Geodesically Reduced Image Primitive Space

J.M. Kinser (USA)

Keywords

Dimensionality reduction, contentbased image retrieval, image primitives, clustering

Abstract

In the pursuit of creating very large scale image databases it is necessary to drastically reduce the dimensionality of the representation of the image content. By creating a linked graph in which the vertices are common image objects and the geodesic edges are similarity measures it is possible to create a highly reduced representation of image content. This representation is developed here and demonstrated through a feasibility example.

Important Links:



Go Back