JPH08293000A - Device and method for image processing - Google Patents

Abstract

PURPOSE: To shorten the processing time by judging the direction of a cut character image and determining the direction of the image on the basis of the judged direction of the character image. CONSTITUTION: A scanner 10 reads the document image and stores the image in a RAM 13 as image data whose document direction should be discriminated. Then, a character area is extracted from the image data stored in the RAM 13. Then, one of extracted character areas is selected and a character is cut out of the character area. Further, the cut character is recognized in four directions of 0, 90, 180, and 279 deg. and respective similarities are calculated. The direction having the largest similarity is determined as the direction of the document. On the basis of the decided direction of the document, the image stored in the RAM 13 is rotated at need and saved as image filing data in an external storage device 15.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method and apparatus.

[0002]

2. Description of the Related Art In a digital copying machine, an original image or the like is optically read and subjected to digital processing to form an image on a recording paper. In such an apparatus, it is possible to read a document composed of a plurality of sheets one after another and form the image on the recording paper. Here, the orientation of the copied document corresponds to the orientation of the original document provided to the device in a one-to-one correspondence. Further, also in the electronic filing device, when the read image is displayed on the display, it is displayed in the direction corresponding to the reading.

[0003]

Therefore, when inputting the image of the document, the document obtained by copying is not in the correct direction unless the original is input in the correct direction. For example, when one of a plurality of documents is turned upside down, only the page of the document obtained by copying is turned upside down. When the copying is executed in such a state, the user is required to confirm the input and then take out the document of the wrong page and rearrange the directions. Further, in the case of the electronic filing device, it is necessary to set the document of the wrong page in the correct direction in the reading device and perform the reading process again, or to rotate the input document image according to the user's designation. Become.

As described above, in the conventional digital copying machine or electronic filing apparatus, when a document is input in the wrong direction, the user needs to make corrections by some work, and the user must be burdened.

The present invention has been made in view of the above problems, and it is an object of the present invention to provide an image processing method and apparatus capable of determining the direction of an image from the read image data.

[0006]

An image processing apparatus of the present invention for achieving the above object determines a cutting-out means for cutting out a character image from an image and a direction of the character image cut out by the cutting-out means. The determination means and the determination means for determining the direction of the image based on the determination result by the determination means.

Further, preferably, the determination means determines that the character image cut out by the cutout means is 0 degrees or 90 degrees.
It is determined which of the four directions of 180 degrees, 180 degrees, and 270 degrees. Normally, the document placement direction is
Since it is one of the directions, it is practical to judge if it is one of these directions.

An image processing apparatus according to claim 1, wherein:

Further, an image processing apparatus having another configuration of the present invention which achieves the above-mentioned object has a cutting-out means for cutting out a plurality of character images from an image, and a plurality of directions for each character image cut out by the cutting-out means. And a determining unit that determines the direction of the image to be one of the plurality of directions based on the similarity calculated by the similarity calculating unit.

Preferably, the plurality of directions are 0
There are four directions of 90 degrees, 90 degrees, 180 degrees, and 270 degrees.

Further, preferably, when the predetermined number of character images are cut out by the cut-out means, the cut-out processing is ended. When cutting out a character image from an image, the processing time can be shortened by ending the cutting out of the character when the number of cut out character areas reaches a predetermined number.

Further, preferably, a plurality of cut out character images are combined to form a new character image, and when the similarity is improved by the new character image, the calculation means is based on the similarity. It further comprises update means for updating the obtained similarity. When the extracted character images are combined to improve the similarity, the combined images are used.For example, even if the character images are disassembled and dissected and divided into parts, the recognition degree in the correct extraction state is improved. Can be calculated.

Further, preferably, when the half-width character size character images are continuous, the updating means combines these to form a full-width size character image, and based on the similarity degree when the similarity degree is improved. Then, the similarity obtained by the calculating means is updated. It is possible to exclude a result recognized by erroneous clipping of a character and prevent a decrease in similarity in the correct direction.

Further, preferably, the updating means performs combining into a full-size character when the similarity of the half-size character image is smaller than a predetermined value. Since the possibility of combining half-size character images with low similarity to full-size characters and the like is examined, the processing becomes efficient and the processing time can be shortened.

Further, preferably, the calculating means calculates the similarity in each direction for the character image which is determined to be a full-width character, and the determining means determines each of the similarity calculated by the calculating means. The direction of the image is determined based on the total for each direction. Since the full-size character image is used, it is possible to prevent the character from being cut out incorrectly and prevent the degree of similarity in the correct direction from decreasing. Moreover, since all the character images are full-size, it is not necessary to combine the character images. Therefore, the number of characters in each direction does not change, the direction of the image can be determined based on the sum of the similarities, and the processing content is simplified.

