Datasheet

Photoshop Elements has three methods of interpolation (methods of figuring out how
to insert new pixels or remove existing ones as you change the size of an image) and five
interpolation options. Nearest Neighbor, Bilinear, and Bicubic are the methods. Bicubic
Smoother and Bicubic Sharper are variations (various levels of sharpening) on the Bicubic
method that were added in Elements 3.
Nearest Neighbor When you resize using Nearest Neighbor interpolation, Photoshop
Elements adds or removes pixels based on pixel information and color that already exist
in the image. Whether upsampling or downsampling, there is no averaging of color and/or
tone to create new colors/tones. Nearest Neighbor is useful, for example, for controlled
upsampling of screenshots without blurring (quadrupling the pixel count can yield an
exact duplication of an image at four times the resolution). Multiply by squares (4, 9, 16,
etc.) to achieve controlled upsampling.
Bilinear Bilinear interpolation behaves much like Bicubic and is supposed to be faster, but
I’ve never clocked them. During the sampling, new tones and colors can be introduced
between existing colors that are not in the original image. This can blur sampling of hard
edges but can provide a smooth transition for tones (Nearest Neighbor might provide a
blockier, stepped result). One thing about Bilinear upsampling is that it does simple, true
averaging between neighboring tones and adds fewer new qualities to an image than any
type of Bicubic resampling. At times these interpolation properties prove to be an advan-
tage in retaining look and feel (when decreasing the image pixel count), and in others they
may result in softening (when increasing the image pixel count). Use Bilinear when you
want to downsample images.
Bicubic The Bicubic resampling process creates new image information by averaging, like
Bilinear, but goes one step further to provide a bit of sharpening to the result. The inten-
sity of the sharpening is stronger depending on the type of Bicubic interpolation selected.
The sharpening is intended to counteract the blurring result of averaging when increasing
the pixel count. Bicubic resampling changes a greater number of pixels with the same
radius setting as Bilinear but may generally give a better visible result in most cases than
Bilinear (when upsampling). This type of interpolation is the real workhorse for sizing
That trick you’ve seen on TV, where a pixilated image gets clearer and clearer as they zoom
in, is reverse engineered. You can never enhance image detail that has not been captured.
The only thing you can really do to reclaim image detail that you don’t already have is
reshoot a subject with higher resolution (e.g., using a longer lens, macro setting, or higher
pixel dimension to capture more detail) or rescan (assuming that the detail is present in what
you are scanning).
resizing images ■ 9
4456c01.qxd 3/1/06 3:04 PM Page 9