Compression

Images always contain large amounts of repetitive information. Typical certificates , for instance, contain large areas of plain paper not covered by either color or printing. Compression is a way of saving file space by representing repetitive information with special codes. You can imagine image compression being like saying, "samples 887 through 921 in row 74 are all light purple."

Different file formats use different compression routines, although they fall into two categories. Lossless file formats faithfully keep all information during compression; lossy formats judiciously discard some information in order to control file size. Every computer program that uses those formats must contain matching "decompression" routines in order to decode files for display and use.

Browser display

A browser is a program that decodes information sent over the internet. Currently, the most popular browsers are (in order) Microsoft Internet Explorer, Mozilla Firefox, Google Chrome, Apple Safari, and Opera. Browsers designed for mobile devices now constitute about 5% of the browser market. Browser programs are constantly being updated, with later versions always exceeding previous versions in their capabilities. Huge numbers of old versions remain in use a decade after initial release.

Bit depth

Bit depth is a measurement of how accurately color or shades of gray can be stored. (See my discussion about Understanding Bits.) It represents how much digital information is used to describe color or shades of gray for every sample measured by a scanner. 1-bit accuracy can record only black or white. 8-bit accuracy can record 256 shades of gray or color. 24-bit accuracy (three color channels using 8 bits each) can record 16.8 million different colors. 48-bit accuracy can store 2.8 billion different colors.

Color model

Entry-level scanners commonly save color information in two ways. 'RGB' (red, green, blue) is the most common color model. Each color (or channel) is saved using 8 bits of memory. Therefore, each color can be described by 256 shades. By combining 256 possible shades of red, green and blue, a total of 24 bits of memory is used for every spot sampled and can record up to 16.8 million colors. Conveniently, color monitors can represent an equal number of colors using red, green and blue light.

A simpler, but less accurate method of recording color is to use a color table. Essentially, a computer program samples images and calculates an ideal selection of 256 colors out of the 16.8 million possible. Each of these 256 colors is given an index number. Next, every sampled spot on the image is assigned one of those 256 numbers based on the closest indexed color. Using a mere 256 colors, each sampled spot requires only 8 bits of memory and so file sizes are much reduced. Or course, color accuracy, is greatly decreased.

While monitors use red, green and blue light to show images, printers use pigmented dyes and inks. Printing with red, green and blue ink would create a sludge-colored mess. In order to create colored images on paper, printers use cyan, magenta and yellow (CMY) pigments. Printed together, CMY pigments have the ability to recreate the same colors as RGB light. This clever phenomena works because cyan is the spectral opposite of red, magenta the opposite of green, and yellow the opposite of blue. Professional offset printers and many ink jet printers add black to the mix to because cyan, magenta and yellow inks rarely create nice, bold blacks. BlacK is abbreviated as 'K' and the full color model becomes CMYK. Many better scanners allow images to be directly saved using CMY or CMYK color models.

LAB is a color model used in higher-priced image editing programs, but is preserved in its native form in only one major file format. 'L' stands for luminance and 'A' and 'B' are two other color channels (Channel A ranges from green to red and B ranges from blue to yellow.) Editing in the LAB color space takes much experience. I've never encountered a scanner that saves images in the LAB color model.

Transparency

When displaying images on color monitors, it is often desirable to make parts of images transparent. Some formats allow one color to be turned transparent and others use a special transparency channel.

Interlacing

As the internet gained popularity through the mid- and late-1990s, web site developers realized that users really hated waiting for images to download. If images took too long to appear when connecting to the web over ordinary phone lines, users would simply give up and move to different sites. In response, image formats were modified in order for web sites to display partial images very quickly and then gradually improve those images as more information flowed across the wires. Gradual image appearance was achieved by interlacing digital information as it was stored in image files. Interlacing was an absolute necessity with old 56K modems. Curiously, today's various browser programs do NOT always recognize interlacing. Today's broadband cable and DSL connections have pretty much eliminated the need for interlacing.

Metadata

Modern digital devices can large amounts of information in addition to images. Therefore, most file formats have been modified to record extraneous information such as creator's name, date, time of day, device settings, originating device type, even image location. This data is called metadata.

TIF (TIFF) = Tagged Image File Format

The TIF format is the most widely used image format among professionals. The format has either minimal or no limits to pixel size, resolution, or bit depth. TIFs can be saved either in non-compressed form or compressed using LZW, ZIP or JPEG compression. Some programs have extended the TIF standard to accept transparency and multiple layers, although many programs ignore those features during use. The format is widely touted as being 'lossless.' I prefer to say that the save, re-open process is lossless. All images deteriorate (or at least change) when they are rotated, re-sized, re-sampled, and distorted. TIF images are not used on the web.

JPG (JPEG) = Joint Photographic Experts Group

JPGs (pronounced 'j-peg') are the most prominent file format used on the web and among non-professionals. The format is superb at saving space. The amount of compression is easily configurable. Professionals advise against saving in JPG format until after all image manipulation and color adjustment is finished. They say this because the format is 'lossy' and images lose resolution every time they are saved. Quality loss depends on the quality of imaging programs, the number of saves, resolution, and the degree of changes made during each open-save cycle.

JPG 2000

The JPG 2000 format is an improvement upon the JPG standard with added lossless compression and alpha channel transparency capabilities. Unfortunately, even eleven years after development, support for the format is still spotty, both among browsers and among image manipulation programs.

GIF = Graphics Interchange Format

When developed in the late 1980s, GIF images were pretty much restricted to the Compuserve system. Since then, GIF has become one of the de facto standard formats for use on the World Wide Web. Because the format relies on indexed colors, it is not recommended for storing high-resolution color. GIFs work very well for graphics that display areas of solid color; GIFs works less well or poorly on photographs displaying wide spectrums of shades. The developer of the format claimed that GIF was meant to be pronounced 'Jif', like the peanut butter. Practically everyone I encounter pronounces 'GIF' like 'gift' without the final 't'.

PNG = Portable Network Graphics

The PNG format was developed for web graphics. It was specifically designed to emulate or improve upon features of the GIF format while avoiding patent issues with the LZW compression routine used in GIFs. It is a well-designed, full-featured format decoded by all major web browsers. Unfortunately, support for all PNG features is spotty. When using PNGs for web display, be sure to test all browsers your audience is likely to use. Developers indicated that PNG should be pronounced 'ping.' I hate to irritate anyone, but I almost always hear it pronounced 'puh-nung'.

BMP = Bitmap Image File

This format is a patent-free file standard popularly developed by Microsoft and used by many Windows and OS/2-based operating systems. While the format can be effectively compressed by typical 'Zip' programs, native BMP files tend to be much too large for easy transmittal over slow web connections. While the format can be opened, manipulated and saved by a large number of Windows-based image manipulation programs, it does not seem superior to the widely-accepted TIF format. The BMP format never caught on among professionals and seems destined to remain that way.

PCX = Personal Computer eXchange

This format was developed by the ZSoft Corporation (now defunct) and was the native format for the PC Paintbrush program. A decade or more ago, it was a heavily-used DOS format but has since been greatly superceded by TIF, JPG and GIF standards. The format is very 'fluffy', meaning file sizes are overly large.