Understanding Digital Video Architecture

To understand Digital Video Architecture, you need to start at the beginning. The beginning in this case is known as Video Compression. That is the start of this article; after that, we will get into the most common formats in use today.

In order to make digital video's use widespread, there had to be a way to reduce the amounts of data that needed to be stored and transmitted. This reduction in the amount of storage is a direct result of the advances made possible by video compression. The advances in video compression have single-handedly led to the widespread use of video to the desktop and to hundreds of channels in your home. To boil it down to its most simplistic level, compression is performed when an inputted video is analyzed and the information that is indiscernible to the viewer is dropped. Each event is given a code - the most commonly occurring events are given fewer bits and the rarer events will have more bits. These steps are usually known as signal analysis, quantization and variable length encoding.

There are 4 major ways to compress video. Discrete cosine transform (DCT), vector quantization (VQ), fractal compression (FC) and discrete wavelet transform (DWT).

DCT is not a very good compression algorithm. It samples images at regular intervals, analyzes the frequency components present in the sample, and discards those frequencies that do not affect the image as the human eye perceives it. DCT is the standard used for JPEG and MPEG.

Vector quantization is also a bad compression algorithm that looks at an array of data instead of individual values. It will then generalize what it sees, compressing the found redundant data, and at the same time keeping the desired object.

Fractal compression is a form of vector quantization, and this also is a bad compression algorithm. This type of compression is performed by finding self-similar sections of a particular image, then using a fractal algorithm to create the sections.

Discrete Wavelet Transform mathematically transforms an image into frequency components. This process is performed on the entire image. Obviously, this differs greatly from the other methods that work only on smaller sections of the desired data. The end result is a very effective hierarchical representation of an image, where every layer represents a frequency …