The Megapixel
Masquerade
Background
Imagine a world where a
camera can be dubbed "14 megapixel" when it has 4.6 million pixels on
its imaging sensor while at the same time, another camera can be
dubbed "10 megapixel" when it has no pixels at all on its
imaging sensor. Sound like a strange world? Maybe, but it's
the world we live in today! Evolving technologies are making it
more difficult to define exactly what is meant by the term "megapixel"
and is blurring camera specifications to the point that many people no
longer know how to compare cameras by the specs alone. In this
article, I will try to explain how manufacturers come up with their
marketing regarding the term "megapixels".
Pixel: a definition
Wikipedia defines a
picture element, or pixel, as "the smallest complete sample of an
image". Others use similar terminology such as "the smallest
discrete component of an image". The key here is that we are
talking about the smallest element in an image: that is,
the final picture or photograph. Obviously, a digital photograph
is made up of millions of tiny points of light, each of which can have
its own unique color and brightness. When these points of light
are displayed next to one another and viewed from a distance, the
individual points of light fade together and we see what appears to be a
smooth, continuous image. There are various ways to represent
color and brightness for each point of light or pixel in the image, but
the most common is to assign each pixel its own set of red, green, and
blue brightness values since you can reproduce a particular color by
combining red, green, and blue intensities. A pixel then, must
have all three (red, green, and blue) components to be a complete sample
of the final image.
Sensor photosites and
pixels
Ever since images from digital
cameras broke the one million pixel boundary more than a decade ago, the
term "megapixel" has been used to describe resolution. Using this
term, buyers could get an idea about how large they could print, how
much leeway they would have to crop images, and so on. While a "10
megapixel" claim is accurate with respect to how many pixels are in the
final (developed) image, somewhere along the way, the megapixel moniker
has gotten confused with "camera resolution". A typical camera
claimed to be a 10 megapixel digital camera may produce 10 megapixel
images, but by definition, the camera itself (the sensor) does not
contain 10 million pixels. Far from it in fact! This "10
megapixel digital camera" actually contains no pixels whatsoever on its
sensor. Instead, the sensor is a conglomerate of 5 million green
photosites, 2.5 million red photosites, and 2.5 million blue photosites.
Sophisticated software takes information from these 10 million
individual samples of red, green, OR blue at each location
in order to predict the missing two color channels at each pixel in the
final image. Since a pixel is defined as a complete picture
element, a typical digital camera cannot be defined as a "10 megapixel
camera" even if it produces a 10 megapixel final image because two
thirds (67%) of that 10 megapixel final image is "predicted" rather than
actual data. For the camera itself to be called 10 megapixels, it
must have 10 million pixels on the sensor, each of which is able to
represent complete information without borrowing information from
neighbors.
Enter Full Color Capture
For about a decade, none of this
pixel definition nit-picking mattered because all cameras were roughly
the same. They all captured only one of the three red, green, or
blue colors at each location on the sensor and they all predicted the
missing two colors by looking at neighboring locations on the sensor and
predicting. The fact that your 10 million pixel image didn't come
from a 10 million pixel camera didn't matter because everyone was
compared on a level playing field. When Sigma introduced the first
consumer full capture camera (the SD9) in 2002, they were faced with a
dilemma. Should they call it a 3.5 megapixel camera because it
delivers 3.5 million pixel final images, or should they call it 10
megapixels since it captures all three red, green, and blue color
primaries at each location on the sensor? Technically (by the
definition of a pixel), they should label it as a 3.5 megapixel camera
but its competition at the time were cameras dubbed as 6 megapixels even
though they were not really 6 megapixel cameras. Now that
technology was changing, the "fuzzy" definition of megapixel that had
worked for years suddenly broke down. People started picking sides
and arguing apples versus oranges.
