What’s In & What’s
Out : Interpolate ?
Interpolation is a mathematical discipline
vital to all image processing. It is
used in everything from space satellite data
to your digital camera. In digital
photography and editing, three basic methods
of interpolation are used; nearest neighbor,
bilinear, and bicubic.
The nearest neighbor method simply takes
adjacent values and copies them to new pixel
positions for filler use. This gives
rough results with continuous-toned images,
like photographs. It works best for
bit-mapped line graphics, text, or anything
with only black and white pixels.
Bilinear interpolation looks at the four
pixels surrounding it’s self; top,
bottom, left and right. It then calculates
new values by averaging the four together. This
method does provide a smoother looking result.
Bicubic interpolation looks at all eight
neighboring pixels and computes a weighted
average to present the best results. The
only draw back is that it is examining so
much data that it requires more time and
greater computing horsepower.
There are a couple of other specialized
interpolation software packages, which keep
their mathematical calculations confidential. Genuine
Fractals and Extensis pxl SmartScale, believe
that their brand of interpolation is better
than that found in Photoshop. However,
both of them cost an additional $150-200
beyond what you already have in Photoshop. Be
certain to download test versions from the
web and test them before you decide to make
a purchase.
Why would anyone want to use an interpolation
(invented made up data) of his or her photo
files anyway? To create bigger prints,
that why. If you begin with a small
resolution file from a 2 or 3 mega-pixel
camera and desire an 11x14 inch print you
are either out of luck, get an ugly ressed-up
jaggy print or interpolate and hope for the
best.
For image printing, bicubic interpolation
is usually successful up to approximately
50 – 100% of the original file size
before too many artifacts are uncomfortably
visible. The success range may be dependent
upon how discerning the print requirements
are or the users eye is. Some claim
it is more advantageous to use Photoshop
to increase file size 10% at a time, until
the desired file size is reached, to prevent
artifact prominence (jaggedness). Under Image>Image
Size check the constrain proportions
and resample boxes and select bicubic for
best results.
If the constrain proportions box is left
unchecked you may, intentionally or surprisingly,
end up with an oddly shaped image. If
the image size was 10 x 8 inches and changed
to 20 x 8, with the box unchecked, the resultant
image would remain 8 inches wide but would
be transformed to a very slim 20 inches tall. It
is the special Defect, rather
than the special Effect,
way of making tall buildings.
If an excessive size image resolution is
sent to the printer software, most often
destructive interpolation occurs. This
means that the software cannot utilize all
of the sent data as the original file and
an irretrievable loss of data is experienced. The
printer software throws away random information
to adjust it’s self to a software predetermined
size limit, often differing in each printing
pass, resulting in inconsistent prints.
For more information on Resolution click
here.
If you want to be certain you know what
is in or what is out of your image file,
you must remain aware of interpolation. Is
interpolation doing useful good work or devastating
your day?
All the best from askRodger@pictureline.com
Figure 1
4x6” image at 72 dpi
constrained proportions & bicubic
interpolation
Figure 2
image example of destructive
interpolation
Figure 3
Image example of unconstrained
proportions
