Texas Instruments DM365, Руководство пользователя

Страница: 20 / 131

Introduction
1-8
Some important H.264 profiles and their special features are (These are
feature as defined by H.264 standard, few of them may not be part of
DM365/DM368 H.264 implementation):
‰
Baseline Profile:
o
Only I and P type slices are present
o
Only frame mode (progressive) picture types are present
o
Only CAVLC is supported
o
ASO/FMO and redundant slices for error concealment is supported
‰
High Profile:
o
Only I, P, and B type slices are present
o
Frame and field picture modes (in progressive and interlaced modes)
picture types are present
o
Both CAVLC and CABAC are supported
o
ASO is not supported
o
Transform 8x8 is supported
o
Sequence scaling list is supported
o
B frames and weighted prediction.
The input to the encoder is a YUV sequence, which can be of format 420
with the chroma components interleaved in little endian. The output of the
encoder is an H.264 encoded bit-stream in the byte-stream syntax. The
byte-stream consists of a sequence of byte-stream NAL unit syntax
structures. Each byte-stream NAL unit syntax structure contains one start
code prefix of size four bytes and value 0x00000001, followed by one NAL
unit syntax structure. The encoded frame data is a group of slices, each is
encapsulated in NAL units. The slice consists of the following:
‰
Intra coded data: Spatial prediction mode and prediction error data,
subjected to DCT and later quantized.
‰
Inter coded data: Motion information and residual error data
(differential data between two frames), subjected to DCT and later
quantized.
The first frame is called Instantaneous Decode Refresh (IDR) picture
frame. The decoder at the receiving end reconstructs the frame by spatial
intra-prediction specified by the mode and by adding the prediction error.
The subsequent frames may be intra or inter coded.
In case of inter coding, the decoder reconstructs the bit-stream by adding
the residual error data to the previously decoded image, at the location
specified by the motion information. This process is repeated until the
entire bit-stream is decoded.
In motion estimation, the encoder searches for the best match in the
available reference frame(s). After quantization, contents of some blocks
become zero. H.264 Encoder tracks this information and passes the
information of coded 4x4 blocks to inverse transform so that it can skip
computation for those blocks that contain all zero co-efficients and are not
coded.