Large-scale cross-media mining
Emerging socical network services and various information obtained from smarphones and wearable devices enable us to capture real-world and web-world objects, events and contexts more clearly. Our research is now focusing on integrated investigations and enhancement of real and web world contents, objects, events and contexts, which we call "large-scale cross-media mining".
Computational model of visual attention
Currently my research effort focuses on salient region extraction
from images and videos. Especially, I am interested in constructing
computational models of the early human visual system.
- Computational model of temporal dynamics of early human
- Stochastic model of selective visual attention with dynamic
- Saliency-based video segmentation with sequentially updated
- Cognitive developmental approach to human-like visual scene
- Single image segmentation with estimated depth
- Visual attention driven by auditory cues
Generalized multivariate analysis
- Semi-supervised canonical correlation analysis
- Generalized multivariate analysis with generalized pairwise expression
Multimedia information retrieval
My research efforts have focused on quick and accurate retrieval from a huge number of multimedia signal archives.
- Very quick retrieval for audio and video archives
- Quick identification of partial images and videos
- Efficient semi-supervised learning of topic models and its
My recent studies include theoretical analysis of communication with
multiple senders, multiple channels and multiple receivers,
especially from the standpoint of data compression.
- Weak variable-length coding in Slepian-Wolf coding system
- Complementary delivery of information
- Shannon-theoretical analysis of index searching