A STUDY OF CHARACTER RECOGNITION PARAMETER ANALYSIS.
Abstract
An adaptive measurement technique for classifying the quality of printed text encountered in applications of optical print reading machines is presented. The measurement technique utilizes the learning properties of adaptive threshold logic. The convergence rate of the adaptive process is taken as a figure-of-merit for classifying print quality. Rapid convergence is associated with good quality printing while slow convergence is an indication of poor quality. A review of adaptive threshold logic is presented as an aid for describing the measurement technique. To implement the technique, threshold logic is simulated with the aid of a digital computer and highresolution images of actual printed text are used as input patterns. The measurement technique is verified and demonstrated by classifying various categories of printed text. The figure-of-merit measurements for each print category are catalogued with an accompanying representative print sample. It is expected that this catalog can be utilized as a comparative aid for estimating the figureof-merit of printed text encountered in applications of automatic print readers. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1965
- Accession Number
- AD0621577
Entities
People
- Harold S. Vitale
Organizations
- General Electric