A Novel Matrix Norm Based Gaussian Kernel for Feature Extraction of Images

Gaussian kernel is widely used in Support Vector Machines and many other kernel methods, and it is most often deemed to provide a local measure of similarity between vectors, which causes large storage requirements and large computational effort for transforming images to vectors owing to its viewin...

Full description

Saved in:
Bibliographic Details
Main Authors: Li, Jun-bao, Chu, Shu-chuan, Pan, Jeng-shyang, Ho, Jiun-huei
Format: Conference Proceeding
Language:English
Subjects:
Automatic control
Automatic testing
Electronic equipment testing
Feature extraction
Image databases
Image storage
Kernel
Polynomials
Principal component analysis
Spatial databases
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Gaussian kernel is widely used in Support Vector Machines and many other kernel methods, and it is most often deemed to provide a local measure of similarity between vectors, which causes large storage requirements and large computational effort for transforming images to vectors owing to its viewing images as vectors. A novel matrix norm based Gaussian kernel (M-Gaussian kernel) which views images as matrices is proposed to solve the problem. Experiments conducted on ORL face database show the effectiveness of the proposed M-Gaussian kernel.
DOI:10.1109/IIH-MSP.2006.265004