I might as well tackle the topic of interpolation because you're going to e-mail me and ask about it anyway. Questions about scanner interpolation are the second-most-common queries (after resolution) that I have received in the last decade.
Interpolation isn't limited to scanners, of course. You would think so only because scanner vendors like to trumpet their products' interpolation prowess in their specifications lists. Interpolation is a process used to create new pixels any time an image is made larger or smaller or changed to a color depth other than the one at which it was captured.
Photoshop uses interpolation all the time. If you enlarge or reduce an image, Photoshop has to create more pixels (when you enlarge) or fewer pixels (when you reduce) to represent the same image. Photoshop also needs to come up with a new set of pixels when you change an image from, say, 24 bits at full color to 8 bits and 256 colors. Within Photoshop, interpolation is almost always a bad thing (even if it's a necessary thing) because you always lose some quality when your image editor obliterates real, actual pixels and replaces them with best-guess estimates. You should, therefore, try to retain your image at its original resolution (such as the resolution it was taken at with your digital camera) until you absolutely have to resize it. And always keep an unedited version of the file to fall back on, should you need to!
Sometimes, the term interpolation is reserved for making larger images from a given resolution. You'll see the term downsampling to describe the type of interpolation used to produce a smaller image. I'm not particularly fussy about terminology that's not often used correctly anyway, so I use interpolation to mean both enlarging and reducing images or scans.
When scanning, interpolation isn't quite as evil because the scanner is working with the original image and can avoid some of the quality loss that plagues your image editor.
Some sort of scanner interpolation takes place whenever you scan at anything other than the native resolution of the scanner. That means if you're using a 1200 spi scanner and scan at 2400 spi, the scanner actually captures 1200 samples per inch but creates 1200 new pixels to make the resolution seem higher. The same scanner at 300 spi must figure out which 300 pixels per inch best represent the full 1200 spi scan.
Actually, when scanning in even fractions or increments of the scanner's optical resolution, the results are usually quite good. The problems start when you try to scan at odd resolutions like 5000 spi or 332 spi.
As the scanner interpolates an image, it doesn't simply duplicate pixels (say, doubling all the pixels in a 1200 spi scan to produce a 2400 spi image) or throw away pixels (to create a 600 spi scan). Instead, the interpolation algorithm examines the pixels surrounding the new pixel and produces one with characteristics that closely match the transition between the pixels. That is, if one pixel is dark tone and the next pixel is light tone, a medium-tone pixel is created to insert between them, as shown in Figure 1-4.
Interpolation is not all smoke and mirrors. A good interpolation algorithm can accurately calculate the pixels that would have been captured by the scanner if it had been able to scan at the interpolated resolution used. That is, you do get additional information through interpolation. However, there are diminishing returns. A 9600 x 9600 spi interpolated image is not eight times as sharp as one scanned at a true 1200 spi resolution.
Following are the most common interpolation algorithms, which have official names:
♦ Nearest Neighbor: With the Nearest Neighbor algorithm, the new pixels are created by examining the pixel nearest where the new pixel will go. This works fine with line art, in which the pixels change in predictable ways. However, the method doesn't allow for the fine and random gradations of tones found in photographs.
♦ Bilinear: This algorithm examines the pixels on either side of the target pixel and produces better quality images than the Nearest Neighbor method, although it is a little slower.
♦ Bicubic: The slowest commonly used method is also the best. Bicubic resampling uses sophisticated formulas to calculate the new pixel based on the pixels found above, below, and to either side of the new pixel.
Photoshop CS2 and the latest version of Elements have two new interpolation options: Bicubic Smoother and Bicubic Sharper. These can be used to fine-tune the most popular interpolation algorithm, making it (you guessed it) either smoother or sharper than the unadorned Bicubic version.
Book VII Chapter 1
Was this article helpful?
To begin with your career in photography at the right path, you need to gather more information about it first. Gathering information would provide you guidance on the right steps that you need to take. Researching can be done through the internet, talking to professional photographers, as well as reading some books about the subject. Get all the tips from the pros within this photography ebook.