JP2006011966A - Character recognition device and character recognition program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device performing OCR processing at very high speed with high precision on a document including mixed characters varied in size and shape. <P>SOLUTION: This character recognition device is provided with a storage means 320 storing hierarchical code data acquired by applying hierarchical coding processing onto raster data of a document, a decoding means 340 performing decoding processing with a first bit rate on the whole or a part of a code string included by the hierarchical code data, and a recognition means 350 recognizing a character from an image obtained by decoding processing. When recognition accuracy by the recognition means 350 is lowered below a predetermined value, a bit rate in decoding processing is switched from the first bit rate to a second rate higher than the first bit rate. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a technique for acquiring text data from an image of a document on which characters are written.

A process for acquiring text data of characters written in a document from raster data of the document is generally called an OCR (Optical Character Recognition) process. In the OCR process, each image in which characters are drawn is cut out from raster data obtained by scanning the document, and each character in the document is recognized by comparing the cut-out image with a pattern prepared in advance as a dictionary. is doing. For this reason, from the viewpoint of improving the character recognition accuracy, it is preferable that the raster data obtained from the document has the highest possible resolution (see Patent Documents 1 and 2). However, if the resolution of the raster data is increased, the capacity of the memory for expanding the data must be increased, and the burden on the processor that processes the expanded data also increases. That is, from the viewpoint of processing cost, it is not preferable that the raster data has a high resolution.
Conventionally, various techniques have been proposed to satisfy these two conflicting requirements. For example, Patent Documents 3 to 5 propose a character recognition technique in which an original on which characters are written is scanned at a relatively low resolution to try character recognition, and when the recognition rate is low, the resolution is increased and then scanned again. . Patent Document 6 also proposes the same kind of recognition technology. The data processing apparatus disclosed in this document varies the zoom magnification of the scanner according to the character recognition situation. The character recognition device disclosed in Patent Document 7 accepts input of an attribute (for example, character size) of a character written on a document, and can be an image scanned at what resolution based on the input attribute. Determine whether sufficient recognition accuracy is secured. The form classification processing system disclosed in Patent Document 8 performs character recognition based on an image scanned with high resolution for parts with complex characters such as kanji, while scanning other parts with low resolution. Character recognition is performed based on the obtained image.
JP-A-6-237379 JP-A-9-289624 JP 2000-29987 A JP 2000-293633 A JP 2001-103311 A JP-A-6-20089 Japanese Patent Laid-Open No. 2002-24766 Japanese Patent Application Laid-Open No. 8-27284

As described above, various techniques for improving the recognition rate of OCR processing with a circuit scale as narrow as possible have been proposed so far, and in many cases, character recognition is performed using images scanned at a low resolution. And the idea of using a case where character recognition is performed using an image scanned at high resolution. However, in the above-described series of conventional techniques, whether to use a high-resolution image or a low-resolution image is determined for each original, and characters that are difficult to recognize in the document to be scanned (for example, It is not suitable for use in the case where characters (e.g., kanji) and characters that are relatively easy to recognize (for example, kana characters and alphabets) coexist. Note that although Patent Document 8 shows that only a part is scanned at a high resolution and used for character recognition, the form classification processing system disclosed in this document has a place where complicated characters are written. A pre-determined form is assumed as a recognition target, and a general document in which kanji, kana characters, alphabets, and the like are mixed is not assumed as a recognition target.
The present invention has been devised in view of such a background, and provides an apparatus for performing OCR processing on a document in which characters of various sizes and shapes are mixed with extremely high speed and high accuracy. Objective.

  A character recognition device according to a preferred aspect of the present invention includes an input unit for inputting an image obtained by scanning a document in which one or more characters are written, and a hierarchical encoding process for the input image. Encoding means for acquiring hierarchical code data including a code string hierarchized according to the bit rate, storage means for storing the acquired hierarchical code data, and hierarchies stored in the storage means Decoding means for performing decoding processing at a specified bit rate for all or part of a code string included in the target code data, and recognition means for recognizing characters from the image obtained by the decoding processing; Control means for controlling a bit rate in the decoding process based on success or failure of recognition by the recognition means.

