Document image processing apparatus, document image processing method, and computer-readable recording medium having recorded document image processing program

• US 8,611,666 B2
• Filed: 03/19/2010
• Issued: 12/17/2013
• Est. Priority Date: 03/27/2009
• Status: Expired due to Fees

• Current Assignee: Konica Minolta Business Technologies Incorporated (Konica Minolta Inc.)
• Original Assignee: Konica Minolta Business Technologies Incorporated (Konica Minolta Inc.)
• Inventors: Komaki, Yoshio
• Primary Examiner(s): Koziol, Stephen R
• Assistant Examiner(s): Lee, Jonathan S
• Application Number:
  US12/727,311
• Publication Number:
  US 20100245875A1
• Time in Patent Office: 1,369 Days
• Field of Search: None
• US Class Current: 382/190
• CPC Class Codes: G06V 30/416 Extracting the logical structure, e.g. chapters, sections or page numbers; Identifying elements of the document, e.g. authors

Abstract

Each small region positioned just before a large region according to a reading order is determined as a first candidate, and an evaluating process to evaluate whether each first candidate is an index or not is performed based on a difference in feature from the related large region with respect to each first candidate. Each small region positioned just before a first index according to the reading order is determined as a second candidate, and an evaluating process to evaluate whether each second candidate is the index or not is performed based on a difference in feature from the related first index with respect to each second candidate. Small regions determined as the first index and the second index are extracted as index regions.

Claims

1. A document image processing apparatus comprising:

a memory for storing a document image; and

a controller for controlling extraction of an index region from said document image, wherein said controller is configured to

i) classify a plurality of character string element regions constituting said document image into small regions and large regions,

ii) determine each small region positioned just before said large region according to a reading order as a first candidate, as a first determining process,

iii) determine at least one part of said first candidates as a first index, by performing an evaluating process to evaluate whether or not each said first candidate is an index, based on a difference in feature from the related large region, with respect to each said first candidate, as a first evaluating process,

iv) determine each small region positioned just before said first index according to the reading order as a second candidate, as a second determining process,

v) determine at least one part of said second candidates as a second index, by performing an evaluating process to evaluate whether or not said second candidate is the index, based on a difference in feature from the related first index, with respect to each said second candidate, as a second evaluating process, and

vi) extract the small regions determined as said first index and said second index, as said index region wherein

in said first evaluating process, said controller sets a first feature section for each said first candidate as for a style type different in feature from a corresponding related large region that represents said related large region corresponding to the intended first candidate among a plurality of style types, said first feature section including a feature of said intended first candidate region but not including a feature of said corresponding related large region,

groups into region groups at least one or both of the related large regions and the first candidates having the feature included in said set first feature section,

calculates a first index evaluation degree, based on a number of members of each region group with respect to each said first candidate, and

determines whether or not a logical element of each said first candidate is the index, based on said calculated first index evaluation degree, and

in said second evaluating process, the controller sets a second feature section for each said second candidate as for a style type different in feature from a corresponding related first index that represents said related first index corresponding to the intended second candidate among said plurality of style types, said second feature section including a feature of said intended second candidate region but not including a feature of said corresponding related first index,

groups into region groups at least one or both of the related first indexes and the second candidates having the feature included in said set second feature section,

calculates a second index evaluation degree, based on a number of members of each region group with respect to each said second candidate, and

determines whether or not a logical element of each said second candidate is the index, based on said calculated second index evaluation degree.

2. The document image processing apparatus according to claim 1, wherein in said second evaluating process, said controller groups both of said related first indexes and said second candidates having the feature included in said second feature section, with respect to each said second candidate, and calculates said second index evaluation degree, based on the sizes of a first index group and a second candidate group.

3. The document image processing apparatus according to claim 1, wherein in said second evaluating process, said controller further groups at least one of the related first indexes and the second candidates having a feature not included in said second feature section, with respect to each said second candidate, and uses a size of each region group not included in said second feature section in calculating said second evaluation degree.

4. The document image processing apparatus according to claim 1, wherein said controller further changes said second feature section when there is no difference in value in said second index evaluation degree provided with respect to each said second candidate in setting said second feature section in said second evaluating process.

5. The document image processing apparatus according to claim 1, wherein said plurality of style types comprises any one of an indent amount, a justified line, distances from previous and next character string element regions, a character string decoration, a head character kind, an end character kind, a language type, a character size, a line height, a character interval, a font, a character decoration, a character color, and a background color.

6. The document image processing apparatus according to claim 1, wherein in said second evaluating process, said controller sets a predetermined range around a feature value of said intended second candidate as said intended second feature section.

