Concept Preserving Hashing for Semantic Image Retrieval With Concept Drift

Current hashing-based image retrieval methods mostly assume that the database of images is static. However, this assumption not true in cases where databases are constantly updated (e.g., on Internet) and there exists problem concept drift. The online (also known as incremental) hashing have been proposed recently for they considered drift problem. Moreover, update hash functions dynamically by generating new codes all accumulated data over time which clearly uneconomical. In order to solve these two problems, preserving (CPH) proposed. contrast existing methods, CPH preserves original concept, is, set representing a preserved time, learning yield same (old new) concept. objective function consists three components: 1) isomorphic similarity; 2) partition balancing; 3) heterogeneous similarity fitness. experimental results 11 scenarios show yields better precisions than does need previously stored images.