  According to another preferred aspect of the present invention, a character recognition device includes an input unit that inputs an image obtained by scanning a document in which one or more characters are written, and a hierarchical encoding process on the input image. And encoding means for acquiring hierarchical code data including a code string hierarchized according to the bit rate, storage means for storing the acquired hierarchical code data, and storing in the storage means Decoding means for performing decoding processing at a first bit rate on all or part of a code string included in the hierarchical code data, and recognition means for recognizing characters from the image obtained by the decoding processing And a control means for switching the bit rate in the decoding process to a second bit rate higher than the first bit rate when the accuracy of recognition by the recognition means falls below a predetermined value.

  According to another preferred aspect of the present invention, a character recognition device includes an input unit that inputs an image obtained by scanning a document in which one or more characters are written, and a hierarchical encoding process on the input image. And encoding means for acquiring hierarchical code data including a code string hierarchized according to the bit rate, storage means for storing the acquired hierarchical code data, and storing in the storage means A decoding means for identifying a code string corresponding to a region of interest for character recognition from the hierarchical code data and performing a decoding process on the identified code string at a first bit rate; and the decoding process Recognizing means for recognizing characters from the image obtained by the above, and a second bit rate higher than the first bit rate when the accuracy of recognition by the recognizing means falls below a predetermined value. In Ri changing and a control unit.

  In this aspect, the encoding means divides the input image into a plurality of drawing areas that do not have overlapping portions, and performs a hierarchical encoding process on each of the divided drawing areas, thereby The decoding means sequentially identifies each of the divided drawing areas as the attention area.

  According to another preferred aspect of the present invention, a character recognition device includes an input unit that inputs an image obtained by scanning a document in which one or more characters are written, and a hierarchical encoding process on the input image. And encoding means for acquiring hierarchical code data including a code string hierarchized according to the bit rate, storage means for storing the acquired hierarchical code data, and storing in the storage means A decoding means for identifying a code string corresponding to a region of interest for character recognition from the hierarchical code data and performing a decoding process on the identified code string at a first bit rate; and the decoding process Recognition means for recognizing characters from the image obtained by the above, area narrowing means for specifying, as the new attention area, an area in which the recognition accuracy by the recognition means falls below a predetermined value among the attention areas, and the new attention Territory And a control means for switching to a higher second bit rate than the bit rate of the first bit rate when performing the decoding means decoding Te.

  In this aspect, the encoding unit divides the input image into a plurality of drawing areas that do not have overlapping portions, and performs hierarchical encoding processing on each of the divided drawing areas, thereby The area specifying means may determine whether or not the character recognition accuracy is an area below a predetermined value for each of the divided drawing areas.

  According to another preferred aspect of the present invention, there is provided a computer program comprising: an input unit that inputs an image obtained by scanning a document in which one or more characters are written; and an information storage unit. A coding function for performing hierarchical coding processing on an input image and acquiring hierarchical code data including a code string hierarchized according to a bit rate; and storing the acquired hierarchical code data in the memory A data storage function stored in the means, a decoding function that performs a decoding process at a specified bit rate on all or part of the code string included in the hierarchical code data stored in the storage means, A recognition function for recognizing characters from an image obtained by the decoding process and a control function for controlling the bit rate in the decoding process based on the success or failure of the recognition by the recognition function are realized.

  According to another aspect of the present invention, there is provided a computer program comprising: an input unit that inputs an image obtained by scanning a document in which one or more characters are written; and an information storage unit. A coding function for performing hierarchical coding processing on an input image and acquiring hierarchical code data including a code string hierarchized according to a bit rate; and storing the acquired hierarchical code data in the memory A data storage function stored in the means, a decoding function that performs a decoding process at a first bit rate on all or part of a code string included in the hierarchical code data stored in the storage means, When the recognition function for recognizing characters from the image obtained by the decoding process and the recognition accuracy by the recognition function fall below a predetermined value, the bit rate in the decoding process is set higher than the first bit rate. To realize a control function to switch to the second bit rate are.

