Share Email Print

Journal of Electronic Imaging

Evaluation metric of an image understanding result
Author(s): Baptiste Hemery; Helene Laurent; Bruno Emile; Christophe C. Rosenberger
Format Member Price Non-Member Price
PDF $20.00 $25.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

Image processing algorithms include methods that process images from their acquisition to the extraction of useful information for a given application. Among interpretation algorithms, some are designed to detect, localize, and identify one or several objects in an image. The problem addressed is the evaluation of the interpretation results of an image or a video given an associated ground truth. Challenges are multiple, such as the comparison of algorithms, evaluation of an algorithm during its development, or the definition of its optimal settings. We propose a new metric for evaluating the interpretation result of an image. The advantage of the proposed metric is to evaluate a result by taking into account the quality of the localization, recognition, and detection of objects of interest in the image. Several parameters allow us to change the behavior of this metric for a given application. Its behavior has been tested on a large database and showed interesting results.

Paper Details

Date Published: 19 February 2015
PDF: 14 pages
J. Electron. Imag. 24(1) 013030 doi: 10.1117/1.JEI.24.1.013030
Published in: Journal of Electronic Imaging Volume 24, Issue 1
Show Author Affiliations
Baptiste Hemery, ENSICAEN (France)
Helene Laurent, Univ. d'Orléans (France)
Bruno Emile, Univ. d'Orléans (France)
Christophe C. Rosenberger, ENSICAEN (France)

© SPIE. Terms of Use
Back to Top