Method for detecting character strings

Claim 1

A method for detecting character string regions in an image obtained by scanning in a direction a document including a plurality of character strings extending substantially in one direction comprising the steps of: (a) storing image data of said document into an image memory with a direction in which character strings extending in the image are aligned with the scanning direction; (b) generating position data of a black run in each line by accessing said image memory to scan said image; (c) comparing data of black runs obtained in the step (b) for each pair of two adjoining lines, and when a part of a black run which does not adjoin a black run in a preceding line exists in a succeeding line, judging the part to be a boundary candidate of the first kind, generating position data thereof, and storing the data in a table of a storage means, and when a part of a black run which does not adjoin a black run in a succeeding line exists in a preceding line, judging the part to be a boundary candidate of the second kind, generating position data thereof, and storing the data in said table, and (d) looking up said table to determine a pair consisting of a boundary candidate of the first kind and that of the second kind, which have a part whose position for said main scanning direction is common, judging a rectangular region whose boundaries include at least a part of each boundary candidate of the determined pair to be a character string region, and generating position data thereof.

Claim 2

A method for detecting character string regions in an image obtained by scanning in a direction a document including a plurality of character strings extending substantially in one direction comprising the steps of: (a) storing image data of said document into an image memory with a direction in which character strings extending in the image are aligned with the scanning direction; (b) generating position data of a black run in each line by accessing said image memory to scan said image; (c) comparing data of black runs obtained in the step (b) for each pair of two adjoining lines, and when a part of a black run which does not adjoin a black run in a preceding line exists in a succeeding line, judging the part to be a boundary candidate of the first kind, generating position data thereof, and storing the data in a table of a storage means, and when a part of a black run which does not adjoin a black run in a succeeding line exists in a preceding line, judging the part to be a boundary candidate of the second kind, generating position data thereof, and storing the data in said table, and (d) looking up said table to determine a pair consisting of a boundary candidate of the first kind and that of the second kind, which have a part whose position for said main scanning direction is common, judging a rectangular region whose boundaries include at least a part of each boundary candidate of the determined pair to be a character string region, and generating position data thereof.

Claim 9

A method as claimed in any one of claims 1 wherein the position data of each boundary generated in said step (c) includes data of the starting position and length of said boundary.