7. The document image processing apparatus according to claim 1 wherein in said second evaluating process, said controller sets a boundary point between a feature value of said intended second candidate and a feature value of said corresponding related first index, and sets said second feature section so as to include the side of the feature value of said intended second candidate from said boundary point.

8. The document image processing apparatus according to claim 1, wherein

in said second evaluating process, said controller further compares features of the small region determined as said second index and other small regions, and determines the small region having a feature similar to that of said determined small region among said other small regions, as a new second index.

9. The document image processing apparatus according to claim 1, wherein

when said second index is extracted, said controller repeats said second determining process and said second evaluating process by replacing said second index with said first index.

10. The document image processing apparatus according to claim 1, wherein

in said second evaluating process, said controller compares features of the intended second candidate and the related large region regarding the related first index corresponding to said intended second candidate, with respect to each said second candidate, and removes the second candidate having a feature similar to the feature of said related large region, from the index candidate.

11. A document image processing method executed by a document image processing apparatus comprising a memory storing a document image, to extract an index region from said document image stored in said memory, and comprising the steps of:

classifying a plurality of character string element regions constituting said document image into small regions and large regions;

determining each small region positioned just before said large region according to a reading order as a first candidate;

determining at least one part of said first candidates as a first index, by performing a first evaluating process to evaluate whether or not each said first candidate is an index, based on a difference in feature from the related large region, with respect to each said first candidate;

determining each small region positioned just before said first index according to the reading order as a second candidate;

determining at least one part of said second candidates as a second index, by performing a second evaluating process to evaluate whether or not said second candidate is the index, based on a difference in feature from the related first index, with respect to each said second candidate; and

extracting the small regions determined as said first index and said second index, as said index region,

wherein said first evaluating process comprises

setting a first feature section for each said first candidate as for a style type different in feature from a corresponding related large region that represents said related large region corresponding to the intended first candidate among a plurality of style types, said first feature section including a feature of said intended first candidate region but not including a feature of said corresponding related large region,

grouping into region groups at least one or both of the related large regions and the first candidates having the feature included in said set first feature section,

calculating a first index evaluation degree, based on a number of members of each region group with respect to each said first candidate, and

determining whether or not a logical element of each said first candidate is the index, based on said calculated first index evaluation degree, and said second evaluating process comprises

setting a second feature section for each said second candidate as for a style type different in feature from a corresponding related first index that represents said related first index corresponding to the intended second candidate among said plurality of style types, said second feature section including a feature of said intended second candidate region but not including a feature of said corresponding related first index,

grouping into region groups at least one or both of the related first indexes and the second candidates having the feature included in said set second feature section,

calculating a second index evaluation degree, based on a number of members of each region group with respect to each said second candidate, and

determining whether or not a logical element of each said second candidate is the index, based on said calculated second index evaluation degree.

12. A computer-readable non-transitory recording medium having a recorded document image processing program comprising the steps of:

classifying a plurality of character string element regions constituting a document image into small regions and large regions;

determining each small region positioned just before said large region according to a reading order as a first candidate;

determining at least one part of said first candidates as a first index, by performing a first evaluating process to evaluate whether or not each said first candidate is an index, based on a difference in feature from the related large region, with respect to each said first candidate;

determining each small region positioned just before said first index according to the reading order as a second candidate;

determining at least one part of said second candidates as a second index, by performing a second evaluating process to evaluate whether or not said second candidate is the index, based on a difference in feature from the related first index, with respect to each said second candidate; and

extracting the small regions determined as said first index and said second index, as said index region wherein said first evaluating process comprises

setting a first feature section for each said first candidate as for a style type different in feature from a corresponding related large region that represents said related large region corresponding to the intended first candidate among a plurality of style types, said first feature section including a feature of said intended first candidate region but not including a feature of said corresponding related large region,

grouping into region groups at least one or both of the related large regions and the first candidates having the feature included in said set first feature section,

calculating a first index evaluation degree, based on a number of members of each region group with respect to each said first candidate, and

determining whether or not a logical element of each said first candidate is the index, based on said calculated first index evaluation degree, and

said second evaluating process comprises

setting a second feature section for each said second candidate as for a style type different in feature from a corresponding related first index that represents said related first index corresponding to the intended second candidate among said plurality of style types, said second feature section including a feature of said intended second candidate region but not including a feature of said corresponding related first index,

grouping into region groups at least one or both of the related first indexes and the second candidates having the feature included in said set second feature section,

calculating a second index evaluation degree, based on a number of members of each region group with respect to each said second candidate, and

determining whether or not a logical element of each said second candidate is the index, based on said calculated second index evaluation degree.