JP2008152502A - Document image retrieval device and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To shorten an index creation period for retrieval of document image, and make it possible to significantly reduce retrieval processing burden. <P>SOLUTION: A document image retrieval device performs index creation process 11b. Here, a text area of a document image transmitted from a scanner 20 is cut out and split into units of punctuation of the document and the number of characters is measured for every split unit to register it as an index. When performing retrieval of a document image data, the number of characters between punctuations of retrieval image data which is a retrieval source is measured (11c), an index having the same number of characters as the measured number of characters is detected (11d), and the document image data corresponding to the detected index is extracted from a memory unit 12 (11e). Information on the extracted document image data is displayed on a display input section 13 by display control function 11a. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a document image search apparatus and program, and more specifically, a document image search apparatus that searches document image data corresponding to search document data from previously registered document image data, and functions of the apparatus. It relates to the program to be realized.

2. Description of the Related Art There is known a document image search apparatus that searches document image data corresponding to search document data as a search source from document image data to be searched that is registered in advance.
In such a document image search apparatus, conventionally, for example, when document image data read from a scanner is registered as a search target, text extracted from the document image data by character recognition is created and registered as an index. When searching for registered document image data, first, an index is referenced to search for the same text as the search document data.

As a technique for generating and registering an index of document image data to be searched in this way, for example, there is a full-text search device disclosed in Patent Document 1.
The full-text search device of Patent Document 1 searches a document of recognition candidate characters that match a keyword with reference to an index to realize a high-speed and high-accuracy full-text search. The document matches the search condition and is searched for by matching the shape features.
JP 2001-175661 A

  However, since the search device of Patent Document 1 performs a search by character recognition, the processing load for character recognition is large. That is, since the document image is subjected to character recognition by OCR (Optical Character Recognition) and registered as a search index, there is a problem that it takes time to generate the index and the search processing load is increased.

  The present invention has been made in view of the circumstances as described above. The document image search apparatus and the function of the apparatus that can reduce the index generation time for document image search and greatly reduce the search processing load. It aims at providing the program which realizes.

  In order to solve the above problems, the first technical means of the present invention includes: an input unit for inputting document image data to be stored as a search target; and search document data serving as a search source for searching the document image data; Character number measurement unit that recognizes punctuation marks from document image data input by the input unit or search document data and measures the number of characters between punctuation marks, and a registration unit that registers the number of characters of document image data measured by the character number measurement unit as an index And a search unit that searches for an index having the same number of characters as the number of characters between the punctuation marks of the search document data measured by the character number measurement unit.

  According to a second technical means, in the first technical means, the character number measurement unit measures the character size in addition to the number of characters between punctuation marks, and the registration unit registers the character size in addition to the number of characters as an index. It is characterized by.

  The third technical means is that in the first or second technical means, when the character number measurement unit has a part in the search document data where there is no punctuation and a line break in a place that is not the final position of the line of the document, Characteristic is that the number of characters is measured by regarding the line break as a punctuation mark.

  According to a fourth technical means, in any one of the first to third technical means, a text area composed of a page unit or a plurality of area units included in one page is extracted from the document image data and the document of the search document data. A character number measuring unit that measures the number of characters between punctuation marks for each text region extracted by the region extracting unit, and the registration unit indexes the number of characters in the document image in association with the position information of the text region. It is characterized by registering as.

  According to a fifth technical means, in the fourth technical means, the character number measuring unit integrates a plurality of text areas including one continuous sentence when one continuous sentence is divided into a plurality of text areas. In this case, the number of characters is measured using the integrated text area as one search unit.

  According to a sixth technical means, in any one of the first to fifth technical means, the search unit includes an array of the number of characters that at least partially matches the character number array of the search document data measured by the character number measurement unit. It is characterized by searching an index having.

  According to a seventh technical means, in any one of the first to sixth technical means, the registration unit weights the punctuation marks and / or reading marks included in the document of the document image when registering the number of characters as an index. The registration and search unit is characterized by performing a search using the weights of the punctuation marks and / or punctuation marks in addition to the number of characters between the punctuation marks extracted from the search document.

  According to an eighth technical means, in the seventh technical means, the registration unit weights the line break point as a line break point when there is a line break at a location other than the final position of the document line. The registration and search unit is characterized in that a search is performed using weighting of line breaks in addition to weighting of punctuation marks and reading marks.

