PPI, DPI, Dots In and Dots Out
A Primer on Digital Images for the Professional
Richard C. Pitt
The following is an answer to a question by my friend
and associate, Gary Bannerman.
Richard:
Forgive me if this is an inane question, but it is becoming vital in every
publication I do and getting an answer is not easy because:
1. Photography people know nothing about computers.
2. Computer people know nothing about photography.
3. While photographers know something about professional printing, computer
people know zilch
4. As you are fond of telling people, web designers (computer) know nothing
about graphic design (print)...and that PC/Mac web
design - still more concerned with engineering than art - is 5 years behind
print design.
5. The "help" menus on computers and the the
manuals accompanying cameras etc. are written by people with profound
ignorance of one sector or the other.
6. The commercial world is obsessed with JPEGs ... without properly
comprehending that JPEG is a compression format and that most of the software
and Internet transmission modalities assume that compression is the intent
(and like GIFs) ...the object is to achieve a
pretty picture on a computer monitor in the smallest digital size
possible...so every time anyone goes "save as" and requests JPEG, it
compresses further.... unless specific contrary instructions are given.
Here's the question:
When digital images directly downloaded from any digital camera show up in
PC software such as Photoshop... the "image size" box shows this as 72 ppi
... which, of course, is the optimum for a PC monitor.....
and it is easy to "save as" the image to 300 ppi or more...
but my question is whether the basic image we are starting with is 72 ppi....
or is that just a reading of how it shows on the monitor?
As you know, saving a truly inferior image to a 300 ppi or more resolution,
merely demonstrates the pixelation in its worst form.
Bottom Line: what does 72 ppi mean in Photoshop after
images are downloaded from digital cameras,
because 72 ppi going to the printer invites disaster...
Gary hit the mark with the question - it is on a
subject that has bugged me (when other people talk about it with what I
consider obviously incorrect understanding) for years.
The quick answer is that resolution of an image is not
the same as the resolution of the display you're viewing it on or the printing
process used to produce printed output. If anything, it is only an indication
of how large the image will be when displayed on the medium - a 720x576 pixel
image will be 10" x 8" on a screen that is 72 PPI (pixels per inch).
The more complex answer is the deceptively simple
statement "it depends."
The problem is one of resolving the differences between the various
technologies you're dealing with in going from original subject to final
product(s). We deal with the differences between continuous tone and filtered
primary colours as well as the difference between additive and subtractive
primary colours - then deal with pixelization, number of light levels
recorded, compression, viewing distance and resolution of the various elements
in the process. Along the way the professional will use
this information to get the best reproduction of the original on the intended
display medium. PPI/DPI is only one
measurement of the technologies in the steps involved,
and many use it completely wrongly.
The professional also has to deal with:
I'll deal with these in subsequent articles.
An Overview of the Process:
The original subject is lit by a continuum of light colors,
typically daylight or flash.
From here, all reproductive technologies rely upon the fact that the human
eye actually only sees 3 colors and uses blends of these to "see" the full
spectrum.
The film camera uses 3 light sensitive emulsions to record the primary
colors with all light for a single "spot" being picked up at that spot on the
stacked emulsions. There is no averaging. Resolution is
based upon the size of the silver grains in the emulsion; thus "fine grained"
film is higher resolution than "normal grain". The tradeoff is that finer
grains are less light sensitive.
The CCD digital camera/scanner on the other hand doesn't register the colors
for any single spot, but instead takes groups of 3 spots and averages them,
using colored filters over the sensors on the flat plane of the chip.
A new digital chip (Foveon) uses a digital process that closely matches the
depth filtering of film to get the color right for each spot by using the
depth of the silicon chip to filter the light and register the color for each
spot correctly - no averaging.
As a counter point, NASA (and other
space agencies) uses technologies that employ moveable filters and a single
CCD chip to achieve very high resolution images - but this only works with
static subjects (subjects that don't move during the time it takes to move in
and out the various filters needed.) Also, in some professional video cameras
(the technology has not yet made it to the mainstream still cameras) a prism
and/or filter system is used to split the incoming image to 3 separate
imaging chips in "real time" to achieve higher resolution (and higher capture
speeds) than can currently be done with all the pixels on a single chip.
At this point (in the camera or scanner) we talk of
resolution in pixels without regard to the size of the imaging technology. It
matters little whether the imaging chip is 10 inches square with a huge lens
and large points on the chip or 1 inch square with the imaging dots
microscopic; if the resulting image contains 3.5 Million pixels then the
resolution is the same. The PPI for the former would be on the order of 180
where the latter would be closer to 1800. The point is that in this case the
PPI relates to the image capture technology.
Again as an aside, I have seen an
interesting item where a camera was made from an old plate-camera body and a
flat-bed scanner. The resulting images (of static items only due to the long
time for the exposure) are incredible - every bit comparable to film of the
same size (8" x 10")
PPI in the camera is roughly equivalent to the
grain size in film. It is the smallest point that a CCD can register any
useful image information. The problem is, with
the typical still camera CCD they
(the vendors) count all the dots when in fact they should count all the
groups of 3 dots since that is actually what is needed to equal the single
spot on a piece of film (or a dot on the foveon chip).
Note that with many of today's
cameras, the size of the imaging chip is so small that, even though it may
have millions of pixels, the lens in front of the camera may not be capable
of resolving the image well enough to take advantage of this resolution. The
whole camera system needs to be of consistent quality. The problem here is
that the technology necessary to create an imaging chip with a size similar
to a 35mm film frame is very expensive. It's not that it can't be done, but
that the number of flawless chips of that size in a batch is lower than the
yield of smaller sizes.
