JP2012190315A - Image processing device and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processing device capable of efficiently determining a feature character string of a document constituted of a plurality of pages compared with a case having no such configuration.SOLUTION: A feature character string candidate extraction section 32 extracts candidates of a feature character string that is a character string on a document, for each page one by one. A page information calculation section 34 calculates page information on pages related to each of the feature character string candidates for each page. A feature character string determination section 36, whenever the feature character string candidates are calculated from each page and page information related to the feature character string candidates are calculated, determines the feature character string of the entire document based on the page information related to the feature character string candidates.

Description

  The present invention relates to an image processing apparatus and a program.

  Patent Literature 1 acquires the position of a title character string based on the position of a keyword character string described in the vicinity of the title character string and the relative position of the title character string with respect to the keyword character string, and the position of the title character string. The title extraction device which outputs the data of a title character string based on this is disclosed.

JP 2008-77454 A

  An object of the present invention is to provide an image processing apparatus that can efficiently determine a character string of a document composed of a plurality of pages as compared with a case where the document does not have this configuration.

  According to the first aspect of the present invention, there is provided candidate extraction means for extracting, for each page, a feature character string candidate that is a character string related to the original document from each page of a plurality of pages. A coefficient for calculating a coefficient on the basis of information on the page in accordance with a predetermined condition for each candidate for a characteristic character string extracted for each page by the candidate extraction unit each time a candidate for a characteristic character string is extracted from a page Whether the coefficient calculated by the coefficient calculation means exceeds a predetermined reference value each time the coefficient is calculated for the candidate of the character string extracted from one page by the calculation means and the coefficient calculation means And when the calculated coefficient exceeds a predetermined first reference value, a candidate for the coefficient exceeding the first reference value is determined as a characteristic character string of the document. An image processing device having a symptom string determination means.

  According to a second aspect of the present invention, there is provided time calculation means for calculating at least the time required for processing of the candidate extraction means for each page every time a candidate for a character string is extracted from one page by the candidate extraction means. In addition, each time the time calculation unit calculates the time required for processing of the candidate extraction unit for one page, the characteristic character string determination unit determines the time for each page calculated by the time calculation unit. It is determined whether or not the accumulated time exceeds a predetermined reference time, and when the accumulated time exceeds a predetermined reference time, a candidate for the maximum coefficient among the calculated coefficients is selected as the characteristic character string of the document. The image processing apparatus according to claim 1, wherein:

  According to a third aspect of the present invention, the characteristic character string determining means is a case where the accumulated time of the time exceeds a predetermined reference time, and the largest coefficient among the calculated coefficients is the first coefficient. 3. The image processing apparatus according to claim 2, wherein a character string related to predetermined information is determined as a characteristic character string of the document when the reference value is equal to or smaller than a second reference value smaller than the reference value.

  The present invention according to claim 4 is characterized in that the characteristic character string determining means determines the maximum of the calculated coefficients when the calculated coefficient does not exceed the first reference value for all pages of the document. The image processing apparatus according to claim 1, wherein a candidate for a characteristic character string related to the coefficient is determined as a characteristic character string of the document.

  The present invention according to claim 5 is characterized in that the characteristic character string determining means is a case where the coefficient does not exceed the first reference value for all pages of the document, and is the largest coefficient among the calculated coefficients. 5. The image processing apparatus according to claim 4, wherein a character string relating to predetermined information is determined as a characteristic character string of the document when is equal to or smaller than a second reference value smaller than the first reference value.

  According to a sixth aspect of the present invention, there is provided candidate extraction means for extracting, for each page, a feature character string candidate that is a character string related to the original document from each page of a plurality of pages. A coefficient for calculating a coefficient on the basis of information on the page in accordance with a predetermined condition for each candidate for a characteristic character string extracted for each page by the candidate extraction unit each time a candidate for a characteristic character string is extracted from a page A calculation unit; a time calculation unit that calculates at least a time required for processing of the candidate extraction unit for each page each time a candidate for a characteristic character string is extracted from one page by the candidate extraction unit; and the time calculation unit Every time the time required for processing by the candidate extracting means is calculated for one page, the accumulated time of each page calculated by the time calculating means is determined. And determining a candidate for the largest coefficient among the calculated coefficients as a characteristic character string of the document when the accumulated time exceeds a predetermined reference time. An image processing apparatus having character string determination means.

  According to a seventh aspect of the present invention, the characteristic character string determining means is a case where the accumulated time of the time exceeds a predetermined reference time, and the largest coefficient among the calculated coefficients is the first coefficient. The image processing apparatus according to claim 6, wherein when it is equal to or smaller than a second reference value smaller than the reference value, a character string related to predetermined information is determined as a characteristic character string of the document.

  According to an eighth aspect of the present invention, there is provided a candidate extraction step for extracting, for each page, a feature character string candidate that is a character string related to a document from each page of a document composed of a plurality of pages. A coefficient that calculates a coefficient based on information about a page according to a predetermined condition for each of the candidate candidate character strings extracted for each page in the candidate extraction step each time a candidate candidate character string is extracted from the page Whether the coefficient calculated by the coefficient calculating step exceeds a predetermined reference value each time a coefficient is calculated for the candidate candidate character string extracted from one page by the coefficient calculating means And when the calculated coefficient exceeds a predetermined first reference value, candidates relating to the coefficient exceeding the first reference value are selected as characteristic characters of the document. An image processing program for executing the feature character string determining step in a computer to determine a.

  According to the first aspect of the present invention, there is provided an image processing apparatus capable of efficiently determining a character string of a document composed of a plurality of pages as compared with a case where the characteristic character string is not provided. Can do.

  According to the second aspect of the present invention, in addition to the effect obtained by the first aspect of the present invention, the time required for determining the character string of the original is compared with the case where the present configuration is not provided. Can be reduced.

