaging. The situation is even worse when
you try to print pictures, as it is very difficult to get printed material that
contains much more than about 6 EVs.
Even if you take high dynamic range pictures using techniques such as
overlaying multiple pictures taken at different EV levels (HDR), you still need
to represent these on paper or an ordinary display. The standard way of doing
this is to use tone-mapping, which essentially is a non-linear transformation
of the large dynamic range back down to around 8 EVs. Generally when people
refer to an HDR picture, what they are really are looking at is a tone-mapped
version of the picture.
When JPEG (but not Raw) is the saved format, all cameras do some
post-processing of the picture to produce a more balanced image. A gamma curve
[5] is typically applied to produce a picture which is closer to normal
human perception. Some cameras have optional facilities to lighten dark areas [6]
and increasingly cameras are becoming available with simple built-in HDR
processing [7].
I believe that tone-mapping is essentially what the eye/brain combination
is doing, and also what landscape painters are doing naturally. I would contend
that a realistic [2] tone-mapped version of a picture is probably a more accurate
representation of a scene than a straight photograph. Note that not all scenes
need a high dynamic range – on an overcast day a conventional picture is fine.
But in bright sunshine I suspect that users will increasingly expect their
camera to be able to provide an accurate representation of the scene as they
perceived it at the time.
Example
The second picture here is a tone-mapped HDR from 5 original photos taken
at -2, -1, 0, +1, +2 EVs, and post-processed using special software
(PhotomatixPro). The first picture is the middle picture of the sequence, and
is what you would get with a straight photograph. Because of the very dark
areas and the bright sunlight in this shot, it is not possible to capture the
full dynamic range with one shot on a conventional camera. Most of the first
picture is either under or over-exposed. The use of tone mapping enabled me to
produce an image that was much closer to what I observed at the time.
Additional Notes
- For more information on
HDR and tone-mapping see:
http://en.wikipedia.org/wiki/High_dynamic_range_imaging - A realistic tone-mapped
image is where tone-mapping software is used to try and create a representation
of the picture as the photographer saw it at the time. Examples can be
seen in this Flickr group:
http://www.flickr.com/groups/realhdr/pool/79968332@N00/ - Tone-mapping can also
be used to produce exaggerated effects similar to surrealist or
impressionist painters. This is moving into the realms of photography as
art, and away from realism. As with all art, some people like the results
and some people do not – but it is possible to produce some dramatic
pictures from what would have been a dull single photograph. Pictures of
dilapidated buildings or old vehicles are popular topics for this treatment.
Various HDR groups on Flickr show examples of this, such as this one:
http://www.flickr.com/groups/hdr_photos/pool/ - New display technologies
currently being developed hold out the prospect of affordable displays
with improved dynamic range. However there are many technical difficulties
and progress is slow. In particular OLED displays do not have a backlight
and so, in theory, should be able to display pure black, and a much better
dynamic range:
http://en.wikipedia.org/wiki/Oled - More information on Gamma Correction can be found here: http://en.wikipedia.org/wiki/Gamma_correction
- The Dynamic Range (DR) facility on Sony Alpha cameras is on example of this. This allows for the lightening of dark shadow areas to bring out more detail. For JPEGs this is done in camera, but if Raw format is used additional control is available using Sony’s supplied software. One advantage with this is that it is done with a single shot and so can be used when there is movement in the subject.
- Manufacturers are now
building-in simple HDR facilities to some of their cameras. Pentax and
Sony have several models that support HDR, but most manufacturers have at
least one camera that supports HDR/tone-mapping. The iPhone 4S has an HDR
facility. Generally with the in-camera facility you have little or no
control over the final appearance, but it is much easier than creating
several images and post-processing them.
Photographs are of the Cloister at Iford Manor, Wiltshire, UK.