  The ninth technical means is a document image search program that realizes the function of the document image search apparatus that is one of the first to eighth technical means.

  According to the present invention, the number of characters between punctuation marks is measured and registered as an index, and by searching for an index having the same number of characters as the search document data, the index generation time is shortened compared to performing character recognition processing, The search processing burden can be greatly reduced.

  That is, according to the present invention, since the number of characters between punctuation marks is measured and registered as an index, the index can be generated in a shorter time than the character recognition processing by OCR, and the same number of characters as the search document data of the search source can be obtained. By searching the index, the search time is shortened, and as a result, the search processing load can be greatly reduced.

Further, according to the present invention, by including the character size in addition to the number of characters between punctuation marks in the search condition, it becomes possible to perform a search with higher accuracy than a search using only the number of characters.
Furthermore, according to the present invention, it is possible to efficiently search for document image data by searching in a text area extracted from document image data to be searched.

Furthermore, according to the present invention, when one continuous sentence is divided into a plurality of text areas, the plurality of text areas including the one continuous sentence are integrated to measure the number of characters as the same measurement target area. By performing, for example, when searching based on a modified document or when searching based on a digitized document, it is possible to search with high accuracy.
Furthermore, according to the present invention, a part of a document is added or deleted by version upgrade or the like by searching an index having an array of character numbers that at least partially matches the measured character array of search documents. Can search efficiently.
Furthermore, according to the present invention, it is possible to improve the search accuracy by weighting each of a punctuation mark, a punctuation mark, and a line feed.

FIG. 1 is a block diagram for explaining a position embodiment of a document image search apparatus according to the present invention.
The document search apparatus according to the present embodiment is realized by an image filing server 10 that functions as a server that searches image data by filed image data. A scanner 20 is connected to the image filing server 10.
The scanner 20 includes a scanning unit 21 that reads a document image, a control unit 22 that controls each unit of the scanner, and a communication I / F 23 that communicates with the image filing server 10. Then, the document image data read by the scanning unit 21 is transmitted from the communication I / F 23 to the image filing server 10.

  The image filing server 10 includes a communication I / F 14 for communicating with the scanner 20, a storage unit 12 that stores document image data read by the scanner 20, and an index used for searching the document image data, a search result, and the like. The display / input unit 13 that allows the user to input operations using a touch panel or the like, and the control unit 11 that controls the entire elements of the image filing server 10 are provided. The communication I / F 14 is configured to be able to communicate with an external device such as a PC (not shown). When the control unit 11 receives the document image data read by the scanner 20 from the communication I / F 14, the control unit 11 registers the document image data in the storage unit 12.

  The control unit 11 reads a program that realizes the functions of the present invention from a memory such as a ROM or an HDD (not shown), and executes the program using a working area such as a RAM. As a result, functions such as a character number measurement unit, a region extraction unit, and a search unit according to the present invention are realized. The function of the registration unit of the present invention is realized by storing the text image data and the index in the storage unit 12 under the control of the control unit 11.

With these functions, the control unit 11 can execute index creation processing 11b for creating an index for the received text image data.
In the index creation process 11b, first, the text area of the document image data transmitted from the scanner 20 is extracted. Here, the text area included in the document image data is discriminated, and the area is cut out and extracted. The text area is composed of page units or area units included in one page.

Then, the extracted text area is divided into punctuation marks of the document. Here, punctuation marks are discriminated from the document data included in the text area, and processing for dividing the punctuation marks into units between punctuation marks is performed. In this case, even when there is no punctuation at a position other than the final position of each line of the document, a line break may be regarded as having a punctuation mark at the line break position.
Then, the number of characters of the document is measured for each divided unit, and the measured number of characters is registered in the storage unit 12 as an index. Here, the document image data and the index are registered in the storage unit 12 in association with each other.

  When extracting the text area, the position of the text area may be set manually by the user, or automatic extraction may be performed by the control unit 11. In the case of automatic extraction, it is possible to determine that the text is written from a specific image pattern in the matrix direction. Alternatively, it is determined that the area represented by gradation is a drawing area for a photo or graphic and is removed from the text area, or a line drawing or curve is recognized from the image to be a graphic drawing area. It may be determined and removed from the text area.