  According to another aspect of the present invention, there is provided a computer program comprising: an input unit that inputs an image obtained by scanning a document in which one or more characters are written; and an information storage unit. A coding function for performing hierarchical coding processing on an input image and acquiring hierarchical code data including a code string hierarchized according to a bit rate; and storing the acquired hierarchical code data in the memory A code string corresponding to a region of interest for character recognition is identified from the data storage function stored in the means and the hierarchical code data stored in the storage means, and the first code rate is applied to the identified code string. A decoding function for performing the decoding process, a recognition function for recognizing characters from the image obtained by the decoding process, and a recognition function by the recognition function when the accuracy of recognition is below a predetermined value, The trait to realize a control function to switch to a higher second bit rate than the first bit rate.

  According to another aspect of the present invention, there is provided a computer program comprising: an input unit that inputs an image obtained by scanning a document in which one or more characters are written; and an information storage unit. A coding function for performing hierarchical coding processing on an input image and acquiring hierarchical code data including a code string hierarchized according to a bit rate; and storing the acquired hierarchical code data in the memory A code string corresponding to a region of interest for character recognition is identified from the data storage function stored in the means and the hierarchical code data stored in the storage means, and the first code rate is applied to the identified code string. A decoding function for performing the decoding process, a recognition function for recognizing characters from the image obtained by the decoding process, and an area of the attention area in which the recognition accuracy by the recognition function falls below a predetermined value. A region narrowing function that identifies the region of interest, and a control function that switches the bit rate when the decoding function performs the decoding process for the new region of interest to a second bit rate that is higher than the first bit rate; make it happen.

  According to the present invention, it is possible to execute OCR processing on a document in which characters of various sizes and shapes are mixed at extremely high speed and with high accuracy while reducing the circuit scale.

(First embodiment)
A character recognition device according to a first embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram showing a schematic hardware configuration of the character recognition apparatus according to the present embodiment. As shown in the figure, this character recognition apparatus is configured by connecting a scanner 100, an operator 200, a controller 300, and a communication interface 400.
The scanner 100 optically scans a document to be scanned to generate gray scale raster data, and supplies the raster data to the controller 300. The operator 200 is responsible for various input operations such as an instruction to start character recognition processing. The controller 300 obtains text data of characters written in the document by performing OCR processing on the raster data. The communication interface 400 transmits the text data acquired by the controller 300 to an external computer device.

FIG. 2 is a block diagram showing an internal configuration of the controller 300. The controller 300 includes an encoding unit 310, a code data storage unit 320, a region of interest specifying unit 330, a decoding unit 340, a character recognition unit 350, and a text data storage unit 360.
The encoding unit 310 performs hierarchical encoding processing on the raster data to obtain hierarchical code data that is code data in a format according to the JPEG (Joint Photographic Experts Group) 2000 encoding standard. That is, the encoding unit 310 includes a DC level shift unit, a tiling unit, a wavelet transform unit, a quantization unit, a coefficient bit modeling unit, a code string generation unit, etc. (not shown), and receives raster data. Each of these means executes a series of processes such as DC level shift, tiling, wavelet transform, quantization, coefficient bit modeling, and code string generation, and acquires hierarchical code data.

Since the behavior of each means incorporated in the encoding means 310 belongs to the category of the prior art, the details are omitted here. Here, the data structure of the hierarchical code data acquired by the encoding means 310 will be described.
FIG. 3 is a data structure diagram of hierarchical code data. As shown in the figure, the code data includes a main header and a plurality of segments called “tiles”. The main header MH includes attribute information such as an image size and an image position offset. Here, the “tile area”, which is a concept unique to the encoding standard JPEG2000, will be described. The “tile area” means a rectangular drawing area of the same size that is obtained by dividing the entire drawing area of the input raster data and has no overlapping portion. In JPEG2000, in various signal processing processes such as wavelet transform and encoding, pixels that cross the boundary of tile areas are not referred to, and each tile area is handled as one independent unit. By eliminating the correlation between the tile areas, memory saving at the time of encoding and decoding is realized. In the hierarchical code data, each segment called a tile corresponds to one tile area.

