MPEG-2 over ATM

This paper overviews the opportunities that the broadband-ISDN provides for future multimedia services, the problems posed by this new network technology and how these are being addressed in the international standardisation and consensus forming arenas. The basic principles of the video coding algorithm of MPEG-2 are the same as those for MPEG-1 and H.261. The algorithm is a hybrid of motion compensated temporal prediction and discrete cosine transform (DCT) based coding. Temporal redundancy, present in stationary areas and regions with linear motion, is removed by subtracting from the input picture a prediction made from offset parts of previously coded pictures. The resulting prediction error is transformed to the frequency domain using the DCT. Although the DCT does not of itself give any data reduction, the values from it are better suited to quantisation, and hence compression. The resulting quantised data, together with the displacement vectors used in the prediction process and various overheads, are variable-length coded for transmission