Browse Prior Art Database

Character Recognition by Adjacent Scan Outline and Centroid Codes

IP.com Disclosure Number: IPCOM000075394D
Original Publication Date: 1971-Sep-01
Included in the Prior Art Database: 2005-Feb-24
Document File: 3 page(s) / 63K

Publishing Venue

IBM

Related People

Cutaia, A: AUTHOR

Abstract

Many character-recognition systems contain "preprocessors", which attempt to simplify and improve the recognition-decision circuits by minimizing the effect of variations which are irrelevant to the correct identification of an input character. Such variations include size, position and rotational skew of the character, line-width variations, overall contrast, noise, and boundary conditions.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 53% of the total text.

Page 1 of 3

Character Recognition by Adjacent Scan Outline and Centroid Codes

Many character-recognition systems contain "preprocessors", which attempt to simplify and improve the recognition-decision circuits by minimizing the effect of variations which are irrelevant to the correct identification of an input character. Such variations include size, position and rotational skew of the character, line- width variations, overall contrast, noise, and boundary conditions.

Fig. 1 shows a recognition system 10, which receives a serial electronic image of character at input 11 and stores it in shift register 12. Element-code generator 13 compares corresponding video from two adjacent raster scans over the input character to produce four element codes, as follows:

(Image Omitted)

Examples of video which may produce these codes is shown in the left column at 41, Fig. 4. The scan direction is shown by arrows 42. More than two adjacent scans, or more than two bits per scan, may be used in deriving the element codes, in order to minimize the effect of noise in the pattern.

Outline-code generator 14 combines the element codes to determine the direction in which a line of the pattern is progressing. The code sequences and their associated direction codes (illustrated at 43, Fig. 4) and sign codes are:

(Image Omitted)

In order to reference these directions and signs spatially within the character, the upper and lower character extremities are determined by profile register 15, and the character height is broken into five regions by generator 16. The regions in which the outline sign is initiated, propagated and terminated are transmitted in real-time to decision logics 17. The character is then classified by predetermined combinations of the outline and region codes, and its identity appears at output
18.

System 20, Fig. 2, codes the centroid motion of pattern elements, rather than their outline motion. Video input 21 is stored in shift register 22, as before. Element-code generator 23 may be essentially the same as generator 13, although it is advantageous here to drop the "C" code, and to introduce an "01" code, which represents either of the sequences "CO" or "OC". These codes are shown in the right-hand column at 41, Fig. 4. Centroid-code generator 24 measures the directional shift of the centroid of each line segment in each scan relative to that of its adjacent line segment in the preceding scan. The following table shows representative element-code sequences and the resulting direction of centroid motion. The sequence "WW" may be assigned the centroid codes "up", "down" or "zero" when the length of the first "W" code is respectively larger than, smaller than or approximately equal to that of the second "W" code. The sequence "BB" is assigned to the "down", "u...