Real time binarization of gray images

• US 6,738,496 B1
• Filed: 05/30/2000
• Issued: 05/18/2004
• Est. Priority Date: 11/01/1999
• Status: Expired due to Fees

• Current Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Original Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Inventors: Van Hall, Richard C.
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): Lu, Tom Y.
• Application Number:
  US09/580,493
• Time in Patent Office: 1,449 Days
• Field of Search: 382/101, 382/270, 382/307, 382/266, 382/256, 382/261, 382/168-172, 382/176, 358/173
• US Class Current: 382/101
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/162 Quantising the image signal

Abstract

A binarization method for gray address images which combines high quality and high speed. The method is designed specifically for efficient software implementation. Two binarization approaches, localized background thresholds and Laplacian edge enhancement, are combined into a process to enhance the strengths of the two methods and eliminate their weaknesses. The image is divided into tiles, making binarization decisions for each tile. Tile decisions are modified based on adjacent tile decisions and then the tiles are binarized. Binarization of pixels is performed by performing background thresholding and edge detection thresholding. Only pixels exceeding both thresholds are selected as “on”.

Claims

1. A computer implemented method for real time binarization of gray scale images, said method comprising the steps of:

receiving an initial gray scale image;

dividing the initial gray scale image into a plurality of tiles;

determining for each tile in the plurality of tiles whether it is to be categorized as a background tile or a binarization tile, based on a sub-sampling of pixels in the tile;

saving the categorization information for the each tile in the plurality of tiles;

updating the saved categorization information for each tile by examining binarization flags for neighboring tiles to determine whether the number of neighboring tiles flagged for binarization is above a desired threshold, and if so, then setting the binarization flag “

on”

for the tile, but if not, then setting the binarization flag “

off”

for the tile;

binarizing the plurality of tiles, including steps of background thresholding and edge detection, in accordance with said binarization flags by combining results of background thresholding and results of edge detection, thereby creating a binarized image readable by an optical character recognition means.

2. A computer implemented method as recited in claim 1, wherein the step of updating the saved categorization information for each tile counts binarization flags for a 3×

3 neighborhood of tiles.

3. A computer implemented method as recited in claim 1, wherein the initial gray scale image is an 8-bit gray scale and each tile comprises 16×

16 pixels of gray scale information.

4. A computer implemented method as recited in claim 3, wherein the step of determining for each tile in the plurality of tiles whether it is to be categorized as a background tile or a binarization tile, further comprises the steps of:

selecting a subset of pixels in the each tile in the plurality of tiles;

examining the selected subset of pixels to generate statistics on variance and to generate a 32-level histogram;

selecting initial tiles for binarization based on information generated in the examining step;

determining a background threshold for the selected tiles; and

determining whether an initial tile is identified for pixel binarization using a count of neighboring tiles selected for binarization compared to a selected threshold.

5. A method as recited in claim 4, wherein each tile comprises 16×

16 pixels, and the selected subset of pixels comprises every fifth pixel in selected rows of pixels 0 to 15, (i.e., pixels 0, 5, 10,

     15).

6. A computer implemented method as recited in claim 4, wherein the step of binarizing the plurality of tiles, further comprises the steps of:

performing background thresholding using a center pixel and pixels one(1) unit distant east, west, north, and south from the center pixel, wherein the center pixel gray-level has a higher weight than the pixels one (1) unit distant;

performing edge detection using pixels one and two (2) units distant east, west, north, south to the center pixel, the edge detection being immune to common line scan camera even/odd channel irregularities; and

determining whether the center pixel is on or off based on combining results of the step of performing background thresholding and results of the step of performing edge detection.

11. A method as recited in claim 1, wherein said computing device provides only a single said categorization information update for respective tiles of said gray scale image.

7. A system for sorting mail comprising:

means for imaging a piece of mail, the means generating a machine readable gray image;

a computing device for generating a binarized image of the machine readable gray image, wherein the machine readable gray image is divided into a plurality of tiles, each tile in the plurality of tiles is categorized as a background tile or a binarization tile based on a sub-sampling of pixels within each tile to provide binarization information, the resulting categorization information for each tile in the plurality of tiles being saved, each tile determined as being selected for a categorization information update, or not, based on neighboring tile categorization, the categorization information for each tile in the plurality of tiles being updated as necessary in accordance with the number of neighboring tiles having different categorization information;

selectively performing binarization of respective tiles in accordance with said categorization information, as updated; and

sorting the piece of mail based on selected criteria relating to the binarized image generated by the computing device.

8. A system for sorting mail as recited in claim 7, wherein each tile of the plurality of tiles comprises 16×

16 pixels of gray scale information.

9. A system for sorting mail, as recited in claim 8, wherein the categorization means selects a subset of pixels in the each tile in the plurality of tiles and for each tile:

averages the selected subset of pixels to reduce noise, utilizing pixels one unit distant east, west, north, and south (right, left, above, below), performs edge detection using pixels two (2) units distant east, west, north, and south, thereby using nine (9) pixels at a time out of the 25 pixels in a 5×

5 array of pixels determined by the selected subset of pixels in the 16×

16 array of pixels in a tile.

10. A system for sorting mail, as recited in claim 7, wherein the step of sorting the piece of mail further comprises the steps of:

identifying alpha-numeric characters in the binarized image; and

comparing the selected criteria relating to the binarized image with the identified alpha-numeric characters, thereby determining a category or bin destination for the piece of mail.

12. A system as recited in claim 7, wherein said computing device provides only a single said categorization information update for respective tiles into which a given gray scale image is divided.