Dot-matrix video enhancement for optical character recognition

• US 5,182,778 A
• Filed: 08/31/1990
• Issued: 01/26/1993
• Est. Priority Date: 08/31/1990
• Status: Expired due to Term

• Current Assignee: Global 360, Inc. (Open Text Corporation)
• Original Assignee: Eastman Kodak Company
• Inventors: Rudak, Peter, Hadgis, George A.
• Primary Examiner(s): Moore, David K.
• Application Number:
  US07/575,715
• Time in Patent Office: 879 Days
• Field of Search: 382/54, 382/51, 382/50
• US Class Current: 382/267
• CPC Class Codes: G06T 5/20 using local operators
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/20 Combination of acquisition, preprocessing or recognition functions

Abstract

A method and apparatus for eliminating discontinuities inherent to binary digital image data when hard copy text is generated by a dot-matrix printer. An M×M matrix is moved over a bit-mapped image of said text one pixel at a time, and for each pixel position, pixels are selectively converted from white to black based on a predetermined set of rules.

Claims

1. A method for use in an apparatus for improving optical character recognition of printed draft quality dot-matrix characters of an image to be scanned, comprising the steps of:

scanning an image with an electronic scanning device and generating signals representing the scanned image as rows and columns of pixels;

sending the signals of the scanned image to a binary digital filter that performs the following steps;

locating kernel of a first integer, M, M pixel rows and a second integer, N, N pixel columns in particular position in said scanned image where M and N are integers greater than two;

modifying the scanned image by moving said kernel between adjacent pixel positions and applying the following set of rules;

a) if two black pixels are separated by no more than N-2 white pixels in the horizontal direction, convert signals representing the white pixels to black;

b) if two black pixels are separated by no more than M-2 white pixels in the vertical direction, convert signals representing the white pixels to black;

c) if two black pixels are separated by no more than M-2 white pixels in the first diagonal direction, convert signals representing the white pixels to black;

d) if two black pixels are separated by more than M-2 white pixels in the second diagonal direction convert signals representing the white pixels to black;

applying the above steps to all pixels in said image;

outputting the modified image for optical character recognition.

2. The method as set forth in claim 1 wherein the white pixels are used as a reference and black pixels are converted to white if the predefined conditions have been met.

3. A method for use in an apparatus for improving character recognition of printed draft quality dot-matrix characters of an image to be scanned, comprising the steps of:

scanning an image with an electronic scanning device an generating signals representing a scanned image as rows and columns of pixels;

sending the signals of the scanned image to a binary digital filter that performs the following steps;

locating a kernel of M pixel rows and N pixel columns in a particular position in said image where M and N are integers greater than two;

moving said kernel to the next adjacent pixel position and applying the following set of rules;

a) if two black pixels are separated by no more than M-2 white pixels in the horizontal direction and the two end pixels in the vertical direction are not black, convert signals representing the white pixels to black;

b) if two black pixels re separated by no more than N-2 white pixels in the vertical direction and the two end pixels in the horizontal direction are not black, convert signals representing the white pixels to black;

c) if two black pixels are separated by no more than N-2 white pixels in either diagonal direction are not black, convert signals representing the white pixels to black;

applying the above steps to all pixels in the image;

outputting the modified image for optical character recognition.

4. A method for use in an apparatus for improving the optical character recognition of printed draft quality dot matrix characters of an image to be scanned, comprising the steps of:

scanning an image with an electronic scanning device and generating signals representing the scanned image as rows and columns of pixels;

sending the signals representing the scanned image to a binary digital filter that performs the following steps;

locating a kernel of M pixel rows and M pixel columns in a particular position in said image where M is an integer greater than two;

modifying the scanned image by moving said kernel to the next adjacent pixel position and applying the following set of rules;

a) if two black pixels are separated by no more than M-2 white pixels in either the horizontal or vertical direction, convert signals representing the intervening white pixels in that direction to black;

b) if two black pixels are separated by no more than M-2 white pixels in either diagonal direction, convert signals representing the intervening white pixels in that direction to black;

applying the above steps to all pixels in the image;

outputting the modified image for optical character recognition.

5. The method as set forth in claim 4 wherein said rules further include the use of qualification pixels which are end pixels located midway on the side of the kernel and perpendicular to either the horizontal or vertical direction being considered for pixel conversion, and also located in opposing corners of the kernel and perpendicular to the diagonal direction being considered for pixel conversion;

said qualification pixels, for each direction must be white before pixel signal conversion can take place in the direction orthogonal to a line connecting said qualification pixels.