Prototype Page Reader Design Study.
Abstract
The past fifteen years have produced significant achievements in the field of Optical Character Recognition (OCR). Today there exist a number of commercially-available OCR systems which are capable of reading a wide variety of type fonts, provided that the input is composed high quality material prepared in a cooperative environment. In terms of human recognition, the restrictions implied by the term high quality material must be considered severe. The machine reject rates on degraded material simply do not correspond to those of the human. To some degree, the performance differential between man and machine is attributable to the machine's incomplete use of available information. This report describes the results obtained on a research project to evaluate a recognition technique for typeset Latin (English) and Cyrillic (Russian) alphabetic characters which have been scanned and digitized at comparatively high spatial resolution and high grey-level quantization by the LIPS scanner. A normalized correlation technique was employed as the recognition algorithm, at the suggestion of the Government. This technique allows recognition in severe noise conditions such as might occur with poor quality printing.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 1976
- Accession Number
- ADA033537
Entities
People
- Earl W. Caldwell
- Edward G. Fisher
- Frank H. Feng
- George E. Forsen
- Paul C. Kane