Cryptography Reference
In-Depth Information
MCTF
Inter-Layer Prediction
Quantization
Qp2
Non-Scalable
CABAC
Quality Enh.-Layer 2
Mo t io n
Compensation
+
In-Loop
Frame Buffer
Refinement
Out-Loop
Frame Buffer
MCTF
Inter-Layer Prediction
Quantization
Qp1
Non-Scalable
CABAC
Quality Enh.-Layer 1
Mo t ion
Compensation
+
In-Loop
Frame Buffer
Refinement
Out-Loop
Frame Buffer
Hiearchical Prediction
Quantization
Qp0
Non-Scalable
CABAC
Quality Base-Layer
Mot io n
Compens ation
{M.V}
+
In-Loop
Frame Buffer
Fig. 3.11.
Prediction structure of coarse granularity scalability.
the scaling of motion vectors are omitted since different quality layers have
the same resolution.
Although the same resolution is used for each layer, different quality layers
are optimized for different bit rates. For instance, if the base layer is coded
and optimized for the bit rate R1, the first quality layer will be coded at
a higher bit rate R2. In the same way, the next quality layer will be coded
for the bit rate R3 with R3 >R2 >R1. To achieve the target bit rate
for each layer, different quantization parameters are applied. In addition, the
redundancy between quality layers is removed by using inter-layer prediction.
Thus, each quality layer records the differences between the signals obtained
with two different quantization parameters. The prediction residues at each
