Ana səhifə

Contents Page


Yüklə 2.59 Mb.
səhifə16/35
tarix25.06.2016
ölçüsü2.59 Mb.
1   ...   12   13   14   15   16   17   18   19   ...   35

7.10 Data partitioning


Data partitioning is a technique that splits a video bitstream into two layers, called partitions. A priority breakpoint indicates which syntax elements are placed in partition 0, which is the base partition (also called high priority partition). The remainder of the bitstream is placed in partition 1 (which is also called low priority partition). Sequence, GOP, and picture headers are redundantly copied in partition 1 to facilitate error recovery. The sequence_end_code is also redundantly copied into partition 1. All fields in the redundant headers must be identical to the original ones. The only extensions allowed (and required) in partition 1 are sequence_extension(), picture_coding_extension() and sequence_scalable_extension().

NOTE - The slice() syntax given in 6.2.4 is followed in both partitions up to (an including) the syntax element extra_bit_slice.

The interpretation of priority_breakpoint is given in Table 7-30.

Table 7-30 Priority breakpoint values and associated semantics

priority_breakpoint

Syntax elements included in partition zero

0

This value is reserved for partition 1. All slices in partition 1 shall have a priority_breakpoint equal to 0.

1

All data at the sequence, GOP, picture and slice() down to extra_bit_slice in slice().


1

127


All syntax elements above, plus up to 64 (run, level) DCT coefficient pairs.



Figure 7-17 A segment from a bitstream with two partitions, with priority_breakpoint set to 64 (one (run, level) pair). The two partitions are shown, with arrows indicating how the decoder needs to switch between partitions.

Semantics of VBV remains unchanged, i.e. the VBV refers to the sum of two partitions, not any single one.

The bitstream parameters bit_rate (bit_rate_value and bit_rate_extension), vbv_buffer_size (vbv_buffer_size_value and vbv_buffer_size_extension) and vbv_delay shall take the same value in the two partitions. These parameters refer to the characteristics of the entire bitstream formed from the two partitions.

The decoding process is modified in the following manner:

Set current_partition to 0, and start decoding from bitstream that contains the sequence_scalable_extension (partition 0).

If current_partition = 0, check to see if the current point in the bitstream is a priority breakpoint.

If yes, set current_partition to 1. Next item will be decoded from partition. 1

Otherwise, continue decoding from partition 0. Remove sequence, GOP, and picture headers from both partitions.

If current_partition = 1, check the priority breakpoint to see if the next item to be decoded is expected in partition 0.

If yes, set current_partition to 0. Next item will be decoded from partition 0.

Otherwise, continue decoding from partition 1.

An example is shown in Figure 7-17 where the priority breakpoint is set at 64 (one (run, level) pair).


7.11 Hybrid scalability


Hybrid scalability is the combination of two different types of scalability. The types of scalability that can be combined are SNR scalability, spatial scalability and temporal scalability. When two types of scalability are combined, there are three bitstreams that have to be decoded. The layers to which these bitstreams belong are named in Table 7-31.

Table 7-31 Names of layers

layer_id

name

0

base layer

1

enhancement layer 1






For the scalability between the enhancement layers 1 and 2, the enhancement layer 1 is its lower layer, and the enhancement layer 2 is its enhancement layer. No layer can be omitted from the hierarchical ladder. E.g., if there is SNR scalability between enhancement layer 1 and enhancement layer 2, the prediction types in enhancement layer 1 are also valid for the combined decoding process for enhancement layers 1 and 2.

The coupling of layers is more loose with spatial and temporal scalability than with SNR scalability. Therefore, in these kinds of scalability, first the base layer has to be decoded and upconverted before it can be used in the enhancement layer. In SNR scalability, both layers are decoded simultaneously. The decoding order can be summarised as follows :

case 1 :

base layer



enhancement layer 1



enhancement layer 2

First decode the base layer, and then decode both enhancement layers simultaneously.

case 2 :


base layer

enhancement layer 1



enhancement layer 2

First decode the base layer and the enhancement layer 1 simultaneously, and then decode the enhancement layer 2.

case 3 :


base layer

enhancement layer 1



enhancement layer 2

First decode the base layer, then decode the enhancement layer 1, and finally decode enhancement layer 2.

1   ...   12   13   14   15   16   17   18   19   ...   35


Verilənlər bazası müəlliflik hüququ ilə müdafiə olunur ©atelim.com 2016
rəhbərliyinə müraciət