When searching for document image data, a process of searching the storage unit 12 for document image data corresponding to the search image data serving as a search source is performed.
In this case, when receiving the search source image data from the scanner 20 or an external device such as a PC (not shown) or a recording medium, the image filing server 10 measures the number of characters between the punctuation marks of the search image data (11c). An index corresponding to the measured number of characters is detected (11d), and document image data corresponding to the detected index is extracted from the storage unit 12 (11e). The extracted document image data information is displayed on the display / input unit 13 by the display control function 11a.
At this time, if there are a plurality of extracted document image data, all of them are extracted, the information is displayed on the display / input unit 13, and the extracted document image is displayed in accordance with user operation instructions and default condition settings. The data is subjected to OCR processing and a detailed search using text data is performed.

FIG. 2 is a diagram showing an image when a user registers document image data in the image filing server according to the embodiment of the present invention.
In this example, it is assumed that the scanner 20 is incorporated in a multifunction machine 30 having a copying function. The MFP 30 can access the image filing server 10 on the network by connecting to the network.

  The user uses the scanner function of the multi-function device 30 to read a document to be registered in the image filing server 10. The multifunction device 30 transmits the document image data read by the scanner function to the image filing server 10 via the network, and registers it in the image filing server 10. In addition, the user can transmit document image data to the image filing server 10 from a network scanner device connected to the network or an information processing device such as a PC. In this case, user authentication processing is performed as necessary. The image filing server 10 accumulates document image data transmitted from the user, and stores and stores it.

FIG. 3 is a diagram for explaining text processing of document image data in the present invention.
In the image filing server 10, the document image data transmitted when the document image data to be registered in the image filing server 10 is transmitted from an external scanner (or a copier or a multifunction machine incorporating the scanner) 20. Is registered and stored in its own storage unit 12 and text processing of the registered document image data is performed.

In the text processing, the control unit 11 of the image filing server 10 first cuts out and extracts a text area in the document image data. FIG. 3A is a diagram illustrating an example of a text area of document image data.
Then, the extracted text area document is divided into units between punctuation marks (FIG. 3B). As described above, here, when there is a part where there is no punctuation and there is a line break in a place other than the final position of the line of the document, the line break may be regarded as a punctuation mark.

  The number of text characters is measured for each division unit. FIG. 3C is a diagram showing the measured number of characters, and the number of characters between a punctuation point, a punctuation point, and a line break point (a line break point without a punctuation mark but not the final position of the line) is measured. As shown in FIG. 3D, the measured number of characters is registered in the storage unit 12 of the image filing server 10 in association with the document image data.

  In the above text processing, not only the measurement of the number of characters in the division unit, but also the text considering the text size, the text direction such as vertical / horizontal writing, and the text divided into a plurality of text areas according to the embodiment of the present invention. Information that weights blocks, punctuation marks, and line breaks is measured and registered in addition to the character count information.

Through the text processing as described above, document image data and an index associated with the document image data are accumulated in the image filing server 10.
When the user actually searches the stored document image data, the user inputs search document data as a search source to the image filing server 10 using a scanner, an external PC, or the like. The search document data can be image data read from the scanner 20 or the like, but may be text data created by a word processor or binary data corresponding to an application.
The image filing server 10 executes text processing similar to the document image data to be registered on the input search document data.

Then, the document image data having the same arrangement is searched for the arrangement of the number of characters for each division unit obtained by the text processing. Further, document image data having a partially matching sequence in addition to the same number of characters may be searched as similar document image data.
In addition to the above arrangement of the number of characters, the search uses text size, text direction such as vertical / horizontal writing, text blocks considering text divided into multiple text areas, information weighted with punctuation marks and line breaks, etc. Is done.

  As a result of the search, if the document image data corresponding to the predetermined search condition as described above can be extracted, the document image data is displayed. If the search fails, the fact is displayed. When a plurality of text image data is extracted, the search target can be determined by performing OCR processing or the like.

4 and 5 are diagrams showing examples of document image data having a plurality of text areas in one page. Here, for example, consider the case of registering document image data 100 as shown in FIG.
By performing the text processing as described above on the sentence image data 100 and extracting the text areas, a plurality of (here, six) text areas R1 to R1 as shown in FIG. Assume that R6 has been extracted.