  According to the third aspect of the present invention, in addition to the effect obtained by the second aspect of the present invention, it is possible to eliminate the determination of an inappropriate character string as a characteristic character string of a document.

  According to the present invention of claim 4, in addition to the effect obtained by the present invention of claim 1, the characteristic character string of the document can be determined after processing all the pages.

  According to the fifth aspect of the present invention, in addition to the effect obtained by the present invention according to the fourth aspect, it can be excluded that a character string inappropriate as the characteristic character string of the document is determined as the characteristic character string.

  According to the sixth aspect of the present invention, there is provided an image processing apparatus capable of reducing the time required to determine the character string of a document composed of a plurality of pages compared to the case where the present configuration is not provided. it can.

  According to the present invention of claim 7, in addition to the effect obtained by the present invention of claim 6, it can be excluded that a character string inappropriate as a characteristic character string of a document is determined as a characteristic character string.

  According to the eighth aspect of the present invention, there is provided an image processing program capable of determining a character string of a document composed of a plurality of pages more efficiently than a case where the character string is not provided with the present configuration. Can do.

It is a figure which illustrates the hardware constitutions of the image processing apparatus concerning this embodiment. 2 is a processing program that operates in the image processing apparatus illustrated in FIG. 1. It is a figure which shows the structure of the characteristic character string candidate extraction part shown in FIG. It is a figure which shows the structure of the page information calculation part shown in FIG. It is a figure which shows the structure of the characteristic character string determination part shown in FIG. It is a figure which illustrates position score reference information, scale score reference information, and frequency score reference information. It is a flowchart which shows the process of a processing program. It is a figure which illustrates the processing result of the image processing apparatus which concerns on this embodiment.

FIG. 1 is a diagram illustrating a hardware configuration of an image processing apparatus 2 according to the present embodiment.
As illustrated in FIG. 1, the image processing apparatus 2 includes a control device 21 including a calculation unit 212 such as a CPU and a storage unit 214 such as a memory, a communication device 22, a recording device 24, and a user interface device (UI). Device) 25, a printing device 26, and an image reading device 27.

The UI device 25 includes a display device such as an LCD (Liquid Crystal Display) display device or a CRT (Cathode Ray Tube) display device, and a keyboard / touch panel.
The printing device 26 is, for example, a printer, and prints character data or image data on a recording medium such as paper.
The image reading device 27 is a scanner or the like, for example, and reads an image or the like from a recording medium such as a document, and converts it into, for example, read information in a bitmap format.
That is, the image processing apparatus 2 has a hardware configuration part as a computer capable of information processing and communication with other image processing apparatuses or terminals.
In the following drawings, substantially the same components and processes are denoted by the same reference numerals.
In the present embodiment, the image processing apparatus 2 includes the printing apparatus 26 and the image reading apparatus 27. However, the image processing apparatus may be, for example, a PC that does not include the printing apparatus and the image reading apparatus. In this case, the image processing apparatus may be connected to the image reading apparatus via a LAN (Local Area Network) or the like.

FIG. 2 is a diagram showing the configuration of the processing program 3 that operates in the image processing apparatus 2 shown in FIG.
As shown in FIG. 2, the processing program 3 includes a document reading information receiving unit 302, a document reading information management unit 304, a layout analysis unit 308, a character string extraction unit 310, a characteristic character string candidate extraction unit 32, and a page information calculation unit 34. , A characteristic character string candidate management unit 358 and a characteristic character string determination unit 36.
The processing program 3 is supplied to the image processing apparatus 2 via, for example, the storage medium 240 (FIG. 1), loaded into the storage unit 214, and installed on the image processing apparatus 2 on the OS (not shown). It is executed by specifically using the hardware resources of the processing device 2.
In the present embodiment, the processing program 3 is realized by software, but all or part of the processing program 3 may be realized by hardware such as an FPGA (Field Programmable Gate Array). .

FIG. 3 is a diagram illustrating a configuration of the feature character string candidate extraction unit 32 illustrated in FIG. 2.
As shown in FIG. 3, the feature character string candidate extraction unit 32 includes a character string position calculation unit 322, a character string position scoring unit 324, a character string scale calculation unit 326, a character string scale scoring unit 328, and a character string frequency calculation unit 330. , A character string frequency scoring unit 332, a scoring total unit 334, a scoring storage unit 336, and a characteristic character string candidate determination unit 338.
FIG. 4 is a diagram showing a configuration of the page information calculation unit 34 shown in FIG.
As shown in FIG. 4, the page information calculation unit 34 includes a scoring point extraction unit 342, a candidate coefficient calculation unit 344, a processing time calculation unit 346, an accumulated processing time calculation unit 348, and a processing time management unit 350.

FIG. 5 is a diagram showing a configuration of the characteristic character string determination unit 36 shown in FIG.
As shown in FIG. 5, the characteristic character string determination unit 36 includes a candidate coefficient comparison unit 362, an accumulated processing time comparison unit 364, a processed page number determination unit 366, and a characteristic character string determination unit 368.

In the processing program 3 (FIG. 2), the document reading information receiving unit 302 receives the reading information (document reading information) obtained from the image reading device 27 and sends the received document reading information to the document reading information management unit 304. Output.
The document reading information management unit 304 stores and manages document reading information from the document reading information receiving unit 302.
Further, when the document is composed of a plurality of pages, the document reading information management unit 304 causes the layout analysis unit 308 and the character string extraction unit 310 to perform processing for each page in response to a command from the characteristic character string determination unit 36. To control.
That is, after the processing of one page is completed, the document reading information management unit 304 receives the instruction indicating that the next page processing is to be performed from the characteristic character string determination unit 36, and the layout analysis unit 308 and The character string extraction unit 310 is controlled to perform processing for each page.

