cca with applications.ppt

典型相关分析及其应用(CCA with Its Applications)
Chen Songcan (陈松灿)
Sun Tingkai (孙廷凯)
s.******@nuaa.
.
南京航空航天大学
Problems and Feature
Problems: Image retrieval
[HSS2004] (NC)
Images retrieved for the text query: “at phoenix sky harbor on July
6, 1997. 757-2s7, n907wa phoenix suns taxis past n902aw teamwork America west America west 757-2s7, n907wa phoenix suns taxis past n901aw Arizona at phoenix sky harbor on July 6, 1997.” The actual match is the middle picture in the first row. (text  image)
Computer vision
detecting pixels correlated to sound A on audio & video [ESE2005] (CVPR05) (Audio Video)
Image segmentation
by expert
A
[LGD2005](PR)
by LDA
Pixels
 Pixels
Image processing
Demosaicing
Object: color bands R, G, B
[Yac2004] (HP report)
Mosaiced
Original
Processed result
The demosaicing problem deals with the reconstruction of a full color image f from its partial sampling m.
Color  Color
Classification (1)
Binary classification [GSB2001] (ICANN)
X (Pattern)  Y (Label)
Classification (2)
Multi-class recognition
[Yo2004]
Samples from 4 classes
Separation between classes
Pose estimation
In this case, CCA is more suitable for clear description of manifold than PCA, and result shows A outperforms PCA in estimation accuracy. [MRB2003] (PR) (Image  Pose)
Sensor Localization
(Sensor signal  location)
[PKY06] . Pan, . Kwok, Q. Yang, Y. Chen, Multidimensional Vector Regression for Accurate and Low-Cost Location Estimation in puting, IEEE Transaction on Knowledge and Data Engineering, 18(2006):1181-1193.