Notes on "Lossiness"

Compression formats, whether they operate on audio, video, images, or random collections of files, are either lossless or lossy. The distinction is simple: Lossless formats are identical to the original(s) after being decompressed, while lossy formats are not. A good example of a lossless compression format is the ubiquitous . zip archiving scheme. When you unpack a zip archive containing a backup of your system from last month, losing even a single byte is unacceptable. However, some types of data can withstand having information thrown away, on the grounds that either you'll never notice what's missing, or you're willing to make a compromise: Smaller files in exchange for missing but unimportant data.

A good example of a lossy compression format is JPEG, which banks on the fact that image files often store more information than necessary to display an image of acceptable quality. By throwing away some of the information, and by encoding redundant information with mathematical algorithms, excellent compression ratios can be achieved for images that don't need to be displayed at high resolutions.

While the JPEG analogy doesn't depict the MP3 compression process accurately, it does illustrate the concept of lossiness, and it's important to understand that all MP3 files, no matter how well-encoded, have discarded some of the information that was stored in the original, uncompressed signal.

Many lossy compression formats work by scanning for redundant data and reducing it to a mathematical depiction which can be "unpacked" later on. Think for a moment of a photograph depicting a clear blue sky, and below it a beach. If you were to scan and store this image on your hard drive, you could end up storing hundreds of thousands of pixels of perfect blue, all identical to one another, and therefore redundant. The secret of a photographic compression method like GIF is that this redundant information is reduced to a single description. Rather than store all the bits individually, they may be represented as the mathematical equivalent of "repeat blue pixel 273,000 times." When the part of the image depicting the sand is encountered, the sand is analyzed for redundancy and similar reductions can be achieved. This is why simple images can be stored as small files, while complex images don't compress as well-they contain less redundancy. On the other hand, JPEG compression works in accord with user-defined "tolerance thresholds"; determining how similar two adjacent pixels (or, more accurately, frequencies) have to be before they're considered redundant with one another is the key to determining the degree of lossiness. If JPEG compression is set high, light blue and medium blue pixels may be treated as being redundant with one another. If JPEG compression is set low, the codec will be more fussy about determining which pixels are redundant. The end result will be a clearer picture and a larger image file.

Next:  Masking Effects