In the text processing, the above-described character count measurement or the like is performed on each extracted text region R1 to R6. FIG. 4B shows the measurement result of the number of characters in each text region R1 to R6.
Also in this example, in the same way as in the above example, each text region R1 to R6 is divided into a plurality of text regions in addition to the above arrangement of the number of characters, a character size, a text direction such as vertical writing / horizontal writing, and the like. Sentence blocks that take into account sentences, information that weights punctuation marks and line breaks, and the like are appropriately measured.

The measurement information measured in the text processing is stored as an index in the storage unit 12 of the image filing server 10. At this time, the measurement information is linked to the position information in the pages of the text areas R1 to R6. Remembered.
FIG. 5 is a diagram showing position information of each text area, and the position information is represented by coordinate information of four vertices of each text area R1 to R6 in the page. The coordinate information may be information indicating the pixel position of the bitmap, for example, or may be position information in an arbitrarily defined coordinate system.

  Then, when the user searches the document image data, the search image data of the search source is subjected to text processing in the same manner as described above, thereby acquiring the text area of the search image data and its position information. Here, with respect to the text area of the document image data at the same position as the text area of the search image data, the arrangement of the number of measured characters is compared, and the document image data with the same arrangement of the measured character number is extracted and displayed. In this case, it may be determined by comparing the arrangement of the number of measured characters for one area among the plurality of text areas, or the arrangement may be compared using the number of measured characters of all or a plurality of text areas. Good.

FIG. 6 is a diagram for explaining an example of document image data search processing, and shows processing for searching an index using the measured number of characters and character size in a text area.
When searching the document image data to be searched using the search document data of the search source, the search is performed using the measured number of characters in the text area as described above, but in addition to this, the text area The character size of the document inside can be used.

For example, as shown in FIG. 6A, it is assumed that documents having different character sizes are mixed in one text region R.
In this case, when the image filing server 10 performs text processing on the text region R, in addition to the measured number of characters between punctuation marks, the character size is associated and stored as an index (FIG. 6B). ). The character size is size information generated according to a predetermined parameter.

When searching the stored document image data, the search accuracy can be improved by comparing the character size with the number of characters measured in the text region R.
Further, for example, when searching for document information data, the search efficiency can be improved by narrowing down the search target using the character size and further using the measured number of characters. For example, it is possible to perform a search by a method of performing a search under the condition of a headline sentence having a character size of 1.5 or more, and performing a search based on the number of measured characters if there are a plurality of corresponding document image data. The headline sentence can be determined based on conditions such as a character string at the beginning of the sentence and having a character size larger than other characters, or a character string in which the characters are bold.

FIG. 7 is a diagram for explaining another example of a document image search process, and shows a process when searching for an index using the number of characters measured in a text area and sentence direction information.
When searching the document image data to be searched using the search document data of the search source, in addition to the number of measured characters in the text area, the text direction (vertical writing / horizontal writing, etc.) in the text area is used as a search condition. be able to.

For example, as shown in FIG. 7A, it is assumed that text regions R1 and R2 in which documents having different sentence directions are described are mixed in the document image data 100.
Here, when the image filing server performs text processing, in addition to the number of characters measured between punctuation marks, the text direction (vertical writing / horizontal writing, etc.) is associated and stored as an index (FIG. 7B).

  The text direction (vertical writing / horizontal writing) can be discriminated from the pattern of image data that differs depending on whether vertical writing or horizontal writing. Alternatively, OCR may be applied to a part of the document, and morphological analysis may be performed on the obtained character data to determine the direction in which a meaningful character string can be extracted as the sentence direction.

When searching the stored document image data, the search accuracy can be improved by comparing the text direction with the number of characters measured in the text area.
For example, when searching for document image data, the search efficiency can be improved by narrowing down the search target using the text direction and further searching using the measured number of characters.

FIGS. 8 and 9 are diagrams for explaining still another example of document image data search processing, and show processing when searching for an index using the number of characters measured in a text area and search unit information. .
When searching the search target document image data using the search source document data, the search unit (sentence block) recognized as the text area is used as the search condition in addition to the number of measured characters in the text area. May be.

  For example, as shown in FIG. 8, it is assumed that one sentence is divided into two text regions R1 and R2. Here, sentences continue from the end of the first page to the beginning of the next page. In this case, in the above text processing, a text area is extracted for each page, so that one sentence is divided into different text areas R1 and R2.