The layout analysis unit 308 analyzes the document reading information, classifies natural images such as characters, tables, and photographs, CG (Computer Graphics), or paintings included in the document (object classification), and associates position information with each of them. .
Further, the layout analysis unit 308 outputs information indicating the analysis result (placement information) to the feature character string candidate extraction unit 32.
Here, the arrangement information is information indicating which object (natural image such as a character, a table, or a photograph, a CG or a painting) is included at which position and in which position in the document corresponding to the document reading information. is there.

This arrangement information includes, for example, position information indicating the position of each object and scale information indicating the scale (size, area, etc.) of each object.
Here, for example, the position information may indicate an absolute position such as position coordinates, or may indicate a relative positional relationship with another character string or the like.
Similarly, the size information may indicate the absolute size of the object, for example, the font size or the occupied area, or may indicate the relative size between other objects. Alternatively, it may indicate a difference from the average value of the object scale.
The above-described classification by the layout analysis unit 308 is performed by, for example, detecting various lines, frame lines, ruled lines, or color information included in the document, edge detection, and pattern matching. However, it is not restricted to these methods.

The character string extraction unit 310 analyzes the document reading information by using, for example, an OCR (Optical Character Recognition) function, and converts the character string included in the document into a predetermined character string alone by, for example, morphological analysis. Extract in a form with meaning.
Here, the character recognition refers to generating character data (character string) by specifying the character by comparing the image data of the character obtained by reading with a pattern stored in advance.
The morpheme analysis is, for example, classifying one sentence into morphemes (the smallest language unit having meaning) based on information on grammatical rules stored in advance and a dictionary in which words are registered. This is a process for discriminating the part of speech of a morpheme.
In this morphological analysis process, the language of the character string is determined (that is, for example, whether the character string is Japanese, English, or another language).
Further, the character string extraction unit 310 outputs each extracted character string to the feature character string candidate extraction unit 32.

The feature character string candidate extraction unit 32 extracts one feature character string candidate (characteristic character string candidate) that is a character string related to the document, and outputs it to the feature character string candidate management unit 358. .
Here, the characteristic character string is, for example, a document title (title) or the like, and is a character string for a human to identify the document.
The characteristic character string may be the name of the electronic data or a path folder (directory) for storing the electronic data when the document reading information is converted into electronic data (electronic file) or the like.
The characteristic character string candidate is a character string that can be a characteristic character string, and is determined for each page constituting the document.
If it is determined by a process described later that a certain characteristic character string candidate satisfies the characteristic character string condition, the characteristic character string candidate is determined as a characteristic character string for the entire document.

In the feature character string candidate extraction unit 32, the character string position calculation unit 322 (FIG. 3) accepts the arrangement information about the processing target page from the arrangement analysis unit 308, and the character string extraction unit 310 receives the arrangement information on the processing target page. Accept a string.
Further, the character string position calculation unit 322 calculates position information for each received character string based on the position information included in the received arrangement information, associates each character string with the position information, and sets the character string position. The data is output to the scoring unit 324.
The character string size calculation unit 326 accepts arrangement information about the processing target page from the arrangement analysis unit 308, and accepts a character string on the processing target page from the character string extraction unit 310.
In addition, the character string scale calculation unit 326 calculates scale information for each received character string based on the scale information included in the received arrangement information, associates each character string with the scale information, and sets the character string scale. The data is output to the scoring unit 328.
The character string frequency calculation unit 330 receives the character string on the processing target page from the character string extraction unit 310.
In addition, the character string frequency calculation unit 330 calculates the appearance frequency in the processing target for each received character string, generates frequency information indicating the appearance frequency, associates each character string with the frequency information, Output to the column frequency scoring unit 332.

The character string position scoring unit 324 stores in advance information (position scoring reference information) indicating a scoring standard indicating the relationship between the position information of the character string and the score calculated for the character string.
The character string position scoring unit 324 calculates a score (position scoring) for each character string according to the position scoring reference information, associates each character string with the position scoring, and outputs it to the scoring total unit 334 To do.
The character string scale scoring unit 328 stores in advance information (scale scoring reference information) indicating a scoring standard indicating the relationship between the scale information of the character string and the score calculated for the character string.
The character string scale scoring unit 328 calculates a score (scale scoring) for each character string in accordance with the scale scoring reference information, associates each character string with the scale scoring, and outputs it to the scoring total unit 334 To do.
The character string frequency scoring unit 332 stores in advance information (scale scoring reference information) indicating a scoring standard indicating the relationship between the frequency information of the character string and the score calculated for the character string.
The character string frequency scoring unit 332 calculates a score (frequency scoring) for each character string according to the frequency scoring reference information, associates each character string with the frequency scoring, and outputs it to the scoring total unit 334 To do.

FIG. 6 is a diagram illustrating position scoring reference information, scale scoring reference information, and frequency scoring reference information, (A) is a diagram for explaining position scoring reference information, and (B) is a position scoring reference. Information is illustrated, (C) illustrates scale score reference information, and (D) illustrates frequency score reference information.
As shown in FIG. 6A, for example, the position scoring reference information represents the vertical direction of the original page by relative position coordinates with the upper end being 0% and the lower end being 100%, and the horizontal direction of the original page is indicated. , And the relative position coordinates where the left end is 0% and the right end is 100%.
Further, for example, the position scoring reference information uses the center of the character string as a reference point for calculating the position of the character string.

