IMAGE PROCESSING TECHNIQUES AND METHODOLOGIES FOR OPTICAL CHARACTER RECOGNITION
Original Publication Date: 1991-Apr-01
Included in the Prior Art Database: 2001-Nov-30
This publication is in defense of the ideas/methodologies/graylevel digital image processing techniques employed in the automatic visual interpretation and recognition task pertaining to "Optical Character Recognition". The term Character is defined as an alpha-numeric symbol, manufacturers logo, or any other solid-connected contiguous blob uniquely recognizable by human capabilities. The characters can be darker or lighter than the back- ground, obscured by the presence of similar characters, and deformed or degraded significantly. The term Recognition is defined as the ability to identify or read automatically individual characters. Characters are identified by the presence of "excitatory features" and the absence of "inhibitory features". Features can be defined as gradient (combined or decoupled direction edges) or solid blob information - reduced to descriptors of relations amongst segments or single points, having assigned weighting. Excitakwy features are considered the critical segments or points that should to properly identify a character. Inhibitory features are considered the critical seiments or points that &&I notbe due to possible mis-identification or confusion with similai characters. The relative significance or importanbe of a feature is determined by weighting - more critical features having larger weighting, less critical features having ,smaller weighting. Critical requ,irements of feature selection and weighting is determined in conjunction with the entire character set of interest.