Vector quantization is both a mathematical model and a technique for data compression, the goal of which is to minimize the transmission and storage rate for a communication system while retaining the best allowable fidelity to the original. Since the use of resources such as bandwidth will generally expand to meet the resources available, it will always be beneficial to apply data compression even though the cost of bandwidth and storage capacity drops steadily. One example is information flow on the World Wide Web (WWW). With the remarkably growing popularity of the WWW, demands on data distribution rise much faster than the speed of transmission channels. Data compression thus remains at the technical core of many communication systems. A brief look at a video transmission system demonstrates this fact. According to the H.263 video coding standard, every video image in the QCIF format contains 176 × 144 pixels. Each pixel is described by one luminance component and two chrominance components, each stored originally in 1 byte. Sin the chrominance components are subsampled every 2× 2 pixels, one image requires 176× 144 × 1.5 bytes, i.e., 304128 bits, to specify. For a video sequence with 30 frames per second, the transmission rate should be above 304128 × 30 ≈ 9.12 X 106 bits per second. Since for the most up-to-date modem, the transmission rate is about 56k bits per second, in order to view real time video sent by a web server, compression ratios of at least 9.12 x 106/5.6 × 104 ≈ 163 to 1 are needed.
KeywordsVector Quantization Lossy Compression Scalar Quantization Code Book Average Distortion
Unable to display preview. Download preview PDF.