Cryptography Reference
In-Depth Information
2.3.4 Forward Prediction and Bi-Directional Prediction
Motion compensated prediction based on the past frame is basically unidirec-
tional and is thus called forward prediction. Such predicted frames are des-
ignated as P-frames. Compared to Intra coded frames (I-frames), P-frames
allow a higher data compression.
However, when an object moves, it conceals the background at its leading
edge and reveals the background at its trailing edge. The newly revealed
background will not have reference in previous frame. Similarly, when the
camera pans, new areas come into view and nothing is known about them
previously. Instead, we can find their reference in the frames after it, i.e., its
future frames. This is possible if the frames are reordered and transmitted
so that the future frames are received before the current frame. This type of
prediction based on future and past frames is called bi-directional predictive
coding.
Bi-directional predictive coding allows the current frame to be predicted
from frames before and after the current frame, which are then interpolated
to form the prediction for the current frame. The frame generated from bi-
directional predictive coding is called the B-frame. B-frames can also be com-
pressed through a prediction derived from a previous frame or a future frame
alone. Similar to P-frames, the prediction error of the B-frame is further com-
pressed by DCT based Intra coding. Fig. 2.14 shows the diagram of forward
predictive coding (P-frame) and bi-directional predictive coding (B-frame).