Fast forward to 2007 and the same
problem exists today. Sigma's updated SD14 produces a 4.6
megapixel final image from 4.6 million sensor pixels. Once again,
Sigma was faced with how to label their product since the competition
was calling their cameras 8 and 10 megapixel yet those cameras recorded
no true pixels at all and the final 8 or 10 megapixel image had to be
"derived" using a lot of educated guessing (read complex predictive
analysis). Had Sigma called their SD14 a 4.6 megapixel camera,
most consumers wouldn't realize that since the camera captures full
color, its final images are comparable to images from typical (non full
color) 10 megapixel cameras. They chose instead to take the "high
road" and label it a 14 megapixel camera figuring that if the rest of
the industry can claim 10 megapixels when only one third of each pixel
is real data, they can claim 14 megapixels when they are capturing all
three primary colors (4.6 x 3). In reality, Sigma marketing was
fighting misleading terminology with more misleading terminology.
They likely felt they needed to because it was easier than reeducating
the masses by writing an article like this and then hoping everyone
reads it. The phrase "damned if you do, damned if you don't" comes
to mind here.
Does it Matter?
It's interesting that some (both
online and hard copy) publications can claim that calling a 4.6
megapixel full capture camera 14 megapixels is hype when no one
complains that a camera advertised as 10 megapixels can't deliver 10
megapixels of real image information. What's the real hype here:
the fact that the SD14 is really 4.6 megapixels and not 14, or the fact
that a typical camera labelled 10 megapixel really only captures one
third of the information at each pixel? The truth here is that
sometimes you have to read the fine print. When comparing single
color capture cameras with full color capture cameras, just keep in mind
that megapixel ratings really cannot be compared directly. Both
technologies work and one is not necessarily better than the other for
all things, but when comparing megapixel numbers on paper, it's
beneficial to note that the term "megapixel" is used rather loosely in
this industry by both camps: the typical single color
capture camp and the full color capture camp, i.e. Foveon/Sigma.
Due to the filtering and reconstruction involved in creating an image
from a typical single color capture camera, it can resolve less detail
per final-image-pixel than a full color capture camera like a Sigma
SD14. How much will depend on the image, but a decent rule of
thumb is that full color capture cameras like the SD14 compare nicely to
cameras with about twice as many pixels in the final image. That
is, the 4.6 megapixel SD14 can resolve detail comparable to a typical
(single color capture) camera rated at about 9.2 megapixels. I
admit it's a bit silly to try to explain "fuzzy" logic with even more
"fuzzy" logic but sometimes it's necessary unless you expect all your
readers to have engineering or computer degrees. :-) If you
want to read (and see) more about how complicated it can get comparing
single color capture to full color capture, read my article on the
SD14 versus Canon 5D
where I take a look at some of the intricacies involved in comparing
typical single color capture cameras to full color capture cameras.
The Eyeball Argument
Some reviewers screaming "hype" on
the 14 megapixel designation of the Sigma SD14 argue that normal single
color capture cameras can actually approach their rated resolution even
when only one color per pixel is captured by the sensor. I've seen
claims that cameras rated at 10 megapixels can approach 10 megapixels of
true resolution especially when capturing black and white detail.
While the algorithms designed to create a full color image from
one-color-per-pixel sensors are actually pretty good at what they do
especially on black and white detail, the edge blurring needed in order
to make single color capture work properly holds them back from their
upper limit potential. Single color capture really starts to fall
short (of rated resolution) when capturing highly detailed colorful
subjects where the red, green, or blue locations on the sensor start to
contribute less information than they would in a B/W scene such as a
resolution chart. I've also heard the argument that single
color capture cameras, particularly those with the Bayer RGBG design,
try to replicate how the human eye works, giving more resolution to
green and less to blue and red, so that design is actually better as a
result. Such arguments are absurd, however, when you realize that
replicating the deficiencies of the human eye is not a
benefit but rather a necessity for single color capture! The goal
of any imaging device should be to produce the highest quality
photographs possible and reproducing the most accurate information for
each pixel is how we accomplish that task. This is how,
resolution-wise, full color capture cameras like the SD14 can
compare nicely to single color capture cameras with much higher final
image resolution. All this just goes to show that single and full
color capture are not comparable on paper no matter what arguments are
used to try to rationalize the comparison.
Summary
Don't be another victim in the
megapixel wars. Arm yourself with a little knowledge and you won't
have to take the manufacturer's word for it when trying to compare
(especially differing) technologies. There's much more to buying a
camera than just megapixels, of course, but if you like to look at
specs, maybe this article will help a bit with understanding some of the
claims made by manufacturers today with regard to megapixels and
resolution.
Mike Chaney