Indexing of handwritten document images

An important problem in the management of scanned handwritten document image collections, is their indexing or retrieval based on word queries. This paper presents a method for fast localization of query words in handwritten images by an adaptation of the principle of geometric hashing. Specifically, a method of location hashing is presented that uses consecutive features along curves to produce small-sized image hash tables that also enable fast indexing. Handwriting variations are handled by assembling groups of word segments separated by inter-letter spacing, which is automatically estimated from sample pages written by an author. Results are presented that indicate the reduction in search as well as precision and recall possible with location hashing of handwritten words