The character recognition method described here is based on the principle of incoherent spatial matched filtering. The input to this matched filter is not the unknown character itself, but its Fraunhofer diffraction pattern. The intensity distribution in this diffraction pattern is insensitive against shifting of the unknown character, avoiding the need for character registration. The incoherent matched filter is easier to implement than the coherent matched filter, since only binary rather than continuous-tone masks are required. The theory and some experiments will be discussed and compared with other optical character recognition methods.
© 1965 Optical Society of America
Equations on this page are rendered with MathJax. Learn more.