  As shown in FIG. 3, the segment of the tile T is divided into a tile header TH and tile data TD. The tile header TH includes attribute information such as a tile number assigned to each tile area in raster order. On the other hand, the tile data TH is divided into a plurality of packets P, and each packet P includes a code string embedded for each predetermined bit rate. “Embedding” means that a code sequence encoded at a plurality of bit rates is embedded in a hierarchical manner in a code sequence representing one pixel. For example, the packet P1 in FIG. 3 includes a code string obtained by encoding raster data at the lowest bit rate, and the packet P2 is encoded at a bit rate one step higher than that of the packet 1 code string. A code string that is a difference between the obtained code string and the code string of packet 1, that is, a code string that is truncated when quantized to the encoding of packet 1 is included. Similarly, each packet after the packet P3 includes recursively a difference from a code string obtained by encoding at a higher bit rate than the code string of the immediately preceding packet.

  When the code string is embedded in this way, a proportional relationship is established between the bit rate of the bit to be decoded and the image quality of the restored image. For example, it is assumed that code data including an a-byte code string is generated by performing hierarchical encoding processing on an image in a certain tile area. When this a-byte code string is embedded, an image can be obtained by decoding only b bytes (b <a) and c bytes (c <b), and the obtained image is the original one. This is the same as the image quality when encoding / decoding is performed with the bit rate set to b bytes or c bytes.

Returning to the description of FIG.
The code data storage unit 320 is a buffer that temporarily stores the hierarchical code data supplied from the encoding unit 310.
The attention area specifying means 330 specifies a segment corresponding to one tile area as the attention area from the hierarchical code data in the code data storage means 320, extracts the code string of the specified segment, and outputs it to the decoding means 340. Supply. This attention area specifying means 330 has an image quality level counter (not shown), and increases the value of this counter by one each time character recognition fails in the same area. When extracting the code string, the code string to be extracted is selected according to the value of the image quality level counter. For example, if the numerical value is “1” as a result of the attention area specifying unit 330 referring to its image quality level counter, only the code string included in the packet of the highest layer is extracted and supplied, If the value of the image quality level counter is “2”, the code string included in the packet of the second layer from the top is extracted, and the code string obtained by concatenating them is supplied. That is, the attention area specifying unit 330 also functions as a control unit that switches the bit rate of the decoding process performed by the decoding unit 340.

  The decoding unit 340 performs a decoding process on the code string extracted from the hierarchical code data by the attention area specifying unit 330 and restores the image of the attention area as a pixel information group. Pixel information means information including the address and pixel value of a pixel for drawing an image. That is, the decoding unit 340 includes an arithmetic decoding unit, a coefficient bit modeling unit, an inverse quantization unit, an inverse wavelet transform unit, a DC level shift unit, and the like (not shown). Each of these means executes a series of processes such as arithmetic decoding, coefficient bit modeling, inverse quantization, inverse wavelet transform, and DC level shift, and acquires a pixel information group of the region of interest.

The character recognizing means 350 recognizes a character from the pixel information group of the attention area, and obtains and outputs the accuracy of character recognition based on the success or failure of the recognition. As will be described in detail later in the description of the operation, when the recognition accuracy falls below a predetermined value, the pixel information group decoded by the decoding unit 340 at a higher bit rate is converted into the character recognition unit 350. Supplied again.
The text data storage unit 360 is a buffer that temporarily stores text data of characters recognized by the character recognition unit 350.

Next, a character recognition process that is a characteristic operation of the present embodiment will be described.
FIG. 4 is a flowchart showing the character recognition process.
This process is started when the user places a document on a document placement table (not shown) and instructs the operator 200 to start the character recognition process.
When the start of the character recognition process is instructed, the scanner 100 optically scans the document to generate grayscale raster data, and supplies it to the controller 300 (S10). The supplied raster data is input to the encoding unit 310 of the controller 300.

