Universal character segmentation scheme for multifont OCR images

Claim 2

In a processing system which presents a matrix of image data, with the matrix comprising rows and columns of binary pixels associated with a document having at least one field of characters thereon, a method of finding the binary pixels associated with a character included in said field comprising the steps: (a) selecting an examining window whose size covers a predetermined number of rows and columns of the pixels associated with a character in the image data; (b) calculating a probability density function, hereinafter referred to as PDF, for each pixel within the examining window for each character in a set of characters to be segmented to generate a composite PDF for each pixel within the examining window; (c) positioning the examining window over a portion of said field; (d) obtaining a total value for the examining window by using each binary one pixel in the examining window and its associated composite PDF; (e) moving said examining window relative to said field and repeating step (d) so as to obtain a peak or maximum total value for the examining window; (f) using the maximum total value obtained from step (e) as an indication that the examining window contains image data associated with a character in the set of characters; and (g) repeating steps (c) through (e) for the remainder of the image data associated with said field.

Claim 3

The method as claimed in claim 1 in which said moving step (e) is used for substantially centering the image data associated with a character within the examining window.

Claim 4

In a processing system which presents a matrix of image data, with the matrix comprising rows and columns of binary pixels associated with a document having at least one field of characters thereon, a method of segmenting the binary pixels associated with a character included in said field comprising the steps: (a) selecting an examining window whose size covers a predetermined number of rows and columns of the pixels associated with a character in the image data; (b) calculating a probability density function, hereinafter referred to as PDF, for each pixel within the examining window for each character in a set of characters to be segmented to generate a composite PDF for each pixel within the examining window; (c) positioning the examining window over a portion of said field; (d) obtaining a total value for the examining window by using each binary one pixel in the examining window and its associated composite PDF; (e) moving said examining window relative to said field and repeating step (d) so as to obtain a peak or maximum total value for the examining window; (f) using the maximum total value obtained from step (e) as an indication that the examining window contains image data associated with a character in the set of characters; (g) segmenting the image data which was included in the examining window when the maximum total value was obtained; and (h) repeating steps (c) through (g) for the remainder of the image data associated with said field.

Claim 5

In a processing system which presents a matrix of image data, with the matrix comprising rows and columns of binary pixels associated with a document having at least one field of characters thereon, a method of segmenting the binary pixels associated with a character included in said field comprising the steps: (a) selecting an examining window whose size covers a predetermined number of rows and columns of the pixels associated with a character in the image data; (b) calculating a probability density function, hereinafter referred to as PDF, for each pixel within the examining window for each character in a set of characters to be segmented to generate a composite PDF for each pixel within the examining window; (c) positioning the examining window over a portion of said field; (d) obtaining a total value for the examining window by using each binary one pixel in the examining window and its associated composite PDF; (e) moving said examining window relative to said field and repeating step (d) so as to obtain a peak or maximum total value for the examining window; (f) using the maximum total value obtained from step (e) as an indication that the examining window contains image data associated with a character in the set of characters; (g) segmenting the image data which was included in the examining window when the maximum total value was obtained; and (h) repeating steps (c) through (g) for the remainder of the image data associated with said field.

Claim 6

The method as claimed in claim 3 in which said selecting step (a) is effected by utilizing the pitch of the set of characters.

Claim 7

The method as claimed in claim 4 in which said positioning step is effected by utilizing the pitch of the set of characters.

Claim 8

The method as claimed in claim 3 in which said obtaining step (d) is effected by multiplying each binary one pixel in the examining window by its associated composite PDF.

Claim 9

The method as claimed in claim 3 in which said obtaining step (d) is effected by adding up the associated composite PDF for each binary one pixel in the examining window.

Claim 10

In a processing system which presents a matrix of image data, with the matrix comprising rows and columns of binary pixels associated with a document having at least one field of characters thereon, a method of segmenting the binary pixels associated with a character included in said field, comprising the steps: (a) selecting an examining window whose size covers a predetermined number of rows and columns of the pixels associated with a character in the image data; (b) presenting image data for a known character in a set of characters to the examining window, with a binary one pixel representing the presence of data and a binary zero pixel representing the absence of data; (c) storing the binary one and zero pixels in a memory according to row and column positions in the examining window for the known character from the presenting step (b); (d) repeating the presenting step (b) and the storing step (c) a predetermined number of times for the same known character so as to obtain a probability density function, hereinafter referred to as PDF, for each pixel in the examining window for the known character being presented; (e) storing the PDF for each pixel in the examining window for the known character being presented; (f) obtaining and storing according to steps (c) through (e) the PDFs for each pixel in the examining window for the remaining characters in a set of characters to be segmented; (g) summing the PDFs for each pixel in the examining window for all the characters in the set of characters so as to arrive at a total weighted value (hereinafter referred to as TWV) for each pixel in the examining window for the set of characters; (h) positioning the examining window over a portion of the matrix of image data including said field of characters; (i) obtaining a total value for the examining window by using each binary one pixel in the examining window and its associated TWV; (j) moving the examining window relative to the field so as to obtain a peak or maximum total value for the examining window; (k) using the maximum total value obtained from step (j) as an indication that the examining window contains image data associated with a character in the set of characters; and (l) segmenting the image data which was included in the examining window when the maximum total value was obtained.

Claim 11

In a processing system which presents a matrix of image data, with the matrix comprising rows and columns of binary pixels associated with a document having at least one field of characters thereon, a method of segmenting the binary pixels associated with a character included in said field, comprising the steps: (a) selecting an examining window whose size covers a predetermined number of rows and columns of the pixels associated with a character in the image data; (b) presenting image data for a known character in a set of characters to the examining window, with a binary one pixel representing the presence of data and a binary zero pixel representing the absence of data; (c) storing the binary one and zero pixels in a memory according to row and column positions in the examining window for the known character from the presenting step (b); (d) repeating the presenting step (b) and the storing step (c) a predetermined number of times for the same known character so as to obtain a probability density function, hereinafter referred to as PDF, for each pixel in the examining window for the known character being presented; (e) storing the PDF for each pixel in the examining window for the known character being presented; (f) obtaining and storing according to steps (c) through (e) the PDFs for each pixel in the examining window for the remaining characters in a set of characters to be segmented; (g) summing the PDFs for each pixel in the examining window for all the characters in the set of characters so as to arrive at a total weighted value (hereinafter referred to as TWV) for each pixel in the examining window for the set of characters; (h) positioning the examining window over a portion of the matrix of image data including said field of characters; (i) obtaining a total value for the examining window by using each binary one pixel in the examining window and its associated TWV; (j) moving the examining window relative to the field so as to obtain a peak or maximum total value for the examining window; (k) using the maximum total value obtained from step (j) as an indication that the examining window contains image data associated with a character in the set of characters; and (l) segmenting the image data which was included in the examining window when the maximum total value was obtained.

Claim 12

The method as claimed in claim 8 in which said selecting step (a) is effected by utilizing the pitch of the set of characters.

Claim 13

The method as claimed in claim 8 in which said positioning step (h) is effected by utilizing the pitch of the set of characters.

Claim 14

The method as claimed in claim 8 in which said obtaining step (i) is effected by multiplying each binary one pixel in the examining window by its associated TWV.

Claim 15

The method as claimed in claim 8 in which said obtaining step (i) is effected by adding up the associated TWV for each binary one pixel in the examining window.

Claim 16

The method as claimed in claim 12 is which said positioning step (h) is effected by first searching for the rows and columns in which said field is located so as to provide the location of said field of said field within said matrix of image data.