OCR Matrix extractor

• US 4,403,340 A
• Filed: 01/06/1981
• Issued: 09/06/1983
• Est. Priority Date: 01/06/1981
• Status: Expired due to Term

• Current Assignee: Caere Corp. (Mexico) (Microsoft Corporation)
• Original Assignee: Caere Corp. (Mexico) (Microsoft Corporation)
• Inventors: Smith, William R., Kumpf, Richard E.
• Primary Examiner(s): Boudreau, Leo H.
• Application Number:
  US06/223,002
• Time in Patent Office: 973 Days
• Field of Search: 340/146.3 H, 340/146.3 AC, 340/146.3 MA, 382/45, 382/48, 382/62, 382/63, 358/260, 358/261, 235/480
• US Class Current: 382/173
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/146 Aligning or centring of the image pick-up or image-field

Abstract

A matrix extractor is described which is particularly useful for extracting data from a memory which contains representations of optically scanned characters. The method and apparatus of the present invention permits the isolation of packets of data in the memory and the determination of the location of these packets. Data representing the packets is then extracted from the memory for character recognition. The isolation of packets in memory greatly reduces the amount of data which must be processed by the recognition processor and provides much more reliable recognition.

Claims

1. In the processing of scanned images for character recognition, a method for isolating areas of the scanned medium for recognition processing comprising the steps of:

storing representations of said scanned images in a memory;

scanning said memory in a first direction so as to detect the end of a packet of data in said memory;

sensing the location of said end of said packet in said memory;

back-scanning said memory through said packet in a second direction, opposite to said first direction so as to detect the end of said packet in said second direction;

recording the extent of said packet in memory in the direction perpendicular to said first and second directions during said back-scanning;

sensing the location of the end of said packet in said second direction;

whereby data in memory is isolated which can be removed to recognition processing.

2. In an optical character recognition system an improvement comprising:

a memory for storing scanned data;

a first, second and third register;

first control means for scanning said memory in a first direction and for writing and removing data from said first register, so as to enable the detection of the beginning and end of a packet of data in said memory in said first direction, said first control means coupled to said memory and said first register;

second control means for scanning said memory in a second direction, opposite to said first direction and for writing and removing data from said second and third registers, so as to enable detection of the end of said packet of data in said second direction and the extent of said packet of data in the direction perpendicular to said second direction, said second control means coupled to said memory and said second and third registers;

whereby data in said memory is isolated for recognition processing.

3. The improvement defined by claim 2 wherein said first control means tracks the tops and bottoms of packets of data in said first register and detects vertical growth, join, decay and drift of data packets within said first register.

4. The improvement defined by claim 2 wherein said second control means writes a signal into said second register corresponding to the detection of the end of said packet in said first direction within said first register.

5. The improvement defined by claim 4 wherein said second control means tracks the tops and bottoms of said packet of data in said second register and detects vertical growth, decay and drift for said packet of data contiguous with said signal written into said second register.

6. The improvement defined by claim 2 wherein said second control means tracks the vertical extent of all packets of data within said third register and upon said detection of said end of said packet of data in said second direction removes all data from said third register not corresponding to said packet of data.

7. The improvement defined by claim 2 including data extraction means coupled to said memory and registers for defining a perimeter about said packet of data and for removing data from said memory from within said perimeter.

8. The improvement defined by claim 7 wherein said perimeter is a rectangle.

9. The improvement defined by claim 8 wherein the extent of said rectangle in said first direction is determined from the detection of said end of said packet of data from said first register.

10. The improvement defined by claim 8 wherein the extent of said rectangle in said second direction is determined from said second register.

11. The improvement defined by claim 10 wherein the extent of said packet of data in a direction perpendicular to said first and second direction is determined from said third register.

12. The improvement defined by claim 11 wherein data defining said rectangle is formatted for a recognition processor.

13. The improvement defined by claim 7 including data comprising means for normalizing data in said first and second direction as said data is extracted for each of said packets of data.

14. The improvement defined by claim 13 wherein for said normalization each slice of data for each of said packets of data is compressed into a fixed number of words.

15. The improvement defined by claim 2 wherein said memory is also used to store data representing the direction in which data is read.