Browse Prior Art Database

# Pattern Filter Algorithm

IP.com Disclosure Number: IPCOM000085200D
Original Publication Date: 1976-Mar-01
Included in the Prior Art Database: 2005-Mar-02
Document File: 2 page(s) / 55K

IBM

## Related People

Reynolds, SW: AUTHOR

## Abstract

The expansion-contraction technique described in an article by Uno, T., Mese, M. and Ejiri, M., "Defect Detection Device for Printed Circuit Boards", Japan Electronic Engineering, January 1974, pp. 52-57 is effective in finding all defects which include "rough" edges on conducting material. It is not directly applicable to the problem of finding only isolated defects. Binary representations of "defects touching smooth pattern" are indistinguishable from binary representations of "rough" edged patterns which have no defects.

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 74% of the total text.

Page 1 of 2

Pattern Filter Algorithm

The expansion-contraction technique described in an article by Uno, T., Mese, M. and Ejiri, M., "Defect Detection Device for Printed Circuit Boards", Japan Electronic Engineering, January 1974, pp. 52-57 is effective in finding all defects which include "rough" edges on conducting material. It is not directly applicable to the problem of finding only isolated defects. Binary representations of "defects touching smooth pattern" are indistinguishable from binary representations of "rough" edged patterns which have no defects.

The algorithm described here requires the following definition with reference to the logical matrix:
A. A pattern is any configuration of contiguous 1 bits,

in the sense of a neighborhood, which contains at

least one three-by-three square of 1 bits.

The algorithm shown in block form accepts as an argument an arbitrary logical matrix LM. The logical negation of the given matrix is bordered and the 1 bits are uniformly "grown", thereby simulating expansion of material originally represented by the 0's in the given matrix. Another negation and growth is required to simulate contraction of the material. This process generated a mask, or initial pattern matrix F, that is then applied to the given matrix to obtain the defects matrix G.

The pattern "growth" is unique in that all patterns grow in the direction of contiguous 1 bits. The method permits isolation of all defects which do not satisfy the definition of pattern.

A lo...