Interpolation and decimation
Sometimes an image cannot be reproduced at the size needed.
It lacks the needed resolution (file size) or has too much for
practical use. In these cases the computer can be called upon
to modify the resolution of an image, reducing its resolution
or increasing it.
When an image is reduced in resolution, the process is called
decimation. The method for subsampling is under the best of circumstances
- harmless to the image The computer "scans" through
a line of pixels, casting off unnecessary information by averaging
together pairs of pixels, or groups of pixels to the average
of the Original group.
The resulting image is a reduced-size mirror of the original
image, faithful in tonality to the original, but smaller in size.
The process of decimation with Photo CD) images comes to play
in the following example. The photo is a 6 x 6 transparency that
has been scanned to Pro Photo CD Master disc. The reproduction
requirements call for an image with a resolution of 2800 pixels
in one dimension. The Base/64 resolution is 4096 pixels, almost
twice the file size needed. In fact, the image as opened from
the Photo CD disc in RGB mode is 45 MB. We need an image of only
23.5 MB, so the file needs to be subsampled by almost half.
The easiest way to subsample a file is to open the image in Adobe
Photoshop software and use the "Image Size" control
to reduce the resolution. Decimation is a mathematical process,
and can be involved. Faster processors, and those with math coprocessor
chips or digital signal processors, will accomplish this task
faster than will slower processors.
Adobe Photoshop's Image Size window is a powerful and highly
misunderstood control By checking the File Size Checkbox at the
bottom of the window, the proportions and dimensions of an image
can be modified when different resolution values are entered
into the Resolution space. This changes the "packing"
of digital data in to a future space. By Rechecking the File
Size box, one can make changes to an image's actual size, either
by decimation or interpolation. Such changes are sometimes harmful
to the nnsge - especially the interpolation option, which creates
new pixels when enlarging an image.
Interpolation is the opposite of decimation. The process is called
into force when there is not enough resolution in an image to
accomplish the reproduction needed. Interpolation can be a daunting
task for a computer, and will always take more time than decimation
of a similar amount, for the computer processor must calculate,
by means of analysis and averaging, the new pixel values and
colors that are inserted in an image.
When an image needs interpolation, the Photoshop program will
add resolution to the file in the necessary amount. Photoshop
software offers two different methods for interpolation: bipolar
and bicubic. One method "scans" the line of pixels
and interpolates horizontally only, while the bicubic method
actually "looks" around to all the adjacent pixels
and calculates a new pixel value based on what it sees in the
area around the new pixel. Bicubic interpolation results in a
vastly superior product, but takes over four times longer than
the linear process.
Is interpolation harmful to an image? Yes, and no. The process
of interpolation requires the computer to measure adjacent pixels
and "guess" a correct value for new pixels inserted
in the data stream. In the process, it can make errors, create
unwanted noise, and introduce abnormal pixels into areas of the
image causing what computer scientists call "artifacts."
The process is good enough, though, to produce an acceptable
image in most cases. The important thing to remember is that
any interpolation is potentially harmful to an image, and some
interpolation may ruin it.
If an image is interpolated to a small degree up to a doubling
of the image's original size - this is usually harmless. Beyond
two times the original size, you will begin to see serious effects
in most images.
When to stop? Trial and error will dictate the limits of interpolation
for your projects. Because printing processes, printing substrates
and intended viewing distances all contribute to the quality
of the final product, interpolation in one case will succeed
while it will cause failure in another.
