User Guide

1. Hardware Implementation

This section describes the different internal modules of the H.264 Encoder. Data input to the H.264 Encoder must

be in the form of a raster scan image in the YCbCr 422 format. H.264 Encoder uses 422 formats as input and

implements compression in 420 formats.

The following figure shows the H.264 Encoder block diagram.

Figure 1-1. H.264 Encoder - Modules

1.1 Intra Prediction

H.264 uses various intra-prediction modes to reduce the information in a 4 x 4 block. The intra-prediction block in the

IP uses only DC prediction on 4 x 4 matrix size. The DC component is computed from the adjacent top and left 4 x 4

blocks.

1.2 Integer Transform

H.264 uses integer discrete cosine transform where the coefficients are distributed across the integer transform

matrix and the quantization matrix such that there are no multiplications or divisions in the integer transform. The

integer transform stage implements the transformation using shift and add operations.

1.3 Quantization

The quantization multiplies each output of integer transform with a predetermined quantization value defined by the

QP user input value. The range of QP value is from 0 to 51. Any value more than 51 is clamped to 51. A lower QP

value denotes lower compression and higher quality and vice versa.

1.4 Motion Estimation

The Motion Estimation searches 8 x 8 block of the current frame in the 16 x 16 block of the previous frame and

generates motion vectors.

Hardware Implementation

and its subsidiaries

User Guide

DS50003366B-page 5