Survey of Dirac: A Wavelet Based Video Codec for Multiparty Video Conferencing and Broadcasting - Intelligent Video Event Analysis and Understanding - page 216

4 Overview of Dirac Video Codec
4.1 Dirac Architecture
Dirac is an open source video codec developed by BBC for video sizes ranging
from QCIF to HDTV progressive and interlaced. The major difference between
Dirac and other video codec is the transform coding algorithm. Dirac uses
Wavelet Transform instead of DCT Transform.
4.2 Dirac Encoder
In Figure 1, Dirac encoder architecture has been shown. It is very simple and
modular. There are four major modules of Dirac including Wavelet Transform,
Motion Estimation and Compensation, Entropy Coding and Quantization. Motion
Estimation and Compensation module uses Hierarchical Motion Estimation and
Compensation. Hierarchical Motion Estimation and Compensation is achieved by
down converting each frame of video by a factor of 2, hence finding motions
vector at different levels. Dirac performs Integral Error Calculations, Langrangian
Calculations and based on these calculations, it selects the best quantizer from 49
quantizers. Dirac uses Arithmetic coding algorithm, but this algorithm is also
context based and adaptive in nature. It first converts the number into binary form
and based on the context of binary symbol it also updates its probability table.
Fig. 3 Dirac encoder architecture
4.3 Dirac Decoder
In Figure 2, Dirac decoder architecture has been shown which performs the inverse
of encoder. The decoding process is carried out in three stages. In the first stage,
encoded bit-stream is decoded using entropy decoding, the next inverse scaling and
quantization is performed and after this inverse transform is performed.