In this case, in the example shown in FIG. 6B, the character string position scoring unit 324 indicates that the character string exists in the vertical direction exceeding 20% (that is, the lower 4/5) according to the position scoring reference information. Calculates 0.0 points for the character string, and if the character string is 20% or less in the vertical direction (that is, the upper 1/5), calculates 0.1 points for the character string.
In the example shown in FIG. 6B, the character string position scoring unit 324 indicates that the character string is 0 for the character string when the character string is less than 40% or more than 60% in the horizontal direction according to the position scoring reference information. 0 points are calculated, and when the character string is present in the horizontal direction of 40% or more and 60% or less, 0.1 point is calculated for the character string.
For example, in the example shown in FIG. 6A, the character string position scoring unit 324 has 0.0 (vertical direction) +0.1 (horizontal direction) = 0.1 point for the character string “author Fujitaro”. Is calculated.

In the example shown in FIG. 6C, the character string scale scoring unit 328 determines that the scale (font size, etc.) of the character string to be processed is all characters of the document page to be processed according to the scale score reference information. If it is less than twice the average of the size of the column, it calculates 0.0 points for that character string, and if it is more than twice the average and less than 5 times, it calculates 0.1 points for that character string If it is 5 times or more, 0.5 points are calculated for the character string.
In the example shown in FIG. 6D, the character string frequency scoring unit 332 performs processing for the character string if the appearance frequency (number of appearances) of the character string to be processed is one or less on the document page to be processed. If 0.0 is calculated, if the appearance frequency is 2 or more and 4 or less, 0.3 points are calculated for the character string, and if the appearance frequency is 5 or more, 0.5 points for the character string are calculated. Calculate points.

When there are a plurality of the same character strings on the document page to be processed, the character string position scoring unit 324 calculates a score each time the character string appears, and adds the calculated scores to the character string. It is good also as a position scoring corresponding to a line | column, and it is good also considering the maximum value of each calculated score as a position scoring corresponding to the character string.
Similarly, when there are a plurality of the same character strings on the document page to be processed, the character string scale scoring unit 328 calculates a score each time the character string appears, and sums the calculated scores. It is good also as a scale score corresponding to a character string, and it is good also considering the maximum value of each calculated score as a scale score corresponding to the character string.

The score totaling unit 334 (FIG. 3) sums the position score, the scale score, and the frequency score for each different character string, gives the total score to the character string, and is given as the character string. The total points (granting points) are associated with each other and output to the point management unit 336.
The score management unit 336 associates the character string with the assigned score and stores and manages each character string different from each other.
Note that the score management unit 336 may store not only the assigned score but also the breakdown of the assigned score (that is, the position score, the scale score, and the frequency score) for each character string.

The feature character string candidate determination unit 338 determines the character string related to the maximum assigned score as the feature character string candidate of the page when the assigned score is calculated for all the character strings on each page of the document. The characteristic character string candidates are output to the characteristic character string candidate management unit 358 and the page information calculation unit 34.
The characteristic character string candidate management unit 358 (FIG. 2) stores and manages the characteristic character string candidates determined by the characteristic character string candidate extraction unit 32 and page information described later for each page.

In the present embodiment, the score totaling unit 334 calculates the assigned score by adding all of the position score, the scale score, and the frequency score. However, at least one of the position score, the scale score, and the frequency score is calculated. Any given total may be used as the assigned points.
In this case, the position score, the scale score, or the frequency score that do not constitute the assigned score need not be the calculation target.
That is, for example, when the assigned score is the sum of the scale score and the frequency score, the character string position calculation unit 322 does not have to calculate the position information for each character string, and the character string position score unit 324 Need not be calculated.
Furthermore, in the present embodiment, the score totaling unit 334 calculates the assigned score by adding all of the position score, the scale score, and the frequency score, but adds the score to the assigned score by other criteria. May be.

The page information calculation unit 34 (FIG. 2) calculates information (page information) related to each of the characteristic character string candidates for each page, and outputs the page information to the characteristic character string candidate management unit 358.
Here, the page information includes at least a candidate coefficient and a processing time, and is calculated for each page.
The candidate coefficient is an index (accuracy) indicating the certainty of the corresponding characteristic character string candidate as the characteristic character string of the entire document. For example, the position, scale, and frequency of the characteristic character string candidate on each page Etc.
The processing time means at least a period required for extracting characteristic character string candidates from each page.

In the page information calculation unit 34, the scoring point extraction unit 342 (FIG. 4) accepts the feature character string candidate from the feature character string candidate extraction unit 32, and extracts the assigned point assigned to the feature character string candidate and extracts the feature character string candidate. Extracted from the point management unit 336 of the unit 32.
In addition, the score extraction unit 342 outputs the extracted assigned score to the candidate coefficient calculation unit 344.
The candidate coefficient calculation unit 344 calculates a candidate coefficient based on the assigned score from the score extraction unit 342.
Further, candidate coefficient calculation section 344 outputs the candidate coefficient to characteristic character string candidate management section 358 in association with the characteristic character string candidate related to the candidate coefficient.

The candidate coefficient calculation unit 344 may use the assigned scoring as it is as the candidate coefficient when calculating the candidate coefficient. However, the weight for the given scoring may be lowered as the processed page advances.
That is, when [candidate coefficient] = a [giving score], a may be 1 for all pages, or a may be reduced as the processed pages progress.

For example, the assigned score of the feature character string candidate # 1 of the first processed page is S1, the assigned score of the feature character string candidate # 2 of the second processed page is S2, and the third process Assume that the assigned score of the feature character string candidate # 3 of the page that has been set is S3.
In this case, the candidate coefficient calculation unit 344 sets the candidate coefficient related to the characteristic character string candidate # 1 to S1, sets the candidate coefficient related to the characteristic character string candidate # 2 to S2 × 0.9, and sets the candidate coefficient related to the characteristic character string candidate # 3. A candidate coefficient may be calculated as S3 × 0.8.

