Discriminative Learing and Recognition of Image set classes Using Canonical Correlation

Abstract:

Canonical angles or principle angle between the two dimensional sub-space.

Canonical angle is compared with the two main classical methods: parametric distribution based and non parametric sample based.

The classical linear discriminate analysis which we have to develop is to be maximizing the correlation with-in the class set and it has to minimize between the class-set.

Image set after transforming the discriminate function are compared with the canonical correlation and also classical orthogonal sub-space method also compared with the proposed set.

The proposed set also used for compared with the ETH-80 database.


Canonical correlation
Each set is represented by a linear sub-space and the angle between two high-dimensional sub-spaces is exploited as a similarity measure of two sets.

HARDWARE SPECIFICATION
Processor : Any Processor above 500 MHz.
Ram : 128Mb.
Hard Disk : 10 GB.
Compact Disk : 650 Mb.
Input device : Standard Keyboard and Mouse.
Output device : VGA and High Resolution Monitor.

Software Specifications:-

Java 1.5

Swings