Further, preferably, when the number of characters used for the direction determination in the calculation means is less than a predetermined number, the character image is cut out by the cutout means, and the calculation means and the determination means are added. It further comprises control means for repeatedly executing. Since the determination process is repeated until the number of character images used for determining the direction of an image becomes equal to or larger than a predetermined value, the influence of a character cutout error is reduced and the accuracy of the direction determination is improved.

Preferably, in the similarity in each direction calculated by the calculating means, when the difference between the first-rank similarity and the second-rank similarity is less than a predetermined value, the clipping means separates the two. A control means is further provided, which cuts out a character image from the character area and repeatedly executes the calculation means and the determination means. The orientation of the image can be determined in the character area where the difference between the first and second similarities exceeds a predetermined value, and the orientation is erroneously determined for a portion that is not a character area or a type of character area that is not targeted by this character recognition method. Therefore, the accuracy of the direction determination is improved.

Further, preferably, in the similarity in each direction calculated by the calculating means, when the first-order similarity is less than a predetermined value, the cutting means cuts out a character image from another character area. And a control means for repeatedly executing the calculation means and the determination means. If the first-ranked degree of recognition is smaller than a predetermined value,
Characters are cut out from another character area, and the determination process is performed until the first recognition degree portion having a predetermined value or more is detected. This prevents a portion that is not a character area or a character area of a type that is not a target in the present character recognition method from being mistakenly used for the direction determination, and the accuracy of the direction determination is improved.

Further, preferably, when the image has an area of a figure or a table, the character area in the figure or the table has a lower priority for cutting out the character image by the cutting means. By lowering the priority used for the determination in the character areas in the figures and tables, erroneous determination of the direction determination is prevented. still,
It may be prohibited to use the character region of the figure or table for the determination.

Also, preferably, the plurality of directions is 0.
Or 180 degrees. In the case where the image reading device can take only two directions with respect to the document feeding direction, two directions of 0 degree or 180 degrees are sufficient.
Therefore, the processing speed can be improved by limiting the calculation of the degree of similarity to two directions.

Further, preferably, the image processing apparatus further comprises an output means for rotating the image as necessary based on the direction of the image judged by the judging means and outputting the image. For example, in an electronic filing device, even if there is a page in which the orientation of the original is wrong, it is possible to automatically align the filing with the correct page for filing, and improve the operability.

Further, preferably, the calculating means calculates the similarity corresponding to the plurality of directions by changing the scanning direction of the cut out character image and calculating the similarity. Since the similarity in each direction is obtained without performing the process of rotating the character image, the processing speed is improved.

Further, preferably, the calculating means divides the cut-out character image into a predetermined number of blocks, scans in a predetermined direction, counts the direction component of the image for each block, and obtains each obtained. The count value for each direction component for each block is stored as feature data, and the feature data for directions other than the predetermined direction is calculated by replacing the count values in the block unit and the direction component unit. The similarity is calculated based on the feature data in each direction. When the character image is operated in a predetermined direction and the characteristic data is acquired,
Since the characteristic data in the other direction is obtained by replacing the obtained characteristic data, the processing speed can be further improved.

[0024]

According to the above construction, a character image is cut out from a given image, and the direction of the cut out character image is determined. Then, based on the determined direction of the character image, the direction of the image is determined.

Further, according to the above-mentioned another configuration of the present invention, a plurality of character images are cut out from the image, and the similarity in a plurality of directions is calculated for each of the cut out character images. The direction of the image is determined as one of the plurality of directions based on the degree of similarity of each direction calculated in this way.

[0026]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT A preferred embodiment of the present invention will be described below with reference to the accompanying drawings.

In the following description of the embodiments, the case where the present invention is applied to an electronic filing apparatus will be described.

FIG. 1 is a block diagram showing the configuration of the electronic filing apparatus in the embodiment. In the figure, 10
Is a scanner, which optically reads a set original image to obtain image data. A CPU 11 controls the entire electronic filing apparatus. Reference numeral 12 denotes a ROM, which stores various control programs to be executed by the CPU 12 and data referred to in various processes.
A RAM 13 provides a working memory area for the CPU 11. In addition, when determining the orientation of the document, which will be described later, 1
It stores the image data of a page sentence document. An interface 14 connects the scanner 10 and the bus 17.

An external storage device 15 stores the document image data to be filed. Reference numeral 16 is a display for performing various displays. For example, it is possible to read the document image data stored in the external storage device 15 and display it on the display 16. A print unit 18 forms a visible image on a recording medium. For example, the document image stored in the external storage device 15 can be formed on the recording paper.

Next, the operation of the electronic filing apparatus of this embodiment having the above configuration will be described.

FIG. 2 is a flow chart showing the operation procedure of the electronic filing apparatus in this embodiment.

