Character recognition device and character segmentation method

1. A character recognition device for use with a medium on which a character string is printed, the character recognition device comprising:

an image reader structured to capture the character string on the medium as image data;

an image memory structured to store the image which is read by said image reader; and

a data processor structured to segment the character string from said image data stored in said image memory and segmenting characters from said character string for character recognition;

wherein said data processor comprises:

a character segmenting unit structured to detect boundary positions of neighboring characters in said character string and segment each character; and

said character segmenting unit comprises:

a boundary search range setting unit structured to set a range to search boundary position of neighboring characters in said character string and

a boundary position setting unit structured to set a boundary position of characters by using a discriminant analysis method within said search range which has been set; and

said boundary position setting unit is structured to divide a projection to said boundary search range by a predetermined width into two regions, calculate the within-region variance in each region, the between-region variance, and the variance ratio between the within-region variance and the between-region variance, and set a segmentation position of characters based on said calculated variance ratio.

2. The character recognition device as set forth in claim 1 wherein

said character segmenting unit comprises a projection curve creating unit which, based on the smallest luminance value of each pixel line arranged in the direction orthogonal to said character string in said image data, is structured to create a projection curve in said image data, and

said boundary search range setting unit is structured to set said boundary search range by associating it with said projection curve.

3. The character recognition device as set forth in claim 1 , wherein

said boundary search range setting unit sets said boundary search range from one end of said character string, and as a boundary position is set within said set boundary search range by said boundary position setting unit, a next boundary search range is set from said set boundary position; and

said boundary position setting unit sets a boundary position for every boundary search range which is set sequentially.

4. A character segmenting method, in a character recognition device for recognizing characters by segmenting each character from a character string which is segmented from image data acquired by capturing a character string on a medium, for processing said image data to segment each character from said character string; the method comprising:

detecting a boundary position of neighboring characters in said character string and segmenting each character;

setting a range to search a boundary position of neighboring characters in said character string; and

acquiring a boundary position of characters by using a discriminant analysis method within said set search range;

wherein, in said acquiring a boundary position, a projection to said boundary search range is divided by a predetermined width into two regions; the within-region variance in each of two division regions, the between-region variance and the variance ratio between the within-region variance and the between-region variance are calculated; and the position to segment a character is set based on said calculated variance ratio.

5. The character segmenting method as set forth in claim 4 wherein

said segmenting comprises detecting a smallest luminance value in each pixel line arranged in the direction orthogonal to the direction of said character string in said image data and creating a projection curve within said image data based on said smallest luminance value of each pixel line; and

said setting a range sets said boundary search range by associating it with said projection curve.

6. The character segmenting method for a character recognition device, which is set forth in claim 4 , wherein:

in said boundary search range setting step, said boundary search range is set from one end of said character string, and as a boundary position is set within said set boundary search range by said boundary position setting unit, a next boundary search range is set from said set boundary position;

in said boundary position setting step, a boundary position is set for every character according to said boundary search range which is set sequentially.