In addition, the candidate coefficient calculation unit 344 may change the weighting of the candidate coefficient according to the creator of the document.
For example, if the creator of the manuscript is the general affairs department, when it is determined that the character string related to the form is a feature character string candidate, the candidate coefficient is increased by increasing the weighting of the feature character string candidate. In the case where the calculated document creator is the development department, when it is determined that the character string related to the drawing is a feature character string candidate, the candidate character coefficient is increased by increasing the weighting to the assigned character string of the feature character string candidate. May be calculated.
As a method for specifying the creator of the document, for example, a method in which the user operates and inputs the UI device 25, or a character string related to the creator described in the document is recognized. There is a way to identify. However, it is not restricted to these methods.
In addition, as a method for determining whether or not the feature character string candidate is a character string related to the creator, for example, a dictionary or the like storing a correspondence table of the character strings related to the creator and the creator is used. There is a way. However, it is not restricted to these methods.

In the above embodiment, the scoring point extraction unit 342 calculates the candidate coefficient based on the assigned scoring extracted from the feature character string candidate extraction unit 32, but does not calculate the candidate coefficient based on the given scoring point. Also good.
For example, any one or more of the position score, the scale score, and the frequency score regarding the feature character string candidates may be summed, and any one of the weights may be changed.

Specifically, for example,
[Formula 1] [Candidate Coefficient] = a × [Location Scoring] + b × [Scale Scoring] + c × [Frequency Scoring]
And
At this time, when the assigned score is used as a candidate coefficient as it is, a = b = c = 1 in Equation 1, but a ≠ b, b ≠ c, c ≠ a may be used.
In Formula 1, any one or two of a, b, and c may be 0.

Further, the page information calculation unit 34 may separately calculate candidate coefficients based on the character strings on each page and their arrangements, without extracting any information from the feature character string candidate extraction unit 32.
In this case, the candidate coefficient may be calculated using a standard different from the position score reference information, the scale score reference information, and the scale score reference information used by the feature character candidate extraction unit 32.

The processing time calculation unit 346 calculates a time (processing time) required to extract at least a feature character string candidate from a certain page.
Specifically, the processing time calculation unit 346 obtains, from the document reading information management unit 304, the time at which the layout analysis unit 308 or the character string extraction unit 310 starts processing the page to be processed, whichever comes first. Extracted as the start time of page processing.
Further, the processing time calculation unit 346 extracts the time when the characteristic character string candidate is extracted for the page to be processed from the characteristic character string candidate extraction unit 32 as the processing end time of the page.
Furthermore, the processing time calculation unit 346 calculates the processing time by subtracting the start time from the end time, and the information (processing time information) indicating the calculated processing time is used as the cumulative processing time calculation unit 348 and the processing time. Output to the management unit 350.

Note that the processing time calculation unit 346 calculates the total time of the time required for extracting the characteristic character string candidate from the certain page and the time required for the processing of the scoring point extraction unit 342 and the candidate coefficient calculation unit 344, The processing time may be calculated.
In this case, the processing time calculation unit 346 extracts the time when the candidate coefficient is calculated from the candidate coefficient calculation unit 344 as the processing end time of the page, and subtracts the start time from the end time, thereby processing the processing. You may calculate time.

The processing time management unit 350 stores and manages processing time information.
When the accumulated processing time calculation unit 348 receives processing time information regarding a certain page from the processing time calculation unit 346, the processing time information regarding all pages processed before the processing for the page is received from the processing time management unit 350. To extract.
Further, the accumulated processing time calculation unit 348 adds up the processing time indicated by the processing time information regarding a certain page and the processing times indicated by the processing time information for all the previously processed pages, thereby calculating the accumulated processing time. calculate.
Further, the cumulative processing time calculation unit 348 uses the information indicating the calculated cumulative processing time (cumulative processing time information) as information indicating the cumulative processing time related to the characteristic character string candidate corresponding to the page, and manages the characteristic character string candidate management. Output to the unit 358.

The feature character string candidate management unit 358 (FIG. 2) stores and manages the feature character string candidate, the candidate coefficient related to the feature character string candidate, and the accumulated processing time related to the feature character string candidate for each page.
Each time the characteristic character string candidate is calculated from each page and the page information (candidate coefficient and cumulative processing time) regarding the characteristic character string candidate is calculated, the characteristic character string determination unit 36 calculates page information regarding the characteristic character string candidate ( The characteristic character string of the entire document is determined based on the candidate coefficient and the accumulated processing time.
In the characteristic character string determination unit 36, the candidate coefficient comparison unit 362 (FIG. 5), when a candidate coefficient is calculated for each page and stored in the characteristic character string candidate management unit 358, the candidate coefficient Extracted from the character string candidate management unit 358.
Further, the candidate coefficient comparison unit 362 compares the candidate coefficient with a predetermined candidate coefficient reference value, and determines whether or not the candidate coefficient exceeds the candidate coefficient reference value.
When the candidate coefficient exceeds the candidate coefficient reference value, the candidate coefficient comparison unit 362 outputs information indicating that (candidate coefficient reference value excess information) to the feature character string determination unit 368, and the candidate coefficient is If the candidate coefficient reference value is not exceeded, the candidate coefficient comparison unit 362 outputs information indicating that (candidate coefficient reference value non-excess information) to the feature character string determination unit 368.

When the characteristic character string determination unit 368 receives candidate coefficient reference value excess information from the candidate coefficient comparison unit 362, the characteristic character string determination unit 368 extracts a characteristic character string candidate related to the candidate coefficient from the characteristic character string candidate management unit 358, and the characteristic character string Candidates are determined as characteristic character strings of the entire document.
Further, the characteristic character string determination unit 368 outputs the determined characteristic character string to the UI device 25 and causes the UI device 25 to display the characteristic character string.
In this case, the processing program 3 ends all the processing without performing the processing for the unprocessed page.
That is, when there is a candidate coefficient exceeding the candidate coefficient reference value, even if there is a page from which no feature character string candidate has been extracted, the unprocessed page is not processed.

