Video Compression - Signal Processing Systems - page 52

a b

Fig. 5 4 modes for luma blocks in H.264/AVC ( a )13

boundary pixels and the predictive directions for different modes; ( b ) Examples of the prediction
data with real images

Illustration of nine kinds of Intra 4 ×

2.2.2 Spatial Prediction

Here we introduce a spatial prediction technique called intra prediction which is
adopted in H.264/AVC. Intra prediction explores the spatial similarity in a region
with a regular texture. It utilizes the boundary pixels in the neighboring pre-coded
blocks to predict the current block. Illustration of Intra 4 × 4 modes in H.264/AVC

isshowninFig. 5 . Each MB is partitioned into 16 4 × 4 blocks. Each 4 × 4block

is predicted by the 13 left and top boundary pixels. There are 9 Intra 4 4 modes

with different predictive directions from the neighboring pixels. The mode with the
prediction data most similar to the current block will be chosen. There is another
scheme call Intra 16

×

16 modes
are suitable for regions with more/less detailed texture. Please refer to [ 25 ] forthe
detailed introduction of intra prediction in H.264/AVC.

× 16 as shown in Fig. 6 . In general, Intra 4 × 4/16 ×

2.2.3 Coding Structure

In a video coding system, a video sequence is partitioned into Group-Of-Pictures
(GOPs). Each GOP contains three kinds of frames—I-frames (Intra Coded Frames),
P-frames (Predictive Coded Frames), and B-frames (Bidirectionally Predictive
Coded Frames). For I-frames, only spatial prediction is supported. For P-frames and
B-frames, both spatial and temporal prediction is supported. P-frames can only use
the preceding frames for temporal prediction (a.k.a forward prediction). B-fames
can use both the preceding and the following (backward prediction) frames for
temporal prediction. In general, B-frames can achieve the best coding performance
because the most data are utilized for prediction. However, B-frames require more