Handwritten digit normalization method

Claim 1

A method for normalization of a character image, comprising the steps of: providing a camera capable of sensing a two-dimensional image and a programmed computer configured to receive an image sensed by said camera; transferring written material containing a character to be normalized into the field of view of said camera; developing a pixelized version of said character image, then, using said programmed computer; determining a skew slope corresponding to the slant angle of the stroke pixels of the character image including: calculating the standard deviation σ Y of the vertical dimension of the stroke pixels of the character image, calculating the standard deviation σ X of the horizontal dimension of the stroke pixels of the character image, calculating the correlation CorrXY of the vertical and horizontal dimensions of the stroke pixels of the character image, and calculating the skew slope of the stroke pixels of the character image employing the following formula ##EQU7## determining the bounds of the smallest parallelogram which completely encloses all of the stroke pixels of the character image having a first pair of opposed horizontal sides and a second pair of opposed sides having said skew slope including: determining a central point of the stroke pixels of the character image, determining a skew axis having a slope equal to said skew slope and passing through said central point of the stroke pixels, determining the leftmost stroke pixel of each horizontal row of pixels, determining the first horizontal distance between the top-left corner of the leftmost stroke pixel of each horizontal row of pixels to said skew axis, determining the second horizontal distance between the bottom-left corner of the leftmost stroke pixel of each horizontal row of pixels to said skew axis, determining the maximum of said first and second horizontal distances, setting the left side of the parallelogram as a line parallel to said skew axis disposed to the left of said skew axis by said maximum of said first and horizontal second distances, determining the rightmost stroke pixel of each horizontal row of pixels, determining the third horizontal distance between the top-right corner of the rightmost stroke pixel of each horizontal row of pixels to said skew axis, determining the fourth horizontal distance between the bottom-right corner of the rightmost stroke pixel of each horizontal row of pixels to said skew axis, determining the maximum of said third and fourth horizontal distances, setting the right side of the parallelogram as a line parallel to said skew axis disposed to the right of said skew axis by said maximum of said third and horizontal fourth distances; and mapping the stroke pixels of said parallelogram into the standard size horizontal row and vertical column pixel dimensions.

Claim 2

A method for normalization of a character image, comprising the steps of: providing a camera capable of sensing a two-dimensional image and a programmed computer configured to receive an image sensed by said camera; transferring written material containing a character to be normalized into the field of view of said camera; developing a pixelized version of said character image, then, using said programmed computer; determining a skew slope corresponding to the slant angle of the stroke pixels of the character image including: calculating the standard deviation σ Y of the vertical dimension of the stroke pixels of the character image, calculating the standard deviation σ X of the horizontal dimension of the stroke pixels of the character image, calculating the correlation CorrXY of the vertical and horizontal dimensions of the stroke pixels of the character image, and calculating the skew slope of the stroke pixels of the character image employing the following formula ##EQU7## determining the bounds of the smallest parallelogram which completely encloses all of the stroke pixels of the character image having a first pair of opposed horizontal sides and a second pair of opposed sides having said skew slope including: determining a central point of the stroke pixels of the character image, determining a skew axis having a slope equal to said skew slope and passing through said central point of the stroke pixels, determining the leftmost stroke pixel of each horizontal row of pixels, determining the first horizontal distance between the top-left corner of the leftmost stroke pixel of each horizontal row of pixels to said skew axis, determining the second horizontal distance between the bottom-left corner of the leftmost stroke pixel of each horizontal row of pixels to said skew axis, determining the maximum of said first and second horizontal distances, setting the left side of the parallelogram as a line parallel to said skew axis disposed to the left of said skew axis by said maximum of said first and horizontal second distances, determining the rightmost stroke pixel of each horizontal row of pixels, determining the third horizontal distance between the top-right corner of the rightmost stroke pixel of each horizontal row of pixels to said skew axis, determining the fourth horizontal distance between the bottom-right corner of the rightmost stroke pixel of each horizontal row of pixels to said skew axis, determining the maximum of said third and fourth horizontal distances, setting the right side of the parallelogram as a line parallel to said skew axis disposed to the right of said skew axis by said maximum of said third and horizontal fourth distances; and mapping the stroke pixels of said parallelogram into the standard size horizontal row and vertical column pixel dimensions.

Claim 6

A method for normalization of a character image, comprising the steps of: providing a camera capable of sensing a two-dimensional image and a programmed computer configured to receive an image sensed by said camera; transferring written material containing a character to be normalized into the field of view of said camera; developing a pixelized version of said character image, then, using said programmed computer; determining a skew slope corresponding to the slant angle of the stroke pixels of said character image; determining the bounds of the smallest parallelogram which completely encloses all of the stroke pixels of the character image having a first pair of opposed horizontal sides and a second pair of opposed sides having said skew slope; and mapping the stroke pixels of said parallelogram into the standard size horizontal row and vertical column pixel dimensions including calculating the length of one of the sides of said parallelogram parallel to said skew axis, calculating a vertical scaling factor from the quotient of the standard vertical column pixel dimensions divided by the length of one of the sides of said parallelogram parallel to said skew axis, calculating the width of said parallelogram, calculating a horizontal scaling factor from the quotient of the standard horizontal column pixel dimensions divided by the width of said parallelogram, determining each stroke candidate pixel within the standard size horizontal row and vertical column pixel dimensions corresponding to any part of a stroke pixel of the character image, dividing each stroke candidate pixel into an array of subpixels, counting the number of subpixels of each candidate stroke pixel which map into any part of a stroke pixel of the character image, and setting a candidate stroke pixel as a stroke pixel if and only if the number of subpixel mapping into a stroke pixel of the character image exceeds a predetermined number.

Claim 7

A method for normalization of a character image, comprising the steps of: providing a camera capable of sensing a two-dimensional image and a programmed computer configured to receive an image sensed by said camera; transferring written material containing a character to be normalized into the field of view of said camera; developing a pixelized version of said character image, then, using said programmed computer; determining a skew slope corresponding to the slant angle of the stroke pixels of said character image; determining the bounds of the smallest parallelogram which completely encloses all of the stroke pixels of the character image having a first pair of opposed horizontal sides and a second pair of opposed sides having said skew slope; and mapping the stroke pixels of said parallelogram into the standard size horizontal row and vertical column pixel dimensions including calculating the length of one of the sides of said parallelogram parallel to said skew axis, calculating a vertical scaling factor from the quotient of the standard vertical column pixel dimensions divided by the length of one of the sides of said parallelogram parallel to said skew axis, calculating the width of said parallelogram, calculating a horizontal scaling factor from the quotient of the standard horizontal column pixel dimensions divided by the width of said parallelogram, determining each stroke candidate pixel within the standard size horizontal row and vertical column pixel dimensions corresponding to any part of a stroke pixel of the character image, dividing each stroke candidate pixel into an array of subpixels, counting the number of subpixels of each candidate stroke pixel which map into any part of a stroke pixel of the character image, and setting a candidate stroke pixel as a stroke pixel if and only if the number of subpixel mapping into a stroke pixel of the character image exceeds a predetermined number.