System and method for efficient determination of recognition initial conditions

• US 20020085756A1
• Filed: 12/28/2000
• Published: 07/04/2002
• Est. Priority Date: 12/28/2000
• Status: Active Grant

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Shah, Kishan B.
• Granted Patent: US 6,961,465 B2
• Time in Patent Office: Days
• US Class Current: 382/181
• CPC Class Codes: G06F 18/00 Pattern recognition

Abstract

A system (200) and method creates (312) a reduced-resolution version of an image and uses the reduced resolution version to identify (324) recognition initial conditions. The recognition initial conditions are then used (326) on the original version of the image to recognize objects such as characters in the image.

Claims

1. A method of recognizing at least one object in a digitized representation of an image, comprising:

receiving the digitized representation of the image, the representation having a first resolution;

creating a reduced-resolution version of the image responsive to the digitized representation of the image, the reduced-resolution version of the image having a second resolution lower than the first resolution; and

identifying a value of each of at least one recognition initial condition responsive to at least a portion of the reduced resolution version of the image; and

recognizing the at least one object represented in the digitized representation of the image responsive to the value of each of the at least one recognition initial condition identified.

2. The method of claim 1 wherein the identifying step comprises:

providing a plurality of sets of values of at least one initial condition;

for each of the sets of at least one initial condition, identifying a confidence level of recognition by attempting to recognize from the reduced-resolution version of the image the at least one object responsive to the at least one initial condition in the set; and

selecting at least one of the values of the at least one initial conditions in the set responsive to the confidence levels identified.

3. The method of claim 2 wherein the selecting step comprises selecting a value of each of at least one initial condition corresponding to a highest confidence level from a plurality of the confidence levels identified.

4. The method of claim 2 wherein the selecting step comprises selecting a value of each of at least one initial condition corresponding to a confidence level exceeding a threshold.

5. The method of claim 1 wherein the creating step comprises calculating an average of at least one value of a plurality of pixels of the digitized representation of the image.

6. The method of claim 1 additionally comprising recognizing at least one additional object represented in the digitized representation of the image responsive to the value of at least one recognition initial condition identified responsive to a confidence level exceeding a threshold.

7. The method of claim 1 additionally comprising:

attempting to recognize at least one additional object represented in the digitized representation of the image responsive to the value of at least one recognition initial condition identified, the attempting step comprising producing a confidence level of the attempt; and

responsive to the confidence level of the attempt below a threshold;

repeating the identifying step; and

recognizing the at least one object represented in the digitized representation of the image responsive to the value of each of the at least one recognition initial condition identified during the repeating step.

8. A computer program product comprising a computer useable medium having computer readable program code embodied therein for recognizing at least one object in a digitized representation of an image, the computer program product comprising:

computer readable program code devices configured to cause a computer to receive the digitized representation of the image, the representation having a first resolution;

computer readable program code devices configured to cause a computer to create a reduced-resolution version of the image responsive to the digitized representation of the image, the reduced-resolution version of the image having a second resolution lower than the first resolution; and

computer readable program code devices configured to cause a computer to identify a value of each of at least one recognition initial condition responsive to at least a portion of the reduced resolution version of the image; and

computer readable program code devices configured to cause a computer to recognize the at least one object represented in the digitized representation of the image responsive to the value of each of the at least one recognition initial condition identified.

9. The computer program product of claim 8 wherein the computer readable program code devices configured to cause a computer to identify comprises:

computer readable program code devices configured to cause a computer to provide a plurality of sets of values of at least one initial condition;

computer readable program code devices configured to cause a computer to, for each of the sets of at least one initial condition, identify a confidence level of recognition by attempting to recognize from the reduced-resolution version of the image the at least one object responsive to the at least one initial condition in the set; and

select at least one of the values of the at least one initial conditions in the set responsive to the confidence levels identified.

