Fuzzy aggregation operators in region-based image retrieval

摘要: We examine the effect of fuzzy aggregation operators on image retrieval performance, by empirically comparing 67 operators, applied to problem computing similarity, given a collection feature similarities regions. While majority existing similarity models express as an similarities, no study presents systematic comparison different operators. compare by: (1) incorporating each operator into hierarchical, region-based model, which expresses region and corresponding similarities; (2) evaluating obtained model(s) five test databases, containing 64,339 general-purpose images, in 749 semantic categories. Results show that performance strongly depends operator(s) incorporated model - difference average precision between best worst performing is up 50%.