Combining MPEG-7 based visual experts for reaching semantics

Abstract

Semantic classification of images using low-level features is a challenging problem. Combining experts with different classifier structures, trained by MPEG-7 low-level color and texture descriptors is examined as a solution alternative. For combining different classifiers and features, two advanced decision mechanisms are proposed, one of which enjoys a significant classification performance improvement. Simulations are conducted on 8 different visual semantic classes, resulting in accuracy improvements between 3.5-6.5%, when they are compared with the best performance of single classifier systems. © Springer-Verlag Berlin Heidelberg 2003.