Character string extraction method and character string extraction device

A character string extraction method in which a character frame is set on a paper sheet image obtained by capturing a paper sheet on which a character string is printed or typed, and an image of the character string is extracted, the character string extraction method comprising:

• setting a character portion that is a region including each character in the character string, a rim portion that is a surrounding region adjacent to and around the character portion, a character frame that is a combination of the character portion and the rim portion, and a character string frame that includes all the character frames;
• calculating feature values of each image in the character portion and the rim portion for each character frame;
• calculating a character string frame evaluation value based on the feature values for the character string frame;
• moving a position of the character string frame on the paper sheet image;
• and
• extracting an image in the character portion by using the character string frame at a position at which the character string frame evaluation value reaches a maximum.

A character string extraction device that receives a paper sheet image obtained by capturing a paper sheet on which a character string is printed or typed, sets a character string frame thereon, and thereby extracts an image of the character string, the character string extraction device comprising:

• a skew correction unit that corrects a skew of the paper sheet image when the paper sheet image is skewed;
• a partial region extraction unit that extracts a region that includes the character string from the paper sheet image or the paper sheet image corrected by the skew correction unit;
• an evaluation value calculation unit that sets a region that includes each character in the character string as a character portion, a surrounding region adjacent to and around the character portion as a rim portion, the character portion combined with the rim portion as a character frame, and a region including all the character frames as a character string frame, calculates feature values of each image in the character portion and the rim portion while moving a position of the character string frame, and calculates a character string frame evaluation value based on the feature values of the rim portion and the character portion calculated at different positions;
• and
• a character position determining unit that determines, as a region for extracting an image, the character portion of the character string frame for which the character string frame evaluation value reaches a maximum.