Abstract | ||
---|---|---|
We propose a fast motion search algoritlm for H.264 motion estimation with variable block sizes. Motion estimation of H.264 encoder has larger Computation complexity than existing video compression standards such as MPEG-4 and H.263 since search steps increase in proportion to the number of block sizes. Here, we employed two-stage motion estimation. In the first stage, 8x8 to 16x16 motion estimation is conducted where only limited areas are searched in 8x16, 16x8, and 16x16 block sizes using the results of 8x8 block search. Then in the second stage, 4x4/8x4/4x8 block size motion search is performed only when 8x8 block mode is chosen in the first stage. Using the above two approaches. limited search area and conditional smaller block size search. very fast and highly accurate variable block size motion estimation has been achieved. The experimental results also confirmed that the proposed algorithm can greatly reduce computational complexity While fully using seven block sizes and maintaining motion estimation efficiency. |
Year | Venue | Keywords |
---|---|---|
2004 | ICIP: 2004 INTERNATIONAL CONFERENCE ON IMAGE PROCESSING, VOLS 1- 5 | computational complexity,motion estimation,video compression,data compression,search algorithm |
Field | DocType | ISSN |
Search algorithm,Computer science,Real-time computing,Artificial intelligence,Jump search,Motion estimation,Block size,Pattern recognition,Algorithm,Encoder,Data compression,Image compression,Computational complexity theory | Conference | 1522-4880 |
Citations | PageRank | References |
3 | 0.51 | 3 |
Authors | ||
4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Tomoyuki Shimizu | 1 | 49 | 2.96 |
Akio Yoneyama | 2 | 117 | 17.49 |
Hiromasa Yanagihara | 3 | 33 | 7.26 |
Yasuyuki Nakajima | 4 | 147 | 23.31 |