he most important difference between analogous real-world imagery and digital pictures is the element-based construction of a digital image. No matter what you see on a digital screen, it always consists of an absolute collection of picture elements, better known as pixels. A digital photograph is essentially a mosaic existing of sampled color values from a real-life scene.

BITMAP VERSUS VECTOR
There are two main categories of digital imagery: bitmap graphics and vector graphics. Bitmap graphics consist of an absolute collection of pixels representing an image, while vector graphics show a pixelized representation of visual math. To say it in an alternative way: a bitmap circle consists of an absolute number of pixels that are arranged in a circular fashion, while a vector circle is an infinitely sharp circle defined by a formula and translated into a given number of pixels. Each of the two mentioned categories comes with its own advantages and disadvantages. The vector approach offers the delight of resolution independency: you can enlarge a vector graphic as much as you want and it will only become more detailed because there will be more pixels to display the vector graphics. Macromedia Flash and Adobe Illustrator are well-known examples of vector-based image software.

The mathematically defined nature of vector graphics can also be a disadvantage. If you take a picture with a digital camera the camera can't mathematically analyze the scene you're capturing and translate it into a full-fledged vector scene. The camera can't recognize all shapes captured by the camera and intelligently turn everything into formulas on the fly. That's why vector graphics are usually created from scratch on the computer and not captured from the real world, while digital photography is confined to bitmap graphics. Adobe Photoshop and Corel Paint Shop Pro are both bitmap-oriented.

A major disadvantage of a bitmap is its absolute nature. If you want to enlarge a bitmap, the computer can not conjure up details that are not saved along with the bitmap file. A bitmap that is enlarged from 16 x 16 pixels to 32 x 32 pixels will transfer the lower detail of the smaller image to the larger image size, although the image could be much more detailed because the new image size includes four times as much pixels as the old image size (twice the old width and twice the old height).

Detail of a pure black and white skull scetch bitmap (© SevensHeaven.nl).

SCALING METHODS
The most basic traditional type of image size reduction or enlargement is the "Nearest Neighbour" method, also known as the "Pixel Resize" method. This results in a straightforward averaged size alteration according to the available image information and the new resolution. Nearest Neighbour works best when you upscale an image by 200 percent or downscale it by 50 percent, because that yields the best averaging results. A better, more computation-demanding scaling method is Bicubic Interpolation, which is the default method in Adobe's Photoshop. Bicubic interpolation yields great results when downscaling, because it will calculate the result on a sub-pixel level, accurately determining the most appropriate values and positions of remaining pixel information. However, when enlarging an image with Bicubic Interpolation the method will fill the stairway-like gaps between the original pixels using transitional color values, resulting in usually unwanted blurriness. An interpolation method that is slightly better than Bicubic Interpolation is Lanczos (named after a Hungarian mathematician), but usually enlarging images with regular bitmap scaling techniques will result in loss of sharpness and exaggeration of pixel squares.

The pure black and white skull bitmap enlarged to 200 percent with Photoshop using the default Bicubic Interpolation method. Notice the jagged pixel edges and some slight blurring between the blown-up pixel data.

The Dutch company Shortcut has introduced a very interesting innovative approach that is based upon the best known interpolation methods for image enlargement and adds intelligent image analysis such as advanced edge detection. Shortcut's approach is based upon the S-Spline algorithm and implemented into their flagship software package Photozoom Pro. Photozoom Pro offers a very clear interface without a confusing collection of functions.

S-SPLINE
Although Photozoom Pro offers a range of different interpolation methods to choose from, such as Lanczos and Catmull-Rom, the star of the show is without a doubt Shortcut's own S-Spline algorithm. There is a substantial difference between S-Spline and conventional image scaling algorithms. S-Spline is an adaptive method that intelligently analyzes edges and color transitions and calculates a weighted interpolation between pixels, using a process that is similar to vectorizing bitmap data. But S-Spline is not the same as a bitmap to vector conversion like you can find in Flash or Illustrator. First of all, S-Spline uses its vector-like algorithm for internal calculations only, as Photozoom Pro works purely in bitmap space. Secondly, the S-Spline method takes more into account than advanced edge detection only: Photozoom Pro also analyzes dynamics and noise in a bitmap in order to get the most out of the source data.

The resize controls in Photozoom Pro's interface.

The S-Spline method offers a number of functions to refine the result of an image enlargement. Next to unsharp masking controls to remove blurriness in the result you can also control the amount of strength and sensitivity for the S-Spline calculations. Higher values result in a more sharp-edged, vector-like result. Very high values can result in a loss of photographic realism because of too much forced edge sharpness. A general rule of thumb is that the crisper your image is (such as clearly edged drawings), the higher you can set Photozoom Pro's strength and sensitivity values. Another interesting option in Photozoom Pro is the artificial detail function. When enlarging imagery with lots of fine detail such as grass and hair, the artificial detail function tries to reconstruct such detail by adaptively refining the image data in the higher resolution using appropriate pixel patterns.

The pure black and white skull bitmap enlarged to 200 percent using Photozoom Pro's S-Spline algorithm with maximum strength and sensitivity settings. Notice the vector-like edge sharpness.

THE ROUNDUP
Photozoom Pro lives up to its name: it's a skilled companion when it comes to image enlargement offering more than just a straightforward solution. However, Photozoom Pro doesn't seem to offer more than Photoshop when it comes to downscaling imagery. As a matter of fact an image that is reduced in size actually looks better using Photoshop's Bicubic Interpolation than any available Photozoom Pro method. But for image enlargement Photozoom Pro is clearly doing a better job than Photoshop. Options are available for batch processing of multiple images and Photozoom Pro comes with a convenient Photoshop plug-in that adds an export option to Photoshop, so you directly send an image from Photoshop to Photozoom Pro for enlargement. Photozoom Pro's manual is as clear as the software itself, but is a bit concise. An explanation of the different available interpolation methods along with example imagery would be a welcome addition.

Photozoom Pro offers great value for a sharp price and should be in the tool arsenal of any graphic designer who needs to enlarge images for purposes such as printed matter.

Metin Seven

www.metinseven.com
www.sevensheaven.nl
www.figurefarm.com