Abstract
In hardware design of MPEG-2 to H.264 transcoder, MPEG-2 motion vector (MV) is often reused as search center to reduce the search range for H.264 motion estimation (ME) module. But MPEG-2 MV is not always accurate search center. The common-used comparison algorithm for motion vector predictor (MVP) is hardly implemented in hardware. In this paper, we propose a MV prediction algorithm for hardware design based on smoothness of MPEG-2 MV field. Experiment results show that the proposed MV prediction method with a small search range can approach the performance of full search algorithm with a large search range. Using the proposed algorithm, the motion prediction accuracy is significantly improved compared with the method only using MPEG-2 MV, especially in sequence with fast motion and complicate background.
| Original language | English |
|---|---|
| Title of host publication | Proceedings - 1st International Congress on Image and Signal Processing, CISP 2008 |
| Pages | 278-282 |
| Number of pages | 5 |
| Volume | 1 |
| DOIs | |
| Publication status | Published - 2008 |
| Event | 1st International Congress on Image and Signal Processing, CISP 2008 - Sanya, Hainan Duration: 2008 May 27 → 2008 May 30 |
Other
| Other | 1st International Congress on Image and Signal Processing, CISP 2008 |
|---|---|
| City | Sanya, Hainan |
| Period | 08/5/27 → 08/5/30 |
Fingerprint
ASJC Scopus subject areas
- Computer Vision and Pattern Recognition
- Signal Processing
Cite this
A hardware-oriented high precision motion vector prediction scheme for MPEG-2 to H.264 transcoding. / Wei, Xianghui; Tang, Wenming; Tian, Guifen; Ikenaga, Takeshi; Goto, Satoshi.
Proceedings - 1st International Congress on Image and Signal Processing, CISP 2008. Vol. 1 2008. p. 278-282 4566163.Research output: Chapter in Book/Report/Conference proceeding › Conference contribution
}
TY - GEN
T1 - A hardware-oriented high precision motion vector prediction scheme for MPEG-2 to H.264 transcoding
AU - Wei, Xianghui
AU - Tang, Wenming
AU - Tian, Guifen
AU - Ikenaga, Takeshi
AU - Goto, Satoshi
PY - 2008
Y1 - 2008
N2 - In hardware design of MPEG-2 to H.264 transcoder, MPEG-2 motion vector (MV) is often reused as search center to reduce the search range for H.264 motion estimation (ME) module. But MPEG-2 MV is not always accurate search center. The common-used comparison algorithm for motion vector predictor (MVP) is hardly implemented in hardware. In this paper, we propose a MV prediction algorithm for hardware design based on smoothness of MPEG-2 MV field. Experiment results show that the proposed MV prediction method with a small search range can approach the performance of full search algorithm with a large search range. Using the proposed algorithm, the motion prediction accuracy is significantly improved compared with the method only using MPEG-2 MV, especially in sequence with fast motion and complicate background.
AB - In hardware design of MPEG-2 to H.264 transcoder, MPEG-2 motion vector (MV) is often reused as search center to reduce the search range for H.264 motion estimation (ME) module. But MPEG-2 MV is not always accurate search center. The common-used comparison algorithm for motion vector predictor (MVP) is hardly implemented in hardware. In this paper, we propose a MV prediction algorithm for hardware design based on smoothness of MPEG-2 MV field. Experiment results show that the proposed MV prediction method with a small search range can approach the performance of full search algorithm with a large search range. Using the proposed algorithm, the motion prediction accuracy is significantly improved compared with the method only using MPEG-2 MV, especially in sequence with fast motion and complicate background.
UR - http://www.scopus.com/inward/record.url?scp=51849109016&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=51849109016&partnerID=8YFLogxK
U2 - 10.1109/CISP.2008.775
DO - 10.1109/CISP.2008.775
M3 - Conference contribution
AN - SCOPUS:51849109016
SN - 9780769531199
VL - 1
SP - 278
EP - 282
BT - Proceedings - 1st International Congress on Image and Signal Processing, CISP 2008
ER -