2006 Fiscal Year Final Research Report Summary
Study of low-delay video encoding and decoding
| Project/Area Number |
15560326
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Communication/Network engineering
|
|---|
| Research Institution | Mie University |
|---|
Principal Investigator |
KONDO Toshio Mie University, Graduate School of Engineering, Professor (60324539)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
OHNO Kazuhiko Mie University, Graduate School of Engineering, Lecturer (20303703)
SASAKI Takahiro Mie University, Graduate School of Engineering, Assistant Professor (20362361)
|
|---|
| Project Period (FY) |
2003 – 2006
|
| Keywords | low delay / IDR / GDR / H. 264 / re-intra coding / rate control / intra slice / overflow |
|---|
| Research Abstract |
We proposed two delayed decoder refresh methods, named Delayed GDR and Delayed IDE, which reduce video decoding delay due to decoder buffer by double encoding of each refresh part and by allowing refresh delay. The Delayed GDR and the Delayed IDR encode each refresh part not only as a P slice or as a P picture for display but also as an I slice or as a re-intra picture of the local decoded picture of the P picture for delayed refresh, respectively.
These two proposed methods were implemented and evaluated on two software encoders, H. 264 JM reference encoder and mpeg2 v12 encoder.
The basic coding characteristics were evaluated by using the H.264 reference encoder. The result shows that Delayed IDR decreases buffering delay for a still scene about 100 ms or more in comparison with GDR which is the current most effective low-delay encoding method and it can provide compression efficiency as well as the GDR under the coding condition of M=1 though both the two proposed methods only provide only 80-96 % compression efficiency of the GDR under the coding condition of M=3, where M is the P picture period.
Another evaluation using mpeg2v12 encoder shows that the Delayed GDR decreases buffering delay of M=3 about 30 ms in comparison with GDR in the coding condition of practical CBR rate control causing no buffer overflow and no buffer underflow.
|