When the accumulated processing time is calculated for each page and stored in the feature character string candidate management unit 358, the accumulated processing time comparison unit 364 displays the accumulated processing time information indicating the accumulated processing time each time as a feature character string candidate. Extracted from the management unit 358.
The accumulated processing time comparison unit 364 compares the accumulated processing time indicated by the accumulated processing time information with a predetermined accumulated processing time reference value, and determines whether or not the accumulated processing time exceeds the accumulated processing time reference value. To do.
When the accumulated processing time exceeds the accumulated processing time reference value, the accumulated processing time comparison unit 364 outputs information indicating that (accumulated processing time reference value excess information) to the feature character string determination unit 368. When the cumulative processing time does not exceed the cumulative processing time reference value, the cumulative processing time comparison unit 364 sends information indicating that fact (cumulative processing time reference value non-excess information) to the feature character string determination unit 368. Output.

The feature character string determination unit 368 extracts, from the feature character string candidate management unit 358, the candidate coefficients for all the processed pages when the accumulated processing time reference value excess information is received from the accumulated processing time comparison unit 364. To do.
Also, the characteristic character string determination unit 368 extracts the characteristic character string candidate related to the largest candidate coefficient from the extracted candidate coefficients from the characteristic character string candidate management unit 358, and the characteristic character string candidate is extracted from the entire document. Determine as a string.
Further, the characteristic character string determination unit 368 outputs the determined characteristic character string to the UI device 25 and causes the UI device 25 to display the characteristic character string.
In this case, the processing program 3 ends all the processing without performing the processing for the unprocessed page.
That is, when the accumulated processing time exceeds the accumulated processing time reference value, even if there is a page for which no feature character string candidate has been extracted yet, the unprocessed page is not processed.

When the maximum candidate coefficient among the extracted candidate coefficients is equal to or smaller than a predetermined minimum candidate coefficient reference value, the characteristic character string determination unit 368 determines whether the predetermined character string is a predetermined character string regardless of the characteristic character string related to the maximum candidate coefficient. A character string related to information may be determined as a characteristic character string.
This lowest candidate coefficient reference value is a value smaller than the candidate coefficient reference value, and indicates the lowest standard for a feature character string candidate to be valid as a feature character string.
That is, when the candidate coefficient is less than or equal to the minimum candidate coefficient reference value, the feature character string candidate has a higher allocation score than the other character strings on the page, but is determined as a characteristic character string of the entire document. Therefore, if the candidate character string candidate is a characteristic character string, the characteristic character string may not represent the content of the document.
Therefore, in such a case, the characteristic character string determination unit 368 determines predetermined information (for example, information regarding date and time or a user ID) as a characteristic character string.

Based on the information managed by the document reading information management unit 304, the processing page number determination unit 366 determines whether processing has been completed for all pages (feature character string candidates are extracted for all pages, candidate coefficients, Whether the processing time has been calculated).
When processing is completed for all pages, the processing page number determination unit 366 outputs information indicating that fact (all page end information) to the characteristic character string determination unit 368 and processes all pages. If the processing has not ended, the processing page number determination unit 366 outputs information indicating that (all page incomplete information) to the characteristic character string determination unit 368.

When the characteristic character string determination unit 368 receives the all page end information from the processing page number determination unit 366, the characteristic character string determination unit 368 extracts, from the characteristic character string candidate management unit 358, candidate coefficients for all pages that have been processed.
Also, the characteristic character string determination unit 368 extracts the characteristic character string candidate related to the largest candidate coefficient from the extracted candidate coefficients from the characteristic character string candidate management unit 358, and the characteristic character string candidate is extracted from the entire document. Determine as a string.
Further, the characteristic character string determination unit 368 outputs the determined characteristic character string to the UI device 25 and causes the UI device 25 to display the characteristic character string.
Note that, in the same way as when the accumulated processing time exceeds the accumulated processing time reference value, the feature character string determination unit 368 has the largest candidate coefficient among the extracted candidate coefficients equal to or less than a predetermined minimum candidate coefficient reference value. In this case, the character string related to the predetermined information may be determined as the characteristic character string regardless of the characteristic character string related to the maximum candidate coefficient.

On the other hand, the characteristic character string determination unit 368 accepts candidate coefficient reference value non-exceeding information from the candidate coefficient comparison unit 362, accepts cumulative processing time reference value non-excess information from the cumulative processing time comparison unit 364, and determines the number of processed pages. When the all-page non-completion information is received from the unit 366, the characteristic character string determination unit 368 outputs a command indicating that processing is to be executed for the next page to the document reading information management unit 304.
Thereby, the document reading information management unit 304 controls the next page to be processed by the arrangement analysis unit 308 and the character string extraction unit 310, and the arrangement analysis unit 308 and the character string extraction unit 310 perform processing. The feature character string candidate extraction unit 32 extracts feature character string candidates for the next page.

FIG. 7 is a flowchart (S10) showing processing of the processing program 3.
In step 100 (S100), document reading information receiving unit 302 receives document reading information obtained by reading a document.
In step 102 (S102), the document reading information management unit 304 sets the page to be processed as the first page.
In step 104 (S104), the layout analysis unit 308 generates layout information for each page based on the document reading information, and the character string extraction unit 310 extracts a character string.

In step 106 (S106), the characteristic character string candidate extraction unit 32 extracts characteristic character string candidates for each page.
In step 108 (S108), the page information calculation unit 34 calculates candidate coefficients.
In step 110 (S110), the page information calculation unit 34 calculates the accumulated processing time.

