Method and system for document image feature extraction

• US 6,006,226 A
• Filed: 09/24/1997
• Issued: 12/21/1999
• Est. Priority Date: 09/24/1997
• Status: Expired due to Term

• Current Assignee: Ricoh Company Limited, Ricoh Corporation (Ricoh Company Limited)
• Original Assignee: Ricoh Company Limited, Ricoh Corporation (Ricoh Company Limited)
• Inventors: Keith, Alex, Zandi, Ahmad, Cullen, John F.
• Primary Examiner(s): Amsbury, Wayne
• Assistant Examiner(s): PARDO, THUY N
• Application Number:
  US08/936,339
• Time in Patent Office: 818 Days
• Field of Search: 707/4, 707/104, 707/512, 707/514, 707/529, 707/6, 348/607, 345/348, 382/248, 382/283, 382/313, 382/232, 382/112, 382/176, 382/296, 382/133, 382/237, 382/244
• US Class Current: 382/232
• CPC Class Codes: G06F 16/5838 using colour
• CPC Class Codes: G06F 16/5846 using extracted text
• CPC Class Codes: Y10S 707/915 Image
• CPC Class Codes: Y10S 707/99934 Query formulation, input preparation, or translation
• CPC Class Codes: Y10S 707/99936 Pattern matching access

Abstract

A set of image feature extraction techniques to locate and group documents based upon appearance in a database management system. The system automatically determines visual characteristics of document images and collects documents together according to the relative similarity of their document images. The system is operable on both binary and grayscale images.

Claims

1. A method for searching a database containing a plurality of document images for a particular document image, said method comprising:

compressing each document image in said plurality of document images to obtain a compressed representation having a low-pass component and a high-pass component;

extracting image feature information from said compressed representation, wherein said extracting further comprises extracting statistical moments from said low-pass component and extracting connected component information from said high-pass component; and

matching said image feature information with image feature information from said particular document image.

2. The method of claim 1 wherein said compressing step comprises recursively applying a pyramidal wavelet decomposition transform to said document images.

3. The method of claim 2 wherein said pyramidal wavelet decomposition transform comprises a CREW algorithm.

4. The method of claim 1 wherein said statistical moments comprise:

a mean value of pixel values; and

a variance value of pixel values.

5. The method of claim 1 wherein said connected component information further comprises:

a number of words in said document image;

a number of pictures in said document image; and

a number of columns in said document image.

6. The method of claim 1 further comprising:

extracting from said low-pass component an iconic representation as an index for searching.

7. A method for searching a database containing a plurality of document images in binary representation for a particular document image, comprising the steps of:

converting each said document image in binary representation to produce a grayscale representation of said document image;

compressing each said grayscale representation of each document image in said plurality of document images to obtain a compressed representation having a low-pass component and a high-pass component;

extracting image feature information from said compressed representation, wherein said extracting further comprises extracting statistical moments from said low-pass component and extracting connected components from said high-pass component; and

matching said image feature information with image feature information from said particular document image.

8. The method of claim 7 wherein said compressing comprises:

recursively applying a pyramidal wavelet decomposition transform to said document images.

9. The method of claim 8 wherein said pyramidal wavelet decomposition transform comprises a CREW algorithm.

10. The method of claim 7 wherein said statistical moments comprise:

a mean value of pixel values; and

a variance value of pixel values.

11. The method of claim 7 wherein said connected component information further comprises:

a number of words in said document image;

a number of pictures in said document image; and

a number of columns in said document image.

12. The method of claim 7 wherein said converting comprises filtering said binary representation to produce a grayscale representation of said document image.

13. The method of claim 7 further comprising:

selecting said low-pass component for use as an icon.

14. A document image database organizing system comprising:

an electronic storage unit that stores a document image database;

a display that displays document images; and

a processor unit coupled to said electronic storage device and said display, said processor unit operative to;

compress document images to obtain a compressed representation having a low-pass component and a high-pass component;

extract image feature information about example document images, said image feature information comprising statistical moments extracted from said low-pass component and connected components extracted from said high-pass component; and

compare said image feature information from a particular document image to said image feature information from said plurality of document images in said database.

15. A document image database organizing system comprising:

an electronic storage unit that stores a document image database;

a display that displays document images; and

a processor unit coupled to said electronic storage device and said display, said processor unit operative to;

convert document image formats;

compress document images to obtain a compressed representation having a low-pass component and a high-pass component;

extract image feature information about example document images, said image feature information comprising statistical moments extracted from said low-pass component and connected components extracted from said high-pass component; and

compare said image feature information from a particular document image to said image feature information from said plurality of document images in said database.

16. A computer program product comprising:

code that compresses document images to obtain a compressed representation having a low-pass component and a high-pass component;

code that extracts image feature information from said compressed representation, said code further comprising code that extracts statistical moments from said low-pass component and code that extracts connected components from said high-pass component;

code that compares image feature information from a particular document image with image feature information from other document images; and

a computer readable storage medium for storing the codes.

17. The computer program product of claim 16 wherein said code that compresses document images to obtain a compressed representation having a low-pass component and a high-pass component comprises code that recursively applies a pyramidal wavelet decomposition transform to said document images.

18. The computer program product of claim 17 wherein said pyramidal wavelet decomposition transform comprises a CREW algorithm.

19. The computer program product of claim 16 wherein said statistical moments comprise:

a mean value of pixel values; and

a variance value of pixel values.

20. The computer program product of claim 16 wherein said connected component information further comprises:

a number of words in said document image;

a number of pictures in said document image; and

a number of columns in said document image.

21. The computer program product of claim 16 further comprising code that extracts from said low-pass component an iconic representation as an index for searching.

22. A computer program product comprising:

code that converts binary represented document images to grayscale document images;

code that compresses document images to obtain a compressed representation having a low-pass component and a high-pass component;

code that extracts image feature information from said compressed representation, wherein said code that extracts further comprises code that extracts image feature information further comprises code that extracts statistical moments from said low-pass component and code that extracts connected components from said high-pass component;

code that compares image feature information from a particular document image with image feature information from other document images; and

a computer readable storage medium for storing the codes.

23. The computer program product of claim 22 wherein said code that compresses document images to obtain a compressed representation having a low-pass component and a high-pass component comprises code that recursively applies a pyramidal wavelet decomposition transform to said document images.

24. The computer program product of claim 23 wherein said pyramidal wavelet decomposition transform comprises a CREW algorithm.

25. The computer program product of claim 22 wherein said statistical moments comprise:

a mean value of pixel values; and

a variance value of pixel values.

26. The computer program product of claim 22 wherein said connected component information further comprises:

a number of words in said document image;

a number of pictures in said document image; and

a number of columns in said document image.

27. The computer program product of claim 22 further comprising code that extracts from said low-pass component an iconic representation as an index for searching.