Recognizing text in a multicolor image

• US 6,148,102 A
• Filed: 05/29/1997
• Issued: 11/14/2000
• Est. Priority Date: 05/29/1997
• Status: Expired due to Term

• Current Assignee: Adobe Systems Incorporated (Adobe Inc.)
• Original Assignee: Adobe Systems Incorporated (Adobe Inc.)
• Inventors: Stolin, Jacob
• Primary Examiner(s): CHANG, JON CARLTON
• Application Number:
  US08/865,021
• Time in Patent Office: 1,265 Days
• Field of Search: 382/164, 382/165, 382/162, 382/173, 382/176, 358/462, 358/464
• US Class Current: 382/164
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/1429 Identifying or ignoring parts by sensing at different wavelengths
• CPC Class Codes: G06V 30/162 Quantising the image signal

Abstract

A method and apparatus for recognizing text in a multicolor image stored in a computer. The image is separated into multiple blocks, and the color distributions of each of the blocks are analyzed. The blocks having two main colors are identified, and two-color blocks have similar colors are grouped into two-color zones. The two colors in each zone are converted to black and white to produce a black and white image. Text is identified in the two-color zones by performing optical character recognition of the black and white image.

Claims

1. A computer-implemented method recognizing text in a multicolor image stored in a computer, the multicolor image comprising a plurality of pixels, each pixel having an associated color, the method comprising:

separating the image into multiple blocks;

analyzing color distributions of the pixels in each of the blocks;

identifying blocks having two main colors;

grouping two-color blocks having similar colors into two-color zones;

identifying text in the two-color zones;

mapping the pixels of each block to a three-dimensional color space;

defining, for each two-color block, a cylinder that encloses the pixels, the cylinder having a height and a radius;

classifying a block as a text block if the ratio of the radius to the height is less than a predefined value,

wherein the text identifying step is performed in the text blocks.

2. The method of claim 1, further comprising:

converting the two colors in each zone to black and white to produce a black and white image.

3. The method of claim 2, further comprising:

performing optical character recognition of the black and white image.

4. The method of claim 1, wherein the predefined value is approximately 0.35.

5. The method of claim 1, wherein the image is stored in the computer as a raster of pixels, the method further comprising:

representing each block as a vector in a three-dimensional color space, the vector originating at a point in a first group of pixels corresponding to a first color and terminating at a point in a second group of pixels corresponding to a second color;

identifying clusters of vectors that point generally in the same directions; and

marking blocks corresponding to clusters that contain more than a predefined number of vectors as text blocks.

6. The method of claim 5, wherein the cluster-identifying step includes:

defining sample points on a sphere in the three-dimensional color space;

identifying a local maximum of a predefined function at a sample point; and

grouping the vectors within a predetermined angle of the sample point into a cluster.

7. The method of claim 6, wherein the sample points are uniformly distributed on the sphere.

8. A computer program residing on a computer-readable medium for recognizing text in a multicolor image, the multicolor image comprising a plurality of pixels, each pixel having an associated color, the computer program comprising instructions for causing the computer to:

separate the image into multiple blocks;

analyze color distributions of each of the blocks;

identify blocks having two main colors;

group two-color blocks having similar colors into two-color zones; and

identify text in the two-color zones;

map the pixels of each block to a three-dimensional color space;

define, for each two-color block, a cylinder that encloses the pixels, the cylinder having a height and a radius; and

classify a block as a text block if the ratio of the radius to the height is less than a predefined value,

wherein the text is identified in the text blocks.

9. The computer program of claim 8, further comprising instructions for causing the computer to:

convert the two colors in each zone to black and white to produce a black and white image.

10. The computer program of claim 9, further comprising instructions for causing the computer to:

perform optical character recognition of the black and white image.

11. The computer program of claim 8, wherein the predefined value is approximately 0.35.

12. The computer program of claim 8, wherein the image is stored in the computer as a raster of pixels, the computer program further comprising instructions for causing the computer to:

represent each block as a vector in a three-dimensional color space, the vector originating at a point in a first group of pixels corresponding to a first color and terminating at a point in a second group of pixels corresponding to a second color;

identify clusters of vectors that point generally in the same directions; and

mark blocks corresponding to clusters that contain more than a predefined number of vectors as text blocks.