In the present embodiment, the control unit 11 of the image filing server 10 determines whether or not there is a punctuation mark at the final position of the text for each of the extracted text regions R1 and R2.
For example, the final position of the text area R1 is confirmed, and if there is no punctuation mark at the final position, it is determined that the sentence in the text area R1 continues to the next text area R2. Then, processing is performed in which these text regions R1 and R2 are integrated as a sentence block. A sentence block is one search unit.

In such a case, when the image filing server performs text processing, in addition to the information on the number of characters between punctuation marks, information on the text block as a search unit is associated and stored as an index (FIG. 9).
In the example of FIG. 9, “document” is set as a search unit indicating a normal text area cut out by text processing, and a search unit which is a text block formed by integrating a plurality of text areas is set as “text block”. ing.

  When searching document image data, for example, when using search document data in which the document is not modified at all with respect to the document image data, it is possible to search in the text area in page units as described above. When a search is performed using modified search document data, a search using a block unit is effective. Alternatively, when a search is performed using search document data such as digitized text data, the text area cannot be extracted, and can be handled by searching in block units.

In addition, when registering the number of measured characters of document image data as an index, the search unit is registered only in the “document” unit, and when the actual document search is performed, it is placed at the final position of the text area to be searched. When there is no punctuation mark, it is also possible to perform a search using a text block as a search target by performing a blocking process with the next text area.
In the above example, the two text areas are integrated into a sentence block, but the sentence block may be an integration of three or more text areas.

10 to 12 are diagrams for explaining still another example of document image data search processing. Document image data is registered by weighting punctuation marks and line breaks, and search can be performed using these weightings. An example of such processing is shown.
Here, for example, it is assumed that there is document image data including a text region R as shown in FIG. The result of performing text processing on the document image data and measuring the number of characters in the extracted text region R is as shown in FIG.

FIG. 11 is a graph showing the measured number of characters in the text area. FIG. 11A is a graph in which the number of characters between punctuation marks measured from the document image data as shown in FIG. 10 is plotted in order. This plot waveform is called a character number information waveform.
The number-of-characters information waveform in FIG. 11 (A) is simply a plot of the number of measured characters between the measured punctuation marks. Here, it is understood that the punctuation marks, punctuation marks, and line breaks included in the text region R are all treated equally. .

On the other hand, FIG. 11B shows character number waveform information weighted by regarding the number of characters at a punctuation mark as “0” in order to provide sentence-by-sentence information. Since the punctuation mark is a feature of the sentence, the accuracy of the search can be improved by weighting and registering it as search information.
In other words, in this example, when text processing is performed on document image data to be registered as a search target, the text area is extracted and the number of characters for each text area is measured. The information is registered as an index together with the number of measured characters.

  When searching for document image data, text processing is similarly performed on the search document data serving as a search source, the number of characters for each text area is measured, and similarly, the punctuation points are weighted. Then, the retrieval of the document image data to be retrieved is executed using the measured number of characters and the weight information of the punctuation marks.

  In FIG. 12, in addition to the above-mentioned weighting of the punctuation points, information in units of clauses is provided. Therefore, for a place where there is a line break, the number of characters for the line break is weighted as two “0” s, and the information is the number of measured characters. Register with. Here, when there is a line feed after a punctuation point, the punctuation point is omitted and weighting is performed so that two “0” s continue. At this time, when the text ends at a portion other than the last position of the line and a line break is made on the next line, it is determined to be a “line feed”.

  In this case, since the line feed and the punctuation are weighted, the search accuracy can be further improved. Furthermore, in addition to the above-described weighting of the punctuation marks, or in addition to the weighting of the punctuation marks and the line feed point, the punctuation marks may be weighted for retrieval.

  However, when document data such as digitized text data is used as search document data of the search source, or when the document is altered due to layout change or editing of the search target document image data, the reverse is true. There is a case where the search accuracy is higher when they are treated equally without giving weights to punctuation marks and line break points. Therefore, it is preferable that the search method can be arbitrarily switched when searching.

FIGS. 13 and 14 are diagrams for explaining an example of processing for searching for document image data similar to search document data.
Here, as in the example of FIG. 11, it is assumed that the character number information waveform as shown in FIG. 13A is obtained by text processing of the document image data to be searched.

When document image data having such a character number information waveform is to be searched, the character number information waveform has the same shape unless the search document data as the search source is altered from the document image data to be searched. Therefore, at the time of search, the search can be performed only with the short number of measured characters corresponding to a part of the character number information waveform.
Here, as shown in FIG. 13B, when a search is performed with search document data that has not been altered at all, the number of characters measured in a part of the character count information waveform (for example, the block D within the dotted line) when performing the search. Just use. This is because the search document data and the search target document image data always match because no modification has been made.

