L
Lai-Man Po
Researcher at City University of Hong Kong
Publications - 207
Citations - 6191
Lai-Man Po is an academic researcher from City University of Hong Kong. The author has contributed to research in topics: Motion estimation & Search algorithm. The author has an hindex of 33, co-authored 199 publications receiving 5608 citations. Previous affiliations of Lai-Man Po include Hong Kong Applied Science and Technology Research Institute & University of Hong Kong.
Papers
More filters
Journal ArticleDOI
FEANet: Foreground-edge-aware network with DenseASPOC for human parsing
TL;DR: A novel framework called Foreground-Edge-Aware Network (FEANet) with DenseASPOC context module to further enhance the segmentation performance for human parsing and introduces the Dense Atrous Spatial Pyramid Object Context module to address the problem of small and ambiguous objects.
Proceedings ArticleDOI
Compensated sum of absolute difference for fast H.264 inter mode selection
TL;DR: In this article, a new compensated sum of absolute difference (CSAD) for fast H.264 inter mode selection algorithm is proposed to determine the best inter mode based on CSAD cost instead of the rate-distortion cost.
Journal ArticleDOI
The training of Karhunen---Loève transform matrix and its application for H.264 intra coding
TL;DR: An optimal frequency matching (OFM) algorithm is proposed to train K LT matrices for residual blocks and nine KLT matrices corresponding to nine prediction modes of 4 × 4 intra blocks are trained, showing that KLT with trained matrices yields a persistent gain over H.264.
Proceedings ArticleDOI
Web-based Beowulf-Class parallel computing on image database indexing and retrieval system
TL;DR: A Web-based image database indexing and retrieval system that uses fast CBIR with Beowulf-Class parallel computing engine (Super Abacus) to efficiently and effectively shorten the retrieval time.
Proceedings ArticleDOI
Multi-direction search algorithm for block-based motion estimation
TL;DR: A novel multi-directional gradient descent search (MDGDS) is proposed in this paper with use of multiple OTSs in eight directions that performs eight one-dimensional gradient descent searches on the error surface and therefore can trace to the global minimum more efficiently.