In step 112 (S112), the characteristic character string determination unit 36 determines whether or not the accumulated processing time exceeds the accumulated processing time reference value. If it is determined that the accumulated processing time exceeds, the process proceeds to S122 and determines that the accumulated processing time does not exceed. If so, the process proceeds to S114.
In step 114 (S114), the characteristic character string determination unit 36 determines whether the candidate coefficient exceeds the candidate coefficient reference value. When determining that the candidate coefficient exceeds the process, the process proceeds to S120. The process proceeds to S116.
In step 116 (S116), the characteristic character string determination unit 36 determines whether or not the processing has been completed for all pages. If it is determined that the processing has been completed, the process proceeds to S122 and determines that the processing has not been completed. In the case, the process proceeds to S118.
In step 118 (S118), the document reading information management unit 304 sets the next page as a processing target.

In step 120 (S120), the characteristic character string determination unit 36 determines a characteristic character string candidate related to a candidate coefficient that exceeds the candidate coefficient reference value as a characteristic character string of the document, and ends the process.
In step 122 (S122), the characteristic character string determination unit 36 determines whether or not the maximum candidate coefficient exceeds the minimum candidate coefficient reference value. If it is determined that the maximum candidate coefficient exceeds, the process proceeds to S124. If so, the process proceeds to S126.
In step 124 (S124), the characteristic character string determination unit 36 determines the characteristic character string candidate related to the maximum candidate coefficient as the characteristic character string of the document, and ends the process.
In step 126 (S126), the characteristic character string determination unit 36 determines a character string related to predetermined information as a characteristic character string of the document, and ends the process.
In the above-described embodiment, the first page to be processed is defined as the first page in S102. However, the order of the pages to be processed may not be the same as the order of the pages of the document.

Hereinafter, the processing of the image processing apparatus 2 according to the present embodiment will be described with specific examples.
8A and 8B are diagrams illustrating the processing result of the image processing apparatus 2 according to the present embodiment. FIG. 8A illustrates a first example, and FIG. 8B illustrates a second example. Show.
In the example shown in FIG. 8A, for the first page, the characteristic character string candidate # 1 “author Fujitaro” is extracted, and the candidate coefficient # 1 of the characteristic character string candidate # 1 “author Fujitaro” is The processing time # 1 for the first page is 2.0. At this time, the cumulative processing time # 1 is 2.0.
The unit of the processing time and the accumulated processing time is, for example, milliseconds, seconds or minutes, but is not limited thereto.

Also, for the second page, the characteristic character string candidate # 2 “Introduction” is extracted, the candidate coefficient # 2 of the characteristic character string candidate # 2 “Introduction” is 0.40, and the processing time # for the second page 2 is 5.0. At this time, the cumulative processing time # 2 is 7.0 (= 2.0 + 5.0).
Further, for the third page, the characteristic character string candidate # 3 “Japanese manufacturing white paper” is extracted, and the candidate coefficient # 3 of the characteristic character string candidate # 3 “Japanese manufacturing white paper” is 0.80, Processing time # 3 for the third page is 8.0. At this time, the cumulative processing time # 3 is 15.0 (= 2.0 + 5.0 + 8.0).

Here, when the candidate coefficient reference value is 0.70, the cumulative processing time reference value is 20.0, and the lowest candidate coefficient reference value is 0.10, the feature character string candidate # 3 “Japanese manufacturing industry” Candidate coefficient # 3 of “white paper” exceeds the candidate coefficient reference value.
Therefore, the characteristic character string determination unit 36 determines the characteristic character string candidate # 3 “Japanese manufacturing white paper” as the characteristic character string of the manuscript.
In this case, even if there are pages after the fourth page, the feature character string candidates are not extracted for these pages.

Further, when the candidate coefficient reference value is 0.85, the cumulative processing time reference value is 10.0, and the lowest candidate coefficient reference value is 0.10, the cumulative processing time # 2 on the second page is the cumulative processing. Although the time reference value has not been exceeded, the accumulated processing time # 3 on the third page exceeds the accumulated processing time reference value.
In addition, candidate coefficient # 3, which is the largest candidate coefficient at that time, exceeds the minimum candidate coefficient reference value.
Therefore, the characteristic character string determination unit 36 determines the characteristic character string candidate # 3 “Japanese manufacturing white paper” related to the candidate coefficient # 3, which is the largest candidate coefficient at that time, as the characteristic character string of the document.
In this case, even if there are pages after the fourth page, the feature character string candidates are not extracted for these pages.

  If the candidate coefficient reference value is 0.85, the cumulative processing time reference value is 20.0, and the third page is the last page, the characteristic character string determination unit 36 determines that the maximum candidate coefficient at that time is The characteristic character string candidate # 3 “Japanese manufacturing white paper” relating to the candidate coefficient # 3 is determined as the characteristic character string of the original.

In the example shown in FIG. 8B, the characteristic character string candidate # 1 “author” is extracted for the first page, and the candidate coefficient # 1 of the characteristic character string candidate # 1 “author” is 0.05. Yes, the processing time # 1 for the first page is 2.0. At this time, the cumulative processing time # 1 is 2.0.
Further, since the second page is blank or the like, the characteristic character string candidate # 2 is not extracted, and therefore the candidate coefficient # 2 of the characteristic character string candidate # 2 is 0.00. The processing time # 2 for the second page is 5.0. At this time, the cumulative processing time # 2 is 7.0 (= 2.0 + 5.0).
Also, for the third page, the characteristic character string candidate # 3 “reference” is extracted, the candidate coefficient # 3 of the characteristic character string candidate # 3 “reference” is 0.06, and the processing time # for the third page 3 is 8.0. At this time, the cumulative processing time # 3 is 15.0 (= 2.0 + 5.0 + 8.0).

