An image database for handwritten text recognition research is described that contains digital images of approximately 5000 city names, 5000 state names, 10000 ZIP Codes, and 50000 alphanumeric characters to overcome the limitations of earlier databases.