Troubleshooting Photos –
About Re-sizing Photos 2 – Why
Why the Recipe works
Digital photos are composed of pixels, a word which is short for picture element (the fellow who invented the term may not have been fussy about spelling, but that's alright because picel is hard to say and probably would have morphed into pixel eventually anyway.)
Pixels are pure information, individual little collections of data about tone and color at a particular point in a photo, that only take physical form when they are presented on your monitor, or by your printer on a piece of paper.
The main thing to remember about pixels is that they are discrete points of information, so unlike a traditional negative, you can't enlarge a digital image by 'stretching' it optically the way you could with a film negative (actually you can, sort of – but the results with photos aren't pretty – the photo may end up with either obvious pixellation or it may look soft and ill-defined).
When you enlarge a digital image, the process begins with expansion of the matrix on which the existing pixels are located. The illustration below shows a 4 x 4 block of pixels that is to be enlarged 400 per cent (depending on content, this is about the greatest possible enlargement by normal means). This initial step leaves great voids between the existing pixels, which must be filled if the image is to look good.
The next move is up to you. When you open the image re-sizing dialog (Photoshop: Image >> Image Size ... ; Elements: Image >> Resize >> Image Size... ), you'll see three items at the bottom of the box (this process is also described on a page devoted to enlarging):
- a checkbox labelled "Resample Image" (which must be checked if you want the real size of the image altered);
- another labelled "Constrain Proportions" (may be called "Preserve Aspect Ratio" or similar – it tells the program to calculate a new second dimension in proportion to the first so the image preserves the original ratio of width to height).
- a set of options for enlarging and reducing size, including "Nearest Neighbor", "Bilinear", "Bicubic", "Bicubic Smoother" and "Bicubic Sharper".
The last bullet point refers to options for interpolation, a math-based process for deriving the color and tone of new pixels from the ones that are already there. In an enlargement, these will lie in the blank spaces between the original pixels. In a size reduction, the new pixel will replace those that are being removed. The more information you allow your application to use in interpolation, the better the results will be.
- "Nearest Neighbor" interpolation simply looks at the color and tone of the pixel nearest the one it wants to create. The effect is simply to make all the existing pixels larger, with little or no reference to adjacent pixels (which is why Nearest Neighbor is sometimes called "Pixel Resize"). When used for photos, it produces results like this pixellated steamer. For this reason, Nearest Neighbor is only suited to reproducing hard-edged graphics like the little view camera at the bottom of the page.
- "Bilinear" interpolation looks at the color and tone of a 2 x 2 group of pixels, and creates a pixel between them that averages the color and tone of the four original pixels. For photos, it's a marked improvement over Nearest Neighbor, but compared to Bicubic, its results are rough and prone to artifacts.
- "Bicubic" interpolation considers the color and tone of a 4 x 4 group of pixels. The new pixel created at the center gives greater weight to the pixels close to the new pixel, less weight to the farther pixels. It takes a little longer, but gives the best results of all the enlargement and reduction routines included with image-editing programs.
- You may also see the options, "Bicubic Smoother" and "Bicubic Sharper". These simply add additional processing steps after interpolation is complete to achieve smoother enlargements or sharper reductions. Many commentators think these steps are best done after you've had a chance to look at the picture and decide how much smoothing or sharpening is necessary, but if you like the results, use them.
This illustration shows what happened to the 16 pixels shown above when the photo was enlarged by different interpolation routines. nearest shows "Nearest Neighbor" interpolation, aka "Pixel Resize", which just makes the pixels larger. The obvious pixels make photos jagged and unappealing. bicubic shows bicubic interpolation, which does some sophisticated averaging to keep the smooth flow of color you expect in a photo. bilinear shows bilinear interpolation, which renders more quickly than bicubic but is less smooth and prone to creating artifacts; fast contemporary computers have made it obsolete.
A real-world example cropped from the photo above after it was enlarged 400% by various interpolation routines. nearest has an unacceptable case of jagged pixellation. bilinear is smoother, but bicubic is sufficiently better that it's worth the brief extra processing time required by its more complex averaging. Note the better definition of edges and the increased clarity of colors, whether the muted earth tones of the background field or the bright red of the jacket.There is a new class of interpolation routines known as "adaptive" because they first analyze the photo for edges of tone and color, then apply different interpolation routines optimized for edges, areas of smooth color, blends, etc. Generally, normal interpolation is good up to 400%; adaptive interpolation can take you quite a bit farther (as ever, 'how far' depends on the image and your expectations, but so far none of them can perform like the computers in movie thrillers that spot bad guys a kilometer away in snapshots taken with a camera phone). All adaptive routines are sold separately and often include both "pro" (for making billboards) and "home" (less expensive) versions. They include:
The list is not exhaustive. This is a field where a lot of new ideas are being developed so new products appear frequently. Google a phrase like 'enlarge photos' if you're interested. As always, try before you buy – run the trial on a few of the shots you think you'd like to enlarge before you reach for your credit card.
