Browse Prior Art Database

Character Data Creation Using Convolution

IP.com Disclosure Number: IPCOM000099402D
Original Publication Date: 1990-Jan-01
Included in the Prior Art Database: 2005-Mar-14
Document File: 2 page(s) / 52K

Publishing Venue

IBM

Related People

Mano, T: AUTHOR

Abstract

Disclosed is an algorithm for creating statistical character data from a small number of samples using convolution. The statistical character data is a collection of the probabilities of pixels to be black, as shown in the figure. The convolution is very popular in post processing of image manipulation such as edge enhancement. The new method shows that it is also effective in the process of statistical data. The same result is obtained both from the gathering of a lot of samples and from the convolution method.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 76% of the total text.

Character Data Creation Using Convolution

       Disclosed is an algorithm for creating statistical
character data from a small number of samples using convolution.  The
statistical character data is a collection of the probabilities of
pixels to be black, as shown in the figure. The convolution is very
popular in post processing of image manipulation such as edge
enhancement.  The new method shows that it is also effective in the
process of statistical data.  The same result is obtained both from
the gathering of a lot of samples and from the convolution method.

      The new method is described as follows:  At first, gather
samples of characters.  Then a statistical data of characters is
established (see the figure).

      Next, take a convolution of the statistical data with a
convolution matrix.  More precisely, this process is done as follows:
 Let stt(i,j) (i = 1, 2, ..., u, j = 1, 2, ..., v) be
the above statistical data, mat(k,l) be a convolution matrix (k = -w,
-w+1, ..., w-1, w, l = -w, -w+1, ..., w-1, w).  Then a new
statistical data new(i,j) created from stt(i,j) using the convolution
matrix mat(k,l) is calculated by

                            (Image Omitted)

    w     w
    new(i,j) =  S     S      stt(i+s, j+t)*mat(s,t) / total
    t=-w  s=-w where
    w     w
    total =  S     S      mat(s,t)
    t=-w  s=-w
    Then the new data new(i,j) will be used in furt...