Quote:
Originally Posted by acropolite
On my monitor the HDR image looks rather flat and unnatural. The colours of the rocks on the middle RHS of the image look washed out and quite unnatural, whereas the colours from the D300 seem more natural and vibrant.
Not having been there, it's difficult to say which image more accurately matches the reality, but, despite being a Canon man, I must say that in this instance I much prefer the image from the Nikon
|
Hi Phil,
Firstly, thanks to the guys for the lovely images!
I just wanted to take a few moments to help clarify some technical points.
It is important to keep in mind that none of the excellent images on this thread
are HDR images. All of them are LDR images.
This is a very important distinction and a key to what is really going on.
Bear with me and I will try my best to explain what I mean.
Ideally, cameras would be able to take a HDR image with a single exposure.
Just as ideally, rendering devices, such as monitors and printers, would then
be able to display them directly with the full range of luminance levels.
Please allow me to give an extreme example. If the technology existed, one should
be able to take a picture of the sun, such that, when it was displayed back on
a monitor, one would need to wear appropriate protective eyewear just to
view it. Putting aside, for one moment, the obvious health & safety ramifications
such a hypothetical technology might pose, it does bring into some context the current
state of the art in imaging and display technologies compared to what we might
dream about as being possible in the future.
The dynamic range of an image is the ratio of the most luminous part in the image
to the least luminous part.
Though a human eye can adjust itself over a dynamic range of about
1,000,000 : 1 and at any one instant, has a dynamic range of about 1000:1,
many consumer grade digital cameras have a dynamic range of only about
500:1 and LCD monitors only about 600:1.
Therefore compared to our own sense of perception, technology as it stands
today just doesn't have the numbers to back it. As tools for capturing
and reproducing images as we actually see them, in some ways, they are still
measurably blunt instruments.
It's a bit like comparing audio reproduction between an old 78 rpm record and today's
Dolby surround systems. The 78 recording and then playing back on the gramophone
just doesn't have the dynamic range to reproduce the original concert performance.
Since common commercial cameras don't currently possess the capability to capture
in one instant anywhere near the same dynamic range that the human eye/brain
can, HDR is currently a bit of a kludge. We take multiple images at different
exposures and use software to combine and convert them to an image
format whereby floating point numbers are used to represent the image in terms
of energy levels of light for a given area that falls with a given solid angle.
If our monitors and printers could then directly render such a HDR format, the
results would be dramatic.
Alas, they cannot. Our monitors and printers have very limited dynamic range
output capability. Whereas we refer to the HDR file as being "scene referred",
we now need to transform the HDR to a Low Dynamic Range (LDR) image
to match the inherent LDR and limited gamut of our output devices.
Such a transformation results in the data now being "output referred" .
The class of transformations that perform this conversion are referred to
as "tone mapping". Tone mapping is an entirely different and distinct process
from the creation of HDR images.
Tone mapping is not an exact science. Instead, there exist various
algorithmic attempts to perform tone mapping and different software
packages may offer one or more different algorithms to perform it.
Whereas when a user synthesizes a HDR image from multiple exposures,
the amount of control they have over the HDR image is limited, during the
tone mapping process, typically they are presented with many user selectable
controls.
Tone mapping throws away information. So what the user is doing in one
sense is making decisions about which information in the HDR image will
be used to create the LDR image and which will be thrown away.
Some of the tone mapping algorithms attempt to mimic an attribute of
human perception known as "locale adaptation". Our eyes can change
their sensitivity in different areas across the field of view. How we do this
is not completely understood. Tone mapping operations that attempt to
do the same are crude attempts. It is still early days.
One side effect of some popular algorithms that attempt to mimic locale
adaptation can be artifacts such as halos.
So what is the future of HDR? It is a reasonably safe bet that, in the future,
all cameras will be HDR. What we now know as "Raw" formats will one day
disappear. The current limitations are purely based in semiconductor
physics and engineering.
It may turn out that HDR output devices are the more challenging.
What I suspect is that the future may bring some new output technologies
with higher dynamic range than today's monitors. Some form of tone
mapping might still need to take place as long as the dynamic range of the
output device is narrower than the input device. However, the photographer
will have a broader gamut from which they can draw and for those interested
in 'photorealism' rather than visual art, the process will be more forgiving.
Just to touch briefly on the differences between the two brands of camera,
you can see if you breath in all of the above that the cameras themselves
only play one part in the whole process of producing the final image that appears
on the screen. The tone mapping operations and the decisions the
photographer makes about how the tone mapping algorithms will be
applied will likely account for more of the differences in color than those that can
be attributed to come from the cameras themselves. So no need to trade
in your camera for some other brand.
So in a nutshell, there are two processes. The first is HDR. Since none of
us have output devices that can directly render HDR, the HDR files
are actually still sitting back on the hard disks of our intrepid photographers.
They can't 'view' them directly either.
The second is tone mapping. We are on an entirely different page now.
Whereas one can say, "I like the look of the results of this tone mapping
operation", it doesn't make much sense to say "I like the look of this
HDR image". Semantically, it is now common to refer to an image that
was originally created with HDR as an "HDR image". But many enthusiasts
confuse the results of HDR with those of tone mapping and they are two
entirely different things. The danger is that some might throw the baby
out with the bath water. When HDR becomes "in built" and the default
image capture mode in cameras, there is the risk that there might be some
that say, "but I don't like HDR" because of their experience with tone
mapped LDR's. It would be a bit like saying, "but I don't like Raw images".
Hopefully the above might be a helpful insight for those unfamiliar with the
underlying technologies involved.
Thanks again to our intrepid photographers for their splendid posts!:
Best regards
Gary
The cameras are only one part of the workflow
