An experimental study on the performance of visual information retrieval similarity models

Abstract

This paper is an experimental study on the performance of the two major methods for macro-level similarity measurement: linear weighted merging and logical retrieval. Performance is measured as the average query execution time for a significant number of tests. The two models were implemented in the standard version (as they are applied in a number of prototypes) and in an optimized version. The results show that optimized logical retrieval clearly outperforms optimized linear weighted merging.