  However, when the search document data as the search source is modified from the search target document image data, the number of characters is likely to change between them. In that case, it is necessary to compare the similarity of the whole document without being caught by a partial difference in the number of measured characters. Therefore, in this case, it is necessary to search using the arrangement information of a relatively long number of measured characters.

For example, as shown in FIG. 14, in the character number information waveform of the entire text area, blocks E and F are unmodified portions, and block G is a modified portion. In such a case, as the number of measurement characters to be searched, an array of the number of measurement characters corresponding to the waveform in the area from the blocks E to G to F is used.
At this time, only the array of the measured number of characters of the block G does not match the array of the measured number of characters of the target sentence image data, but if the array of at least a part of the measured characters (here, the blocks E and F) matches. The whole image is determined to be similar, and is extracted as document image data to be searched.

FIG. 15 is a flowchart for explaining an example of index creation processing in the document image retrieval apparatus according to the present invention.
First, the document retrieval apparatus receives document image data from the scanner (step S1). Then, the text area of the document image data received from the scanner is extracted (step S2). Here, for example, one page of document image data is set as one text area, or if there are a plurality of text areas in one page, the plurality of areas are cut out and extracted.

Then, it is determined whether or not there are a plurality of extracted text areas (step S3). If there are no plurality of text areas, the number of characters in the text area is measured (step S4). In the character count measurement, a process for measuring the number of characters between punctuation marks of a document in the target text area is performed. The measured number of characters between punctuation marks is registered as an index (step S5).
Here, depending on the embodiment of the character count measurement process, along with the number of characters between punctuation marks, character size, sentence direction such as vertical / horizontal writing, sentence block considering sentences divided into multiple text areas, punctuation marks and line breaks Measures weighted information and registers it in addition to the number of characters measured.

If it is determined in step S3 that there are a plurality of text areas, the first text area is selected (step S6), and the number of characters in the selected text area is measured (step S7). This process is the same as the measurement process in step S4.
When the character measurement process for the selected text area is completed, it is determined whether there is another text area that has not been measured (step S8). If there is a text area that has not been measured, It selects as a text area (step S10), and progresses to the character number measurement process of step S7.

  If there is no other text area that has not been measured in step S8, the character count measurement processing for all text areas has been completed. Therefore, the position information of each text area and the number of characters between punctuation marks in each text area are used as indexes. Register (step S9). Also in this case, as in step S5, the text size, the text direction such as vertical writing / horizontal writing (the progress direction of the text string of the text), the text block considering the text divided into a plurality of text areas, according to the embodiment, Register information including weighted punctuation marks and line breaks in addition to the number of measured characters.

FIG. 16 is a flowchart for explaining an example of the character region count process in the text region. In this example, a processing example including vertical / horizontal writing determination of a document in a text area, measurement of the number of characters in units of division between punctuation marks, and character size measurement processing will be described.
First, it is determined whether the document in the target text area is vertically written or horizontally written (step S11). Here, as described above, whether the vertical writing or the horizontal writing may be determined from the pattern of the document image, or a part of the document image data is subjected to the OCR process and then the morphological analysis is performed to determine whether the writing is the vertical writing or the horizontal writing. You may do it.

Then, the document in the text area is divided into punctuation marks (step S12), and the number of characters in the division unit between punctuation marks is measured (step S13). In this case, the line break point may be included in the dividing point for processing.
Further, it is determined whether or not the character size of the document is constant within the text area (step S14). If the character size is constant, the character count measurement process is terminated. If the character size is not constant, the character size is measured for each division unit (step S15), and the process is terminated.
In this example, an example is shown in which the character size is measured. However, processing that does not measure the character size as described above may be used.

FIG. 17 is a flowchart for explaining an example of division processing when one sentence is described separately in a plurality of text areas.
First, processing for dividing the document image data in the text area into punctuation marks according to the determined text direction (vertical writing / horizontal writing) is started (step S21). Then, it is determined whether or not the final position of the document in the text area is a punctuation mark (step S22).