If we ignore for the moment the use of compression and take the image
directly to the output medium, we deal only with the direct mapping of the
pixels to the output.
If we are going to view the image on a computer screen and want it at full
resolution, we'll need a screen with as many pixels on it as the original
image has: at 3.5 Mpixels it would be about 1870x1870 for a square image and
at 90 ppi (the resolution of an excellent monitor)
it would be about 20" x 20" - a pretty large screen even by today's standards.
If we print it at full resolution on a 600dpi printer
the image would be about 3" x 3"
If we print it using a 100 screen on newsprint we would get about 18" x 18"
The point is that we'd probably hold the print up close, the newspaper far
away, and sit a couple of feet from the computer
screen - adapting our viewing so that the important part of the image we are
looking at will fill the center part of our field of vision - between 10 and
20 degrees on either side of directly in front of us.
In both the printing systems (the computer printer and the newspaper) the
colors are built up one at a time in registration, whereas the computer screen
uses side-by-side color dots the same as the CCD uses image elements - in
groups of 3 to average the actual color. Note that the screen also uses
additive color where the printing uses subtractive.
Now we come to why DPI, and PPI (and printing screen size) have any need to
exist at all. It is the same reason why TV news anchors are told not to wear
plaid suits or striped shirts - moiré patterns. See the bottom image at
http://www.foveon.com/X3_comparison.html for a prime example.
If you don't match the image to the viewing/printing technology, you may get
visual artifacts that are displeasing to the eye.
Some of the best information on the web about the overall subject can be seen
on the web site of the company Foveon. They released a new imaging system
(competition to the CCD found in most digital cameras today) last year. See
the images at
http://www.foveon.com/X3_tech.html for a comparison of film "pixels" with
CCD pixels and the new foveon technology.
Note that with a CCD imaging system, the placement of the picture elements
(pixels) is such that it only averages the color over a fairly large area
(that of 3 actual sensors) instead of getting the color through the filtering
of a single sensor spot (as in foveon) or the tight grain area of 3 layers of
film emulsion in real film. If you were to focus a light beam such that it hit
a spot 1 micron by 1 micron on each of the 3 types of image capture (film,
CCD, foveon) the CCD might register it only on a single element of one color
(green for example) where both film and foveon would register it on all 3
color layers/sensors and record its actual color which might be white (all 3
primary colors)
So the final answer to Gary's
question:
"When digital images directly downloaded from any digital camera show up in PC
software such as Photoshop... the "image size" box shows this as 72 ppi ...
which, of course, is the optimum for a PC monitor..... and it is
easy to "save as" the image to 300 ppi or more... but my
question is whether the basic image we are starting with is 72 ppi.... or is
that just a reading of how it shows on the monitor?"
... is that this matches the "grain" of the image to the
"grain" of the monitor so you don't get image artifacts caused by the
mismatching of dots. You should save an image at the resolution of the output
medium you are going to use, or at some even fraction of that resolution so
that the dots line up (if the printer can do 600DPI then you might save at
600, 300, 200, 150, 100 but not at 72 or at 720) and
yes, this means that even saving a poor image at higher resolution may make a
difference - but probably not enough to warrant it. Better to save it at a
fractional PPI/DPI instead.
Note that all through this I've ignored compression for file size savings and
re-sizing for image sizing on final output. These also can affect visual
artifacts.
There are two distinct classes of compression: Lossy and
Loss-less.
Lossless compression generally looks at the actual bits
of the data file without regard to the fact that it contains an image. It
looks for patterns that can be encoded using fewer bits, and does this in
various ways depending on whose algorithm is being used. The most typical is
called "LZW" for the initials of the name Lempel-Ziv Welch. See:
http://dictionary.reference.com/search?q=lempel-ziv%20welch%20compression
for a description.
Lossy compression works by comparing the values in
adjoining pixels and averaging them. The higher the compression ratio, the
larger the area that is averaged, until the fact that it is averaged becomes
obvious to the viewer. The most typical used today is JPEG. It also works by
palette simplification - which can lead to bands of colour through areas of
constant gradation; not as noticeable on things with lots of random detail.
See:
http://www.google.ca/search?q=define:JPEG+compression for more
information.
The only compression the professional photographer should use at any time
other than when creating a specific output image for an end use where file
size is a concern is a loss-less one. JPEG includes loss-less to 25%
compression (75% of "normal" file size) but it is better to stick to
TIFF and use the compressed TIFF (LZW) which is
loss-less - this avoids the potential of inadvertently going past 25% with
JPEG and losing detail.
All in all, the image creator should know what the final output needs to be -
and should have an understanding of the technologies involved along the way so
that the final product doesn't suffer from the manipulation it undergoes in
the process. There are all sorts of tools that can help but most degrade the
original in some fashion if used improperly. You should
always start from the highest resolution possible and enlarge as little as
possible. More than any other format, this means doing your cropping in the
camera as much as possible. Today's high quality zoom lenses on good digital
cameras are a far cry from some of the original "variable focal length" lenses
of 30-40 years ago. If you have one, use it to best advantage and get as much
of your intended subject as possible in the frame.
If you are using a digital conversion and are unsure of
your lenses, use as high an f-stop as possible to maximize sharpness;
remember, the image area is smaller than you are used to and tests the
resolution of the lenses.
Other pointers:
