Algorithms of DVC - Distributed Multiple Description Coding - page 151

HDVC Encoder HDVC Decoder

Decoded
WZ
frames

WZ
frames

WZ bits

'

C

C

D

ʴ

ʴ D SW-SPIHT

encoding

SW-SPIHT
decoding

DWT -1 Refining W '

W DWT

-

y y

W

C C

W

re re

DWT

D

MC
Interpolation

WA
Interpolation

W

Y ʴ - WA

Interpolation

re

Intra-frame
encoding

Intra-frame
decoding

K

K '

Key
frames

Decoded
Keyframes

Fig. 5.23 HDVC framework used in the MDDVC

sequence using optimized 1-D DCT zero-padding [ 42 ].Andthenbymeansof
interleaving-splitting, the new sequence is divided into two descriptions, which
can be compressed by any DVC scheme. Here, a proposed hybrid DVC (HDVC)
scheme is employed for its efficient compression performance and low-complexity
encoding. At the decoder, firstly, the central decoder merges the two decoded
descriptions in case of no loss, or the side decoder uses error concealment method
to estimate the lost description and create the concealed big-size video in loss case,
and then in the post-processing, the inversed 1-D DCT zero-padding is implemented
to recover the pixels. The details of HDVC, preprocessing, and post-processing are
shown in the following two sections, respectively.

5.6.3 Hybrid DVC

To each description, an independent DVC codec is used to its big-size sequence.
In view of low-complexity encoding and high rate-distortion performance, a new
HDVC codec combining residual DVC [ 18 ] and Slepian-Wolf set partition in
hierarchical trees (SW-SPIHT) algorithm [ 32 ] is proposed. The codec including
HDVC encoder and decoder is shown in Fig. 5.23 .

5.6.3.1 HDVC Encoder

In each description, the frames are classified as the key frames and Wyner-Ziv
frames according to the coding method adopted. To the key frames, the intra-
frame encoding of H.264/AVC is used. While to the other frames, the Wyner-Ziv
ones, the following processes are implemented. Firstly, the weighted average (WA)