  If the final position is not a punctuation point, it is further determined whether or not there is a text area that has not been subjected to the character count measurement process in the same page (step S23). If there is a text area that has not been subjected to character count processing on the same page, the text area whose final position is not a punctuation mark and the next text area are integrated into the same text area to form a sentence block. The block document image data is divided into punctuation marks (step S24).

  On the other hand, if the final position of the text area is a punctuation mark in step S22, the division process ends. If there is no text area that has not been subjected to character count measurement processing in the same page in step S23, the text area whose final position is not a punctuation mark and the text area of the document image data on the next page are made the same text area. The text blocks are integrated to divide the document image data of the text blocks into punctuation marks (step S25). In this case, if there are a plurality of text areas on the next page, a text block is set using the first text area.

FIG. 18 is a flowchart for explaining an example of a document search process in the document search apparatus according to the present invention.
When searching for document image data, first, search document data as a search source is input (step S31). In the document search apparatus, a process for extracting a text area of the input search document data is performed (step S32). Then, it is determined whether or not there are a plurality of extracted text areas (step S33). If there are not a plurality of text areas, a process for measuring the number of characters of search sentence data in the text area is performed (step S34).
The character count measurement process is a process for measuring the number of characters between punctuation marks as described above, and further considers text size, text direction such as vertical / horizontal writing, and text divided into multiple text areas according to the embodiment. Measurement processing such as information weighted with blocks, punctuation marks and line breaks is performed.

Then, using the measurement result of the character count measurement process, a process for searching from a pre-registered index is performed (step S35).
As a result of the search, it is determined whether or not there are a plurality of extracted document image data (step S36). If there is only one extracted document image data, the extracted sentence image data is displayed as a search result ( Step S42). If there are a plurality of extracted document image data, OCR processing is performed on all the extracted document image data, and the document image data corresponding to the search document data is selected and displayed as a search result (step S37).

  On the other hand, if there are a plurality of cut-out text areas in step S33, the first text area is first selected (step S38), and the number of characters in the selected text area is measured (step S39). In this case, the same processing as the character count measurement processing in step S34 is performed.

  When the character count measurement process for the selected text area is completed, it is further determined whether there is another text area that has not been measured (step S40). If there is another text area that has not been measured here, the next text area that has not been measured is selected (step S41), and the character count measurement process is executed for the selected text area (step S39).

On the other hand, if there is no other text area that has not been measured in step S40, the character count measurement process for all text areas has been completed, and therefore, a search is made from a pre-registered index using the measurement result of the character count measurement process. Is performed (step S35).
In this case, it is also possible to perform the character count measurement process in only one text area among the plurality of text areas in the search document data, and perform an index search using the result.

FIG. 19 is a flowchart for further explaining an example of search processing from an index in search processing.
In the search processing from the index, it is first determined whether or not the same document is specified to be searched (step S51). Here, for example, a mode in which only completely identical text image data is searched and extracted without searching text image data partially modified by editing or the like is partially the same in the text area and similar. If it is possible to select the mode for extracting the text image data being performed, it is determined whether any mode is designated. In the mode for searching for similar document image data, it is possible to search for document image data that has been modified by adding a version upgrade or editing to the original document.

  If it is designated to extract the same document image data, an index having the measured number of characters in the text area of the search document data is searched (step S52). If it is specified to extract similar document image data instead of the same document image data, an index having a measured number of characters similar to the search document data is searched (step S54). Here, when the arrangement of the number of measured characters between the punctuation marks matches in at least a part of the text area, it can be determined that they are similar.

  In the index search process in step S52, not only the search based on the number of measured characters, but also the text size (vertical / horizontal writing) and other text directions (progression direction of the text string) and a plurality of text areas according to the embodiment. It is possible to use sentence blocks that take into account separated sentences, information that weights punctuation marks and line breaks, and the like.

  If an index is detected in step S52 or step S54, document image data registered corresponding to the detected index is extracted (step S53).

  A program according to the present invention is a program for realizing the function of the document image search apparatus. The program is stored in a memory such as a ROM or HDD provided in the document image search device, and the functions of the document image search device described in the above embodiments are realized by a control unit such as a CPU reading and executing the program. can do. Further, by executing the program recorded in the memory, not only the functions of the above-described embodiments are realized, but also by cooperating with the operating system or other application programs based on the instructions of the program, The functions of the invention may be realized.