10. The computer program product of claim 9 wherein the computer readable program code devices configured to cause a computer to select comprise computer readable program code devices configured to cause a computer to select a value of each of at least one initial condition corresponding to a highest confidence level from a plurality of the confidence levels identified.

11. The computer program product of claim 9 wherein the computer readable program code devices configured to cause a computer to select comprise computer readable program code devices configured to cause a computer to select a value of each of at least one initial condition corresponding to a confidence level exceeding a threshold.

12. The computer program product of claim 8 wherein the computer readable program code devices configured to cause a computer to creating comprise computer readable program code devices configured to cause a computer to calculate an average of at least one value of a plurality of pixels of the digitized representation of the image.

13. The computer program product of claim 8 additionally comprising computer readable program code devices configured to cause a computer to recognize at least one additional object represented in the digitized representation of the image responsive to the value of at least one recognition initial condition identified responsive to a confidence level exceeding a threshold.

14. The computer program product of claim 8 additionally comprising:

computer readable program code devices configured to cause a computer to attempt to recognize at least one additional object represented in the digitized representation of the image responsive to the value of at least one recognition initial condition identified, the computer readable program code devices configured to cause a computer to attempt comprising computer readable program code devices configured to cause a computer to produce a confidence level of the attempt; and

computer readable program code devices configured to cause a computer to, responsive to the confidence level of the attempt below a threshold;

repeat the identifying step; and

recognize the at least one object represented in the digitized representation of the image responsive to the value of each of the at least one recognition initial condition identified during operation of the computer readable program code devices configured to cause a computer to repeat.

16. The system of claim 15 wherein:

the system additionally comprises a initial condition selector for selecting and providing at an output a plurality of sets of initial conditions, each set different from at last one of the other sets; and

the recognition engine additionally has a third input coupled to the initial condition selector output for receiving the plurality of sets of initial conditions and the recognition engine performs the attempt on the at least one object at least one time for each of the sets of initial conditions received at the third recognition input.

17. The system of claim 16:

wherein the recognition engine;

additionally has a fourth input for receiving an additional set of initial conditions;

performs the recognizing responsive to the additional set of initial conditions;

is additionally for providing a recognition confidence level at a second output responsive to said attempt, for each of the at least one times; and

the system additionally comprising a initial condition identifier having a first input coupled to the recognition second output for receiving the recognition confidence level for each of the at least one times, and a second input coupled to the initial condition selector output for receiving each of the sets of initial conditions, the initial condition identifier for selecting and providing at an output coupled to the fourth recognition engine input the additional set of initial conditions responsive to the sets of initial conditions received at the initial condition identifier second input and the recognition confidence level for each of the at least one times received at the initial condition identifier first input.

18. The system of claim 17 wherein the initial condition identifier selects the additional set of initial conditions additionally responsive to a threshold confidence level.

19. The system of claim 17 wherein:

the at least one time comprises a plurality of times; and

the initial condition identifier selects the additional set of initial conditions responsive to a confidence level for at least one of the at least one times relative to at least one other confidence level for at least a different of the at least one times.

15. A system for recognizing objects, the system comprising:

a downsampler having an input coupled to a system input operatively coupled for receiving a representation of an image having a first resolution, the downsampler for producing and providing at an output a reduced-resolution version of the image responsive to the representation of the image received at the downsampler input, the reduced resolution version of the image having a second resolution lower than the first resolution; and

a recognition engine having a first input coupled to the downsampler output for receiving the reduced-resolution version of the image and a second input coupled to the system input for receiving the representation of the image, the recognition engine for;

at least attempting to recognize at least one first object in the reduced resolution version of the image received at the first input, at least one time;

recognizing at least one second object in the representation of the image received at the second input; and

providing a representation of the at least one object recognized at a first output coupled to a system output.

20. The system of claim 20 wherein the recognition engine additionally recognizes at least one third object in the representation of the image received at the second input responsive to the additional set of initial conditions and a confidence level corresponding to the at least one third object.