First, in step S1, the scanner 1
The document image is read at 0 and stored in the RAM 13 as image data for which the document direction should be determined. Next, in step S2, a character area is extracted from the image data stored in the RAM 13. In step S3, one of the character areas extracted in step S2 is selected, and a character is cut out from the character area. The character cutout is once terminated when the cutout character area reaches a predetermined number. Although the details will be described later, the number of characters increases or decreases when the recognition processing is performed. Therefore, the number of character cutouts in step S3 is empirically determined so that the final number of characters becomes equal to or larger than the threshold value. Note that a well-known technique such as a character recognition technique can be applied to the extraction of the character region and the cutting of the character.

In step S4, the cut-out character is subjected to recognition processing in four directions of 0, 90, 180, and 270 degrees, and each average similarity is calculated. Then, in step S5, the direction having the highest average similarity is determined as the direction of the document. In step S6, the RAM 1 is determined based on the original orientation determined in step S5.
The image stored in No. 3 is rotated as necessary, and is stored in the external storage device 15 as image filing data.

On the other hand, if the character area cannot be extracted in step S2, the orientation of the image cannot be detected in this embodiment. In this case, the process proceeds to step S7,
The image in the RAM 13 is stored as it is in the external storage device 15 as electronic filing data. Step S
If the predetermined number of character areas cannot be cut out in 3, the direction detection may be performed by setting a higher threshold value for determining the direction described later.

The above steps S3, S4, S5, S6
Each of the processes will be described in more detail.

FIG. 3 is a flowchart showing the procedure of the character cutting process (step S3) in this embodiment.

In the character cutting process, first, in step S31, one of the character regions extracted in step S2 is selected. Next, in step S32, a character is cut out in the selected character area. When the characters can be cut out, the process proceeds from step S33 to step S34, and it is determined whether the number of characters cut out has reached a predetermined value. If a predetermined number of cutouts have been made, this processing is terminated, but if the cutout number has not reached the predetermined value, the process returns to step S32.

If the character cannot be cut out in step S32, the process proceeds from step S33 to step 35, and it is determined whether or not there is an unprocessed character area (a character area that has not been selected yet). If there is an unprocessed character area, the process returns to step S31, a new character area is selected, and the process proceeds to step S32.

In this way, the processing is carried out until the predetermined number of characters have been cut out, but when the unprocessed place area has disappeared before the predetermined number of characters have been cut out. Goes from step S35 to step S7 in FIG.

Next, recognition processing in four directions (step S4)
The process will be described. FIG. 4 is a flowchart showing the procedure of the four-direction recognition process in this embodiment.

First, in step S41, each area cut out in the previous step S3 is scanned in the direction of 0 °, the feature amount is extracted, and the similarity is calculated. And
The sum of the similarities for all the cut out areas is calculated, and this is divided by the number of characters to calculate the average similarity. Hereinafter, in steps S42, S43, and S44, 9
Scanning is performed in the 0 ° direction, the 180 ° direction, and the 270 ° direction, and the average similarity is calculated in the same procedure as in step S41.

FIG. 5 is a diagram for explaining the scanning direction for feature extraction in this embodiment. As described above, in this embodiment, scanning in four directions of 0 °, 90 °, 180 °, and 270 ° and extraction of the characteristic amount are performed.

Here, in order to perform recognition processing in each direction, it is possible to actually rotate the image before recognition. However, in this embodiment, a method of changing the scanning direction is used as shown in FIG. 5 in order to shorten the processing time.

The recognition process is divided into a feature extraction process for extracting a feature of the character image as a feature amount and an identification process for performing matching with the obtained feature amount. When extracting character features, by changing the scanning direction of the character image for each angle, the same feature amount as when extracting the feature by rotating the image is extracted without rotating the image. it can. That is, as shown in FIG. 5, if the feature amount is extracted from the upper left to the lower right of the character image at 0 degrees,
In the case of 90 degrees, the feature amount should be extracted from the lower left to the upper right.

Next, the calculation of the average similarity in this embodiment will be described. FIG. 6 is a flowchart showing a procedure of calculating the average similarity in a scanning direction according to this embodiment. For example, in step S41, the average similarity by scanning in the 0 degree direction is calculated by executing each step in FIG. 6 with the designated angle as 0 degree. In the following description, each step in FIG. 6 will be described by taking the scanning in the 0 ° direction as an example, but the same processing is performed in the scanning in each of the 90 °, 180 °, and 270 ° directions.

First, in step S401, one cutout area is scanned in a designated direction to acquire a feature amount. In this case, 0 in FIG.
The scanning in the degree direction is performed. Then, the degree of similarity is calculated by performing matching with each character pattern for the cutout area, and cumulative addition is performed. In addition, the exposed region that has been scanned is counted as one character, and the number of characters is counted. The above processing is executed for all the cutout areas cut out in the previous step S3 (steps S404 and S405). As a result, it is possible to obtain the sum of the similarities and the number of characters for all the cutout areas cut out in step S3.

When the calculation of the degree