TABLE OF CONTENTS A key element in MPEG inter compression is motion compensation. In inter compression the pels in a region of a reference picture (or pictures) are used to predict pels in a region of the current picture. Differences between the reference picture and the current picture are then coded to whatever accuracy is affordable at the desired bitrate. If there is motion, some parts of the image used for prediction may be shifted relative to the current image, and motion compensation is used to minimize the effects of these shifts.
In this chapter we look at motion compensation from the decoder's point of view. The motion vector data elements have already been decoded from the bitstream as described in Chapter 8, and now must be converted into horizontal and vertical motion displacements.
Chapter 12 describes how the motion displacements are used in the pel reconstruction process. Motion estimation, the process by which the encoder selects these motion vectors, is discussed in Chapter 13.
The pictures in moving picture sequences often have large regions in which very small changes occur from one picture to the next in the sequence. The picture-to-picture correlation in these regions is very high and any reasonable video compression scheme should take advantage of this. Obviously, there are usually some changes from picture to picture, or we would only have to send the first picture of the sequence. Changes are caused by a number of different mechanisms, among them noise, scene lighting,...