Here, when the candidate coefficient reference value is 0.85, the accumulated processing time reference value is 10.0, and the lowest candidate coefficient reference value is 0.10, the accumulated processing time # 2 on the second page is accumulated. Although the processing time reference value is not exceeded, the cumulative processing time # 3 on the third page exceeds the cumulative processing time reference value.
Further, the candidate coefficient # 3 which is the maximum candidate coefficient at that time is equal to or less than the minimum candidate coefficient reference value.
Therefore, the characteristic character string determination unit 36 determines a character string related to the predetermined information as a characteristic character string of the document regardless of the characteristic character string candidate # 3 “reference”.
In this case, even if there are pages after the fourth page, the feature character string candidates are not extracted for these pages.

2 ... Image processing device,
3 ... Processing program,
302: Document reading information receiving unit,
304: Document reading information management unit,
308 ... Placement analysis unit,
310 ... character string extraction unit,
32 ... Characteristic character string candidate extraction unit,
322: Character string position calculation unit,
324: Character string position scoring section,
326... Character string scale calculation unit,
328 ... Character string scale scoring section,
330 ... character string frequency calculation unit,
332: Character string frequency scoring section,
334 ... Scoring total part,
336 ... Scoring storage unit,
338: Characteristic character string candidate determination unit,
34: Page information calculation unit,
342 ... Scoring extraction unit,
344 ... Candidate coefficient calculation unit,
346... Processing time calculation unit,
348 ... Cumulative processing time calculation unit,
350 ... processing time management part,
358... Character string candidate management unit,
36... Character string determination unit,
362 ... Candidate coefficient comparison unit,
364 ... Cumulative processing time comparison unit,
366... Processed page number determination unit,
368 ... Characteristic character string determination unit,

Claims (8)

Candidate extracting means for extracting, for each page, feature character string candidates that are character strings related to the document from each page of the document composed of a plurality of pages;
Each time a candidate for a characteristic character string is extracted from one page by the candidate extraction unit, information on the page is obtained according to a predetermined condition for each candidate for the characteristic character string extracted for each page by the candidate extraction unit. A coefficient calculating means for calculating a coefficient based on;
Each time a coefficient is calculated for the candidate candidate character string extracted from one page by the coefficient calculating means, it is determined whether the coefficient calculated by the coefficient calculating means exceeds a predetermined reference value, An image having characteristic character string determining means for determining, as a characteristic character string of the original, a candidate for a coefficient exceeding the first reference value when the calculated coefficient exceeds a predetermined first reference value; Processing equipment. Each time a candidate for a characteristic character string is extracted from one page by the candidate extraction unit, it further includes a time calculation unit that calculates at least the time required for processing of the candidate extraction unit for each page,
Each time the time required for processing of the candidate extracting unit is calculated for one page by the time calculating unit, the characteristic character string determining unit determines whether the accumulated time of each page calculated by the time calculating unit is a predetermined time. If the accumulated time exceeds a predetermined reference time, a candidate for the largest coefficient among the calculated coefficients is determined as a characteristic character string of the document. Item 8. The image processing apparatus according to Item 1. The characteristic character string determination means is a second case in which the accumulated time of the time exceeds a predetermined reference time, and a maximum coefficient among the calculated coefficients is smaller than the first reference value. The image processing apparatus according to claim 2, wherein when it is equal to or less than a reference value, a character string related to predetermined information is determined as a characteristic character string of the document. The characteristic character string determination means determines a characteristic character string candidate related to the maximum coefficient among the calculated coefficients when the calculated coefficient does not exceed the first reference value for all pages of the document. The image processing apparatus according to claim 1, wherein the image processing apparatus determines the characteristic character string of the document. The characteristic character string determination means is a case where the coefficient does not exceed the first reference value for all pages of the document, and the maximum coefficient among the calculated coefficients is greater than the first reference value. 5. The image processing apparatus according to claim 4, wherein a character string relating to predetermined information is determined as a characteristic character string of the document when the second reference value is equal to or smaller than the second reference value. Candidate extracting means for extracting, for each page, feature character string candidates that are character strings related to the document from each page of the document composed of a plurality of pages;
Each time a candidate for a characteristic character string is extracted from one page by the candidate extraction unit, information on the page is obtained according to a predetermined condition for each candidate for the characteristic character string extracted for each page by the candidate extraction unit. A coefficient calculating means for calculating a coefficient based on;
A time calculation unit that calculates at least the time required for processing of the candidate extraction unit for each page each time a candidate for a characteristic character string is extracted from one page by the candidate extraction unit;
Whether the accumulated time of the time for each page calculated by the time calculation means exceeds a predetermined reference time each time the time calculation means calculates the time required for processing of the candidate extraction means for one page And a characteristic character string determining means for determining, as the characteristic character string of the document, a candidate for the largest coefficient among the calculated coefficients when the accumulated time exceeds a predetermined reference time. apparatus. The characteristic character string determination means is a second case in which the accumulated time of the time exceeds a predetermined reference time, and a maximum coefficient among the calculated coefficients is smaller than the first reference value. The image processing apparatus according to claim 6, wherein when it is equal to or less than a reference value, a character string related to predetermined information is determined as a characteristic character string of the document. A candidate extraction step for extracting, for each page, a feature character string candidate, which is a character string related to the document, from each page of the document composed of a plurality of pages;
Each time a candidate for a characteristic character string is extracted from one page by the candidate extraction unit, information on the page is obtained according to a predetermined condition for each candidate for the characteristic character string extracted for each page in the candidate extraction step. A coefficient calculating step for calculating a coefficient based on the
Each time a coefficient is calculated for the candidate candidate character string extracted from one page by the coefficient calculation means, it is determined whether the coefficient calculated by the coefficient calculation step exceeds a predetermined reference value, A characteristic character string determination step for determining, as a characteristic character string of the document, a candidate for a coefficient that exceeds the first reference value when the calculated coefficient exceeds a predetermined first reference value; An image processing program to be executed.

Images