Face recognition using a similarity-based measure in image database for crime investigation

A method to handle searching a face image database (FID) is proposed to support police officers when searching criminal records from a central registration database system (CRDS). The proposed method assumes that each FID consists of a fixable object and object correlation. The proposed method employs a database search, so that all images with a similarity-based measure are retrieved. Consequently, the proposed method is much faster than sequential searching, especially when an additional set of attributes, like scar, is defined. Moreover it requires less storage space.