作者: Masami Shishibori , Satoru Tsuge , Zhang Le , Minoru Sasaki , Yoshiki Uemura
关键词: Feature (computer vision) 、 Similarity (geometry) 、 Image (mathematics) 、 Earth mover's distance 、 Computer science 、 Computation 、 Artificial intelligence 、 Histogram 、 Multimedia information retrieval 、 Image retrieval 、 Data mining 、 Pattern recognition
摘要: Earth Mover’s Distance (EMD) is a distance measure between two distributions, and have been widely used in multimedia information retrieval systems, especially content-based image systems. When the EMD applied to problems based on color or texture, reflects human perceptual similarities. Its computations, however, too expensive use large-scale databases. In order achieve efficient computation of during query processing, we developed “fastEMD”, library for high-speed feature-based similarity retrievals large This paper introduces techniques that are implementation fastEMD demonstrates efficiency extensive experiments.