LAMP Seminar
Language and Media Processing Laboratory
Conference Room 4406
A.V. Williams Building
University of Maryland

Image Retrieval by Appearance

R. Manmatha
Multimedia Indexing and Retrieval Group
Center for Intelligent Information Retrieval
University of Massachusetts @ Amherst
Joint work with S. Ravela


The goal of image retrieval is to retrieve images "similar" to a given query image by comparing the query and database using visual attributes like color, texture and appearance. Here, we discuss how to characterize appearance and use it for image retrieval. We use two different approaches to image retrieval - using a local and a global representation of appearance.

For the local approach, visual appearance is represented by the outputs of a set of Gaussian derivative filters applied to an image. These outputs are computed offline and stored in a database. A query is created by outlining portions of the query image deemed useful for retrieval by the user (this may be changed interactively depending on the results). The query is also filtered with Gaussian derivatives and these outputs are compared with those from the database. The images in the database are ranked on the basis of this comparison. The technique has been experimentally tested on a database of 1600 images which includes a variety of images. The system does not require prior segmentation of the database. Objects can be embedded in arbitrary backgrounds. The system handles a range of size variations and viewpoint variations up to 20 or 25 deg.

Global representations of appearance may be compared by using histograms of suitable features. We describe an approach which compares histograms of curvature features. The main advantage of such global representations is the speed of retrieval.

For more information, contact Alice O'Dea at (301) 405-6444 or

home | language group | media group | sponsors & partners | publications | seminars | contact us | staff only
© Copyright 2001, Language and Media Processing Laboratory, University of Maryland, All rights reserved.