Cryptography Reference
In-Depth Information
Coding Cycle 1
Coding Cycle 2
Coding Cycle 0
Bit-plane 7
Bit-plane 6
Bit-plane 5
Bit-plane 4
Bit-plane 3
Bit-plane 2
Bit-plane 1
Bit-plane 0
Run
AC
1
AC
2
AC
2
DC
0
AC
1
AC
2
DC
0
AC
1
Block 0
Block 1
Block 2
Block 3
End-of-Block (EOB) = 0
ZigZag Scan
Raster Scan
Fig. 3.13.
Illustration of cyclical block coding.
Table 3.2.
Comparison of fine granularity scalability and coarse granularity scala-
bility.
FGS
CGS
Quantization
Smaller step size Larger step size
Motion Vector of Quality Layers
Shared
Independent
Prediction Mode of Quality Layers
Shared
Independent
Prediction of MCTF Residues
Yes
Adaptive
Entropy Coding of Quality Layers
Scalable
Non-Scalable
coe
cient. To further reduce the bit rate, all the symbols are coded by a
binary arithmetic coder.
In addition to using a Scalable Entropy Coding, Table 3.2 highlights other
differences between FGS and CGS. In FGS, the quality layers are generated
by successive quantization with each succeeding smaller step size. The quality
layers of FGS use the same motion vector and prediction mode. The inter-
layer prediction of MCTF residues is always enabled. Since the Motion Vector
and Prediction Mode are not refined along with the increase of bit rate, the
FGS is preferable for the applications with a smaller range of scalability. The
flexibility of CGS is suitable for the requirement of larger scalable range. In
recent studies [14], it has been proved that applying Adaptive Refinement of
Motion Vector and Prediction Mode in FGS can further improve the PSNR
by 1 dB. Thus, in the latest joint draft of SVC, the FGS maintains the same
flexibility as CGS for better coding e
ciency [14]. Multiple prediction loops
may be created in the quality enhancement-layer.
