Browse Prior Art Database

FAX Compression using Clustering

IP.com Disclosure Number: IPCOM000052045D
Original Publication Date: 1981-Apr-01
Included in the Prior Art Database: 2005-Feb-11
Document File: 2 page(s) / 49K

Publishing Venue

IBM

Related People

Casey, RG: AUTHOR [+2]

Abstract

This invention relates to a facsimile compression method for minimizing the number of mismatches between scanned characters and a reduced number of templates. The method steps comprise: (1) buffering a page of characters; (2) scanning the page and assigning scan characters into similarity classes; (3) constructing the best matched prototype for members of each class; and (4) transmitting a number indicative of a match between each character occurrence and one of the set of prototypes.

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 73% of the total text.

Page 1 of 2

FAX Compression using Clustering

This invention relates to a facsimile compression method for minimizing the number of mismatches between scanned characters and a reduced number of templates. The method steps comprise: (1) buffering a page of characters; (2) scanning the page and assigning scan characters into similarity classes; (3) constructing the best matched prototype for members of each class; and (4) transmitting a number indicative of a match between each character occurrence and one of the set of prototypes.

Advantageously, the method reduces both the number of prototypes and the amount of mismatch data. This is accomplished by transmitting not a selected scan pattern as is found in the prior art, but an "average" pattern representing all occurrences of a given symbol. The average prototype is constructed so as to best match all the occurrences of the symbol. This has the effect of minimizing the mismatched data. At the same time, only one such prototype is required for each symbol class. This minimizes the number of prototypes.

Prototypes can be obtained by means of a clustering procedure such as is set out in the figure. The procedure assigns the scan patterns into similarity classes and constructs a best match prototype for the members of each class. A sample of clustering algorithm is given, for example, in [*]. Significantly, the clustering algorithm need not operate perfectly. Compression advantage is obtainable over, for example, the prior art sam...