Combining Intraframe and Frame-to-Frame Coding for Television

More than forty years ago it was first realized that channel capacity requirements could be significantly reduced by transmitting only those parts of a television signal that represent the changes from one frame of an image to the next.1 However, only recently technology has been 1137 available to store a complete frame of video information to enable such a system to become practicable. 2,3 In addition to the high correlation from frame to frame (temporal correlation), quite high correlation also exists from line to line and between adjacent elements along a line. It is these spatial forms of correlation which have been most widely exploited in coding television signals. For example, within a single frame we can switch between previous element prediction and previous line prediction, depending on whether there is more horizontal or more vertical similarity between adjacent picture elements.4 Similarly, in frame-to-frame coding the element in the previous frame corresponding to the element being encoded is a good prediction when an object is moving slowly, whereas a spatially adjacent element in the same frame is a better prediction of the current element when the object is moving fast. In an ideal situation, it is easy to determine the changeover point at which the element difference is smaller than the frame difference. Consider an image moving horizontally at a constant speed of one picture element per frame period (pef). This speed is quite slow; it would take about 8 seconds for an object to cross from one side of the screen to the other.