The encoding unit 310 of the controller 300 performs hierarchical encoding processing on the input raster data to obtain hierarchical code data (S11). As described above, the hierarchical code data includes a code string embedded in each tile area.
Next, the attention area specifying means 330 specifies one of the tile areas obtained by dividing the raster data during the hierarchical encoding process as the attention area (S12).

The attention area specifying means 330 refers to each tile header included in the hierarchical code data, and specifies tile data corresponding to the tile area specified in step 12 (S13).
The attention area specifying unit 330 extracts the code string of the hierarchy corresponding to the image quality level indicated by the value of its image quality level counter from the tile data specified in step 13, and outputs it to the character recognition unit 350 (S14). For example, if the value of the image quality level counter is “1”, only the character string included in the packet of the highest layer in the tile data is extracted, and the value of the image quality level counter is “2”. For example, the character string included in the packet of the second layer in the tile data is extracted, and the character string obtained by connecting them is supplied.

The decoding unit 340 performs a decoding process on the character string input from the attention area specifying unit 330 to acquire a pixel information group for drawing an image of the tile area, and the acquired pixel information group is used as the character recognition unit 350. (S15).
The character recognizing unit 350 attempts to perform character recognition based on the pixel information group input from the decoding unit 340 (S16). Specifically, an image in which each character is drawn is cut out from a bitmap drawn by the pixel information group, and a feature amount extracted from the cut out image and a character prepared in advance as a dictionary (hereinafter referred to as “reference character”) Measure the logical distance from the feature quantity. Then, as a result of the measurement, it is determined that the reference character having the smallest feature distance is drawn.

  The character recognition unit 350 determines whether the recognition accuracy in step 16 is below a predetermined value (S17). The recognition accuracy in this step corresponds to the distance of the feature amount. That is, if the distance of the feature amount from the standard character determined to be drawn is smaller than the predetermined value, the determination result of this step is “NO”, and if the distance is larger than the predetermined value, the determination result of this step is “YES”. " When the attention area includes an image of a plurality of characters, the determination result of this step is “NO” if the average value of the recognition accuracy obtained for each character is smaller than a predetermined value, and “ YES ”.

  When the determination result in step 17 is “YES”, the character recognition unit 350 supplies a signal indicating that the character recognition has failed to the attention area specifying unit 330 (S18). The attention area specifying means 330 that has acquired the signal adds “1” to the value of its image quality level counter (S19). Then, returning to step 14, the attention area specifying means 330 extracts a code string corresponding to the new image quality level, and the subsequent processing from step 15 is sequentially executed.

When the determination result in step 17 is “NO”, the character recognition unit 350 outputs the text data obtained as a result of the recognition in step 16 to the text data storage unit 360 (S20). The text data storage unit 360 sequentially stores the output text data. Subsequently, the character recognition unit 350 supplies a signal indicating that the character recognition is successful to the attention area specifying unit 330 (S21). The attention area specifying means 330 that has acquired the signal resets the value of its image quality level counter to “0” (S22). Thereafter, returning to step 12, the attention area specifying means 330 specifies another tile area as the attention area, and the processing after step 13 is repeated with this new attention area as the processing target.
When the above-described series of processing has been executed with all tile regions as the attention region, the text data stored in the text data storage unit 360 is read out and transmitted to an external computer device via the communication interface 400.

  In the present embodiment described above, the encoding unit 310 performs hierarchical encoding processing on raster data of a document in which characters are written, and acquires hierarchical code data. The decoding unit 340 that supplies the pixel information group obtained by decoding the hierarchical code data to the character recognition unit 350 receives more feedback from the character recognition unit 350 that character recognition has failed. A pixel information group obtained by decoding at a high bit rate is recursively supplied. According to this embodiment having such a functional configuration, it is possible to obtain a character recognition result with extremely high accuracy while suppressing the burden of decoding processing as much as possible.