13. The computer program of claim 12, wherein the cluster-identifying step includes:

defining sample points on a sphere in the three-dimensional color space;

identifying a local maximum of a predefined function at a sample point; and

grouping the vectors within a predetermined angle of the sample point into a cluster.

14. The computer program of claim 13, wherein the sample points are uniformly distributed on the sphere.

15. Apparatus to recognize text in a multicolor image, the multicolor image comprising a plurality of pixels, each pixel having an associated color, the apparatus comprising:

storage medium to store the image; and

a processor operatively coupled to the storage medium and configured to;

separate the image into multiple blocks;

analyze color distributions of each of the blocks;

identify blocks having two main colors;

group two-color blocks having similar colors into two-color zones; and

identify text in the two-color zones;

map the pixels of each block to a three-dimensional color space;

define, for each two-color block, a cylinder that encloses the pixels, the cylinder having a height and a radius; and

classify a block as a text block if the ratio of the radius to the height is less than a predefined value,

wherein the text identifying step is performed in the text blocks.

16. The apparatus of claim 15, wherein the processor is further configured to:

convert the two colors in each zone to black and white to produce a black and white image.

17. The apparatus of claim 15, wherein the image is stored in the computer as a raster of pixels, and wherein the processor is further configured to:

represent each block as a vector in a three-dimensional color space, the vector originating at a point in a first group of pixels corresponding to a first color and terminating at a point in a second group of pixels corresponding to a second color;

identify clusters of vectors that point generally in the same directions; and

mark blocks corresponding to clusters that contain more than a predefined number of vectors as text blocks.

18. A computer-implemented method recognizing text in a multicolor image stored in a computer, the multicolor image comprising a plurality of pixels, each pixel having an associated color, the method comprising:

separating the image into multiple blocks;

analyzing color distributions of each of the blocks;

identifying blocks having two main colors;

grouping two-color blocks having similar colors into two-color zones;

identifying text in the two-color zones;

representing each block as a vector in a three-dimensional color space, the vector originating at a point in a first group of pixels corresponding to a first color and terminating at a point in a second group of pixels corresponding to a second color;

identifying clusters of vectors that point generally in the same directions; and

marking blocks corresponding to clusters that contain more than a predefined number of vectors as text blocks.

19. The method of claim 18, wherein the cluster-identifying step includes:

defining sample points on a sphere in the three-dimensional color space;

identifying a local maximum of a predefined function at a sample point; and

grouping the vectors within a predetermined angle of the sample point into a cluster.

20. The method of claim 19, wherein the sample points are uniformly distributed on the sphere.

21. A computer program residing on a computer-readable medium for recognizing text in a multicolor image, the multicolor image comprising a plurality of pixels, each pixel having an associated color, the computer program comprising instructions for causing the computer to:

separate the image into multiple blocks;

analyze color distributions of each of the blocks;

identify blocks having two main colors;

group two-color blocks having similar colors into two-color zones;

identify text in the two-color zones;

represent each block as a vector in a three-dimensional color space, the vector originating at a point in a first group of pixels corresponding to a first color and terminating at a point in a second group of pixels corresponding to a second color;

identify clusters of vectors that point generally in the same directions; and

mark blocks corresponding to clusters that contain more than a predefined number of vectors as text blocks.

22. The computer program of claim 21, wherein the cluster-identifying step includes:

defining sample points on a sphere in the three-dimensional color space;

identifying a local maximum of a predefined function at a sample point; and

grouping the vectors within a predetermined angle of the sample point into a cluster.

23. The computer program of claim 22, wherein the sample points are uniformly distributed on the sphere.

24. Apparatus to recognize text in a multicolor image, the multicolor image comprising a plurality of pixels, each pixel having an associated color, the apparatus comprising:

storage medium to store the image; and

a processor operatively coupled to the storage medium and configured to;

separate the image into multiple blocks;

analyze color distributions of each of the blocks;

identify blocks having two main colors;

group two-color blocks having similar colors into two-color zones;

identify text in the two-color zones;

represent each block as a vector in a three-dimensional color space, the vector originating at a point in a first group of pixels corresponding to a first color and terminating at a point in a second group of pixels corresponding to a second color;

identify clusters of vectors that point generally in the same directions; and

mark blocks corresponding to clusters that contain more than a predefined number of vectors as text blocks.