Multilevel bit-mapped image analysis method

• US 20110091111A1
• Filed: 03/13/2003
• Published: 04/21/2011
• Est. Priority Date: 12/17/2002
• Status: Active Grant

• Current Assignee: ABBYY Development LLC
• Original Assignee: ABBYY Software
• Inventors: Vnuchkov, Dmitry, Tereshchenko, Vadim, Anisimovich, Konstantin, Rybkin, Vladimir
• Granted Patent: US 8,103,103 B2
• Time in Patent Office: Days
• US Class Current: 382/195
• CPC Class Codes: G06V 30/40 Document-oriented image-based pattern recognition

Abstract

The present invention discloses a multilevel method of bit-mapped image analysis that comprises a whole image data representation via its components—objects of different levels of complexity—hierarchically connected therebetween by spatially-parametrical links.

The said method comprises preliminarily generating a classifier of the objects that possibly may be present in the image consisting of one or more levels differing in complexity; parsing the image into objects; attaching each object to one of predetermined levels; establishing hierarchical links between objects of different levels; establishing links between objects within the same level; performing an objects features analysis.

The objects feature analysis comprises at least generating and examining a hypothesis about object features and correcting the concerned objects'"'"' features of the same and other levels in response to the hypothesis examination results. The step of objects'"'"' features analysis may also comprise execution of a recursive RX-cut within the same level.

Claims

1. A method of bit-mapped image analysis comprising

classifying the types of objects, that possibly may be present within the said image, into one or more levels differing in complexity;

parsing the said image into objects;

attaching every said object to one of said levels of complexity;

establishing links between said objects of said different levels of complexity;

establishing links between said objects within each level of complexity;

performing an object'"'"'s features analysis, said analysis comprising at least the following steps

forming a hypothesis about object features,

examining the hypothesis about object features,

correcting the concerned objects'"'"' features of the same and other levels.

2. The method as recited in claim 1, wherein said step of objects'"'"' features analysis further comprises execution of at least one recursive RX-cut on objects within the same level of complexity.

3. The method as recited in claims 1 and 2, further outputting the results to a subsequent process.

4. The method as recited in claims 1 through 3, wherein the subsequent process is a character recognition process.

5. The method as recited in claims 1 through 4, wherein said hypothesis about objects features comprises at least the following data

a list of objects of lower complexity incorporated into the object;

a relationship factors estimation of objects within the same level of complexity, said estimation comprising at least

spatial estimations;

parametrical estimations.

6. The method as recited in claims 1 through 4, wherein said analysis is processed by moving from a higher complexity level to a lower complexity level.

7. The method as recited in claims 1 through 4, wherein said analysis is performed by moving from a lower complexity level to a higher complexity level.

8. The method as recited in claims 1 through 4, wherein said analysis is processed by applying to any level and further moving to any level in the very order and the very number of times as determined by the selected analyzing strategy.

9. The method as recited in claims 1 and 2, wherein the method of bit-mapped image analysis comprises at least the following steps

classifying the bit-mapped image into a set of levels, comprising

characters,

word tokens,

words,

lines,

paragraphs,

regions,

tables,

non-text regions;

generating a hypothesis about characters'"'"' features,

examining the hypothesis about characters'"'"' features,

correcting a concerned object'"'"'s features within the same and other levels,

generating a hypothesis about word token features,

examining the hypothesis about word token features,

correcting a concerned object'"'"'s features within the same and other levels,

generating a hypothesis about word features,

examining the hypothesis about word features,

correcting a concerned object'"'"'s features within the same and other levels,

generating a hypothesis about the features of set of characters, word tokens and words comprising a line,

examining the hypothesis about the features of characters, word tokens and words comprising a line,

correcting a concerned object'"'"'s features within the same and other levels,

generating a hypothesis about the features of the set of lines comprising a paragraph,

examining the hypothesis about the features of the set of lines comprising a paragraph,

correcting a concerned object'"'"'s features within the same and other levels,

generating a hypothesis about the features of the set of lines and paragraphs comprising a region,

examining the hypothesis about the features of the set of lines and paragraphs comprising a region,

correcting a concerned object'"'"'s features within the same and other levels,

generating a hypothesis about the features of the set of lines comprising a table,

examining the hypothesis about the features of the set of lines comprising a table,

correcting a concerned object'"'"'s features within the same and other levels,

generating a hypothesis about object features comprising non-text object,

examining the hypothesis about object features comprising non-text object,

correcting a concerned object'"'"'s features within the same and other levels.

10. The method as recited in claim 9, wherein after examining a hypothesis at least at one level, a recursive RX-cut is executed to obtain subsidiary data about objects features of the same and other levels.

11. The method as recited in claim 9, 10, further outputting the results to a character recognition procedure.