Experiments with Linear Prediction in Television
01 July 1952
Linear prediction is perhaps the most expedient elementary means of removing first order correlation in a television message. Before discussing the advantages and disadvantages of linear prediction, it might be well to consider what is generally meant by correlation in a television picture and why it should be removed. Almost every picture that has recognizable features contains both linear and non-linear correlation. Each type of correlation helps in identifying one picture from another; however, linear prediction is only effective in removing linear correlation, and for this reason, future references to correlation will refer only to its linear properties. With television, a signal is obtained as the result of scanning; hence, the cor764 LINEAR PREDICTION" IN TELEVISION 765 relation is evident in both space and time. Briefly, correlation is that relation which the "next," elemental part of the signal has with its past. To leave correlation in a message is to be redundant, and this effectively loads the transmission medium with a lot of excess "words" not necessary to the description of the picture at the receiving end. It is then more "efficient" to send only the information necessary to identify the picture, and to restore the redundancy at the receiver. EFFICIENT TRANSMISSION