Classification Error Rate for Quantitative Evaluation of Content-based Image Retrieval Systems

A major problem in the field of content-based image retrieval is the lack of a common performance measure which allows the researcher to compare different image retrieval systems in a quantitative and objective manner. We analyze different proposed performance evaluation measures, select an appropriate one, and give quantitative results for four different, freely available image retrieval tasks using combinations of features. This work gives a concrete starting point for the comparison of content-based image retrieval systems. An appropriate performance measure and a set of databases are proposed and results for different retrieval methods are given.

• Type: Conference
• Year: 2004
• Where: ICPR
• Authors: Daniel Keysers, Hermann Ney, Thomas Deselaers
• Keywords: Computer Vision, Content-based Image Retrieval, ICPR 2004, Image Retrieval Systems, Image Retrieval Tasks