(Second Embodiment)
In the above embodiment, the decoding process and the character recognition process are performed for each tile area obtained by dividing the raster data. After the character recognition is completed for one tile area, another tile is processed. The processing target was sequentially moved to the area.
On the other hand, in this embodiment, first, the character string is tried by decoding the code string of the entire drawing area of the raster data at the lowest bit rate, and the tile area where the accuracy of character recognition is below a predetermined value is narrowed down. Thereafter, the code sequence of the extracted tile area is decoded at a higher bit rate to try character recognition, and the process of further narrowing down the tile area whose character recognition accuracy is below a predetermined value is recursively repeated.

Since the hardware configuration of this embodiment is the same as that of the first embodiment, the description thereof will be omitted.
Next, a character recognition process that is a characteristic operation of the present embodiment will be described.
FIG. 5 is a flowchart showing the character recognition process.

In this figure, the processing contents of steps 20 to 21 are the same as those of steps 10 to 11 shown in FIG. Further, in the present embodiment, processing corresponding to Step 12 to Step 13 in FIG. 4 is not performed.
When the hierarchical code data is acquired in step 21, the attention area specifying unit 330 encodes the code of the hierarchy corresponding to the image quality level indicated by the numerical value of its image quality level counter from all the tile data included in the hierarchical code data. The column is extracted and output to the character recognition means 350 (S24).

Then, the decoding unit 340 performs a decoding process on the code string input from the attention area specifying unit 330, thereby acquiring a pixel information group for drawing an image of each tile area, and acquiring the acquired pixel information group. It outputs to the character recognition means 350 (S25). That is, in the present embodiment, a set of pixel information groups for each tile area is collectively supplied to the character recognition unit 350.
The character recognizing means 350 tries character recognition based on the pixel information group input from the decoding means 340 (S26). Then, the character recognizing unit 350 determines whether there is a tile area in which the recognition accuracy in step 26 falls below a predetermined value (S27).

  When the determination result in step 27 is “YES”, the character recognizing unit 350 supplies the tile number of the tile area whose accuracy of character recognition is lower than the predetermined value to the attention area specifying unit 330 (S28). At this time, the character recognizing unit 350 stores the text data output as the character recognition result of the area in a memory (not shown) provided for the tile area where the accuracy of character recognition exceeds a predetermined value.

  When acquiring the tile number, the attention area specifying unit 330 adds “1” to the value of its image quality level counter (S29). Then, the attention area specifying unit 330 refers to each tile header included in the hierarchical code data, and specifies tile data of one or more tile areas indicated by the acquired tile number (S30). Further, the attention area specifying unit 330 extracts the code string of the hierarchy corresponding to the image quality level indicated by the value of the image quality level counter of the tile data specified in step 30 and outputs the code string to the character recognition unit 350 (S31). . That is, the attention area specifying means 330 in this embodiment also functions as an area narrowing means for specifying, as a new attention area, an area where the accuracy of recognition by the character recognition means 350 is below a predetermined value. After this processing, returning to step 26, the character recognizing means 350 tries character recognition, and the subsequent processing from step 27 is sequentially executed. Then, the series of processing from step 16 to step 31 is repeated until the accuracy exceeding the predetermined value is output for all tile regions.

When the determination result in step 27 is “NO”, the character recognition unit 350 outputs a series of text data obtained as a result of the character recognition in step 26 to the text data storage unit 360 (S32). The text data storage unit 360 sequentially stores the output text data, and the process ends.
According to the present embodiment described above, the bit rate for decoding the code string of the step-wise code data is stepwise while narrowing down the tile area where the character recognition accuracy falls below a predetermined value every time the character is recognized. Increase it. For this reason, the area for character recognition based on a high-quality image can be narrowed in stages, and the processing burden for character recognition can be reduced.

