Recognition kernel for content-based search

A search based on image contents is an important issue in a digital library which usually relies on large image and video databases. A technique for content-based search (CBS) using a multi-level recognition kernel is presented. Features of model objects are extracted at levels that are most appropriate to yield only the necessary yet sufficient details. Together they form the kernel. It is shown that the deployment of the recognition kernel in a pyramidal (multiresolution) framework facilitates multi-level abstraction of the model and hence improves the matching efficiency and quality. Preliminary experimental result from a small image database is presented