Self-centering character stroke thickening for optical character recognition
AssigneeEastman Kodak Company
InventorRoger S. Gaborski
Life of this patent
Estimating the term…
Abstract
In an optical character recognition system, character strokes are automatically enhanced prior to processing by matching the character bit map with a set of templates. Each template has a stripe of thickness W composed of binary ones surrounded by zeroes along a horizontal, vertical, major diagonal or minor diagonal direction. If the template matches a character stroke to a predetermined degree, the stroke is enhanced by increasing its thickness to a predetermined number N of binary ones.
Claims20
A character image pre-processor which locates and widens thin character strokes in a character image comprising an array of columns and rows of pixels, said pixels characterized by individual values, said values lying within a range including a relatively high ON value and a relatively low OFF value, said pre-processor comprising:
- template memory means for storing a set of n templates comprising an array of M columns and N rows of elements, each template comprising a stripe of width W of high-valued individual template elements in a field of low-valued template elements, the stripe in each of said n templates oriented in a corresponding one of n directions;
- template matching means for summing the products obtained from multiplying pixels of a successive one of a set of M column-by-N row kernels of said image by corresponding elements of a successive one of said templates;
- stroke strengthening means responsive whenever said means for summing obtains a sum greater than a threshold value for raising toward said On value the values of pixels of said character image in said one kernel which are laterally adjacent a line of pixel locations corresponding to the high-valued elements in the stripe of said one template.
A character image pre-processing method for locating and widening thin character strokes in a character image comprising an array of columns and rows of pixels, said pixels characterized by individual values, said values lying within a range including a relatively high ON value and a relatively low OFF value, said method employing a set of n templates comprising an array of M columns and N rows of elements, each template comprising a stripe of width W of high-valued individual template elements in a field of low-valued template elements, the stripe in each of said n templates oriented in a corresponding one of n directions, said method comprising:
- (a) for individual ones of a set of M pixels column-by-N pixel row kernels of said image and for a successive one of said templates, summing the products obtained from multiplying pixels of the one kernel by corresponding elements of the one template;
- and
- (b) whenever the summing step obtains a sum greater than a threshold value, raising toward said ON value the values of pixels of said character image in said one kernel which are laterally adjacent a line of pixel locations corresponding to the high valued elements in the stripe of said one template.
A character image pre-processor which locates and widens thin character strokes in a character image comprising an array of columns and rows of pixels, said pixels characterized by individual values, said values lying within a range including a relative high ON value and a relatively low OFF value, said pre-processor comprising:
- template memory means for storing a set of n templates comprising an array of M columns and N rows of elements T.sub.ij where i and j are integers such that 1.ltoreq.i.ltoreq.N and 1.ltoreq.j.ltoreq.M+1, each template comprising a stripe of width W of high-valued individual template elements in a field of low-valued template elements, the stripe in each of n templates oriented in a corresponding one of n directions;
- template matching means for computing for individual kernels of pixels P.sub.ij in said image and for each one of said templates ##EQU2## stroke strengthening means responsive whenever S is greater than a threshold value for raising toward said ON value the values of pixels of said character image in said one kernel which are laterally adjacent a line of pixel locations corresponding to the high valued elements in the stripe of said one template.
Dependencies are inferred from the claim text.