(Other embodiments)
The present invention can be modified in various ways.
In the above embodiment, encoding and decoding are performed according to the JPEG2000 encoding standard, but encoding and decoding are performed according to other encoding standards such as JBIG (joint bi-level image experts group). Also good.
A program that realizes functions equivalent to those of the encoding unit 310, the attention area specifying unit 330, the decoding unit 340, and the character recognition unit 350 in the above embodiment is mounted on a general-purpose computer device, and the processor of this computer device In addition, processing equivalent to each of the above means may be executed. This type of program may be distributed by being stored in a storage medium such as a CD-ROM, or may be distributed in response to a request from a client device from a server device provided on a network.
In the above embodiment, the character recognition unit 350 has both a function of recognizing characters from the pixel information group of the region of interest and a function of obtaining and outputting the accuracy of character recognition based on the success or failure of the recognition. These functions may be realized by individual means. In the second embodiment, the attention area specifying unit 330 has a function of specifying, as a new attention area, an area in which the accuracy of recognition by the character recognition means 350 falls below a predetermined value. Another means independent of the means 330 may be used.
The code string obtained by encoding the raster data does not need to be embedded. For example, raster data is individually encoded at a plurality of bit rates to obtain code data including each code string, and character recognition is performed using these code data in order from the lowest image quality. May be. Conceptually showing the configuration and operation of such a modified example, “the input means for inputting an image obtained by scanning a document in which one or more characters are written, and the input image at different bit rates. An encoding process is performed individually, encoding means for acquiring each code data obtained by encoding the image at different bit rates, storage means for storing each acquired code data, and the storage means, Extraction means for extracting code data of a first bit rate; decoding means for performing decoding processing on the extracted code data; recognition means for recognizing characters from an image obtained by the decoding processing; Control for switching the bit rate, which is a key for extracting code data by the extraction unit, to a second bit rate higher than the first bit rate when the accuracy of recognition by the recognition unit falls below a predetermined value Character recognition device and a stage. Becomes. "

It is a hardware schematic block diagram of a character recognition apparatus. It is a block diagram which shows the internal structure of a controller. It is a data structure figure of hierarchical code data. It is a flowchart which shows a character recognition process (1st Embodiment). It is a flowchart which shows a character recognition process (2nd Embodiment).

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Scanner, 200 ... Operator, 300 ... Controller, 310 ... Coding means, 320 ... Code data storage means, 330 ... Region of interest specifying means, 340 ... Decoding means, 350 ... Character recognition means, 360 ... Text data storage means.

Claims (10)

