A next obvious improvement in prediction, but one which increases the complexity of the encoder, is prediction from a previous video frame.  This requires storage of a frame of video in the encoder for comparison with the present frame, and introduces an encoding latency of one frame time.  This technique  reduces the information to be sent for still images, but leaves significant data for moving images, even for a simple panning movement of the camera.  (In the case of still images, data can be sent only for the first instance of a frame, and all subsequent prediction error values are zero. However, it is desirable to retransmit the frame occasionally to allow receivers that have just been turned on to have a starting point.)  

NEXT - Motion Compensated Prediction