  The program can be recorded on a recording medium and distributed. Recording media include semiconductor media (eg, ROM, nonvolatile memory card, etc.), optical recording media (eg, DVD, MO, MD, CD, BD, etc.), magnetic recording media (eg, magnetic tape, flexible disk, etc.) Etc. are applicable. In addition, when distributing to the market, it can be held in a server computer connected via a network such as the Internet and transferred to a document image search apparatus.

It is a block diagram for demonstrating the position embodiment of the document image search device by this invention. It is a figure which shows an image when a user registers document image data with respect to the image filing server which is embodiment of this invention. It is a figure for demonstrating the text processing of the document image data in this invention. It is a figure which shows the example of the document image data which has a some text area in one page. It is another figure which shows the example of the document image data which has a some text area in one page. It is a figure for demonstrating the example of a search process of document image data. It is a figure for demonstrating the other example of the search process of a document image. FIG. 10 is a diagram for explaining still another example of document image data search processing. It is another figure for demonstrating the further another example of the search process of document image data. FIG. 10 is a diagram for explaining still another example of document image data search processing. It is another figure for demonstrating the further another example of the search process of document image data. It is another figure for demonstrating the further another example of the search process of document image data. It is a figure for demonstrating the process example which searches the document image data similar to search document data. It is another figure for demonstrating the process example which searches the document image data similar to search document data. It is a flowchart for demonstrating an example of the creation process of the index in the document image search device by this invention. It is a flowchart for demonstrating an example of the character number measurement process of a text area. It is a flowchart for demonstrating the example of a division | segmentation process when one sentence is divided and described in several text area | regions. It is a flowchart for demonstrating an example of the document search process in the document search apparatus by this invention. It is a flowchart for demonstrating further the example of a search process from the index in a search process.

Explanation of symbols

R1 to R6 ... text area, 10 ... image filing server, 11 ... control unit, 12 ... storage unit, 13 ... display / input unit, 14, 23 ... communication I / F, 20 ... scanner, 21 ... scanning unit, 22 ... Control unit, 30 ... MFP, 100 ... sentence image data.

Claims (9)

An input unit for inputting document image data to be stored as a search target, and search document data serving as a search source for searching the document image data;
A character number measuring unit that recognizes punctuation marks from the document image data input by the input unit or the search document data, and measures the number of characters between punctuation marks;
A registration unit that registers the number of characters of the document image data measured by the character number measurement unit as an index;
A document image retrieval apparatus comprising: a retrieval unit that retrieves an index having the same number of characters as the number of characters between punctuation marks of the retrieved document data measured by the character number measurement unit.   The character count measuring unit measures a character size in addition to the number of characters between punctuation marks, and the registration unit registers the index including the character size in addition to the number of characters. Document image search device.   The character number measurement unit, when there is a part in the search document data where there is no punctuation at a place that is not the final position of the document line, the line break is regarded as a punctuation mark and the number of characters is measured. The document image retrieval apparatus according to claim 1, wherein: An area extracting unit that extracts a text area composed of a plurality of area units included in a page unit or one page from the document image data and the document of the search document data;
The character number measurement unit measures the number of characters between the punctuation marks for each text region extracted by the region extraction unit,
4. The document image search apparatus according to claim 1, wherein the registration unit registers the number of characters of the document image as an index in association with position information of the text area.   When the continuous text is divided into a plurality of text areas, the number-of-characters measurement unit integrates a plurality of text areas including the continuous text, and the integrated text area is used as one search unit. 5. The document image retrieval apparatus according to claim 4, wherein the number of characters is measured.   The search unit according to any one of claims 1 to 5, wherein the search unit searches for an index having an array of at least partially matching characters from the array of the number of characters of the search document data measured by the character number measurement unit. The document image search device according to any one of the above. The registration unit, when registering the number of characters as an index, weights and registers the punctuation marks and / or reading marks included in the document of the document image,
7. The search unit according to claim 1, wherein the search unit performs a search using a weight of the punctuation marks and / or punctuation marks in addition to the number of characters between the punctuation marks extracted from a search document. Document image search device. The registration unit, when there is a part that is line-breaked at a place that is not the final position of the line of the document, the weighted point as a line-break point is registered and registered,
The document image search apparatus according to claim 7, wherein the search unit performs a search using the weight of the line break point in addition to the weight of the punctuation mark and the reading point.   A document image search program for realizing the function of the document image search device according to claim 1.

Images