Input means for inputting an image obtained by scanning a document in which one or more characters are written;
Encoding means for performing hierarchical encoding processing on the input image and acquiring hierarchical code data including a code string hierarchized according to a bit rate;
Storage means for storing the acquired hierarchical code data;
Decoding means for performing a decoding process at a specified bit rate on all or part of a code string included in the hierarchical code data stored in the storage means;
Recognition means for recognizing characters from the image obtained by the decoding process;
A character recognition device comprising: control means for controlling a bit rate in the decoding process based on success or failure of recognition by the recognition means. Input means for inputting an image obtained by scanning a document in which one or more characters are written;
Encoding means for performing hierarchical encoding processing on the input image and acquiring hierarchical code data including a code string hierarchized according to a bit rate;
Storage means for storing the acquired hierarchical code data;
Decoding means for performing a decoding process at a first bit rate on all or part of a code string included in the hierarchical code data stored in the storage means;
Recognition means for recognizing characters from the image obtained by the decoding process;
And a control means for switching the bit rate in the decoding process to a second bit rate higher than the first bit rate when the recognition accuracy by the recognition means falls below a predetermined value. Input means for inputting an image obtained by scanning a document in which one or more characters are written;
Encoding means for performing hierarchical encoding processing on the input image and acquiring hierarchical code data including a code string hierarchized according to a bit rate;
Storage means for storing the acquired hierarchical code data;
Decoding means for identifying a code string corresponding to a region of interest for character recognition from the hierarchical code data stored in the storage means, and performing decoding processing at the first bit rate for the identified code string When,
Recognition means for recognizing characters from the image obtained by the decoding process;
And a control means for switching the bit rate in the decoding process to a second bit rate higher than the first bit rate when the recognition accuracy by the recognition means falls below a predetermined value. The character recognition device according to claim 3.
The encoding means includes
The input image is divided into a plurality of drawing regions that do not have overlapping portions, and hierarchical coding processing is performed on each divided drawing region to obtain hierarchical code data for each drawing region,
The decoding means includes
A character recognition device for sequentially specifying each of the divided drawing areas as the attention area. Input means for inputting an image obtained by scanning a document in which one or more characters are written;
Encoding means for performing hierarchical encoding processing on the input image and acquiring hierarchical code data including a code string hierarchized according to a bit rate;
Storage means for storing the acquired hierarchical code data;
Decoding means for identifying a code string corresponding to a region of interest for character recognition from the hierarchical code data stored in the storage means, and performing decoding processing at the first bit rate for the identified code string When,
Recognition means for recognizing characters from the image obtained by the decoding process;
Of the attention area, an area narrowing means for specifying, as a new attention area, an area where the accuracy of recognition by the recognition means falls below a predetermined value;
A character recognition device comprising: control means for switching a bit rate at which the decoding means performs a decoding process for the new attention area to a second bit rate higher than the first bit rate. The character recognition device according to claim 5.
The encoding means includes
The input image is divided into a plurality of drawing areas that do not have overlapping portions, and hierarchical coding processing is performed on each divided drawing area to obtain hierarchical code data for each drawing area,
The area narrowing means is
A character recognition device that determines whether or not the accuracy of character recognition is an area below a predetermined value for each of the divided drawing areas. Input means for inputting an image obtained by scanning a document in which one or more characters are written;
In a computer device comprising information storage means,
A coding function for performing hierarchical coding processing on the input image and acquiring hierarchical code data including a code string hierarchized according to a bit rate;
A data storage function for storing the acquired hierarchical code data in the storage means;
A decoding function that performs a decoding process at a specified bit rate on all or part of a code string included in the hierarchical code data stored in the storage unit;
A recognition function for recognizing characters from the image obtained by the decoding process;
And a control function for controlling a bit rate in the decoding process based on success or failure of recognition by the recognition function. Input means for inputting an image obtained by scanning a document in which one or more characters are written;
In a computer device comprising information storage means,
A coding function for performing hierarchical coding processing on the input image and acquiring hierarchical code data including a code string hierarchized according to a bit rate;
A data storage function for storing the acquired hierarchical code data in the storage means;
A decoding function for performing a decoding process at a first bit rate on all or part of a code string included in the hierarchical code data stored in the storage unit;
A recognition function for recognizing characters from the image obtained by the decoding process;
And a control function for switching the bit rate in the decoding process to a second bit rate higher than the first bit rate when the recognition accuracy by the recognition function falls below a predetermined value. Input means for inputting an image obtained by scanning a document in which one or more characters are written;
In a computer device comprising information storage means,
A coding function for performing hierarchical coding processing on the input image and acquiring hierarchical code data including a code string hierarchized according to a bit rate;
A data storage function for storing the acquired hierarchical code data in the storage means;
A decoding function for identifying a code string corresponding to a region of interest for character recognition from the hierarchical code data stored in the storage means, and performing a decoding process at a first bit rate on the identified code string When,
A recognition function for recognizing characters from the image obtained by the decoding process;
And a control function for switching the bit rate in the decoding process to a second bit rate higher than the first bit rate when the recognition accuracy by the recognition function falls below a predetermined value. Input means for inputting an image obtained by scanning a document in which one or more characters are written;
In a computer device comprising information storage means,
A coding function for performing hierarchical coding processing on the input image and acquiring hierarchical code data including a code string hierarchized according to a bit rate;
A data storage function for storing the acquired hierarchical code data in the storage means;
A decoding function for identifying a code string corresponding to a region of interest for character recognition from the hierarchical code data stored in the storage means, and performing a decoding process at a first bit rate on the identified code string When,
A recognition function for recognizing characters from the image obtained by the decoding process;
Among the attention areas, an area narrowing function for specifying, as a new attention area, an area in which the recognition accuracy by the recognition function is lower than a predetermined value;
A program for realizing a control function for switching a bit rate at which the decoding function performs a decoding process for the new attention area to a second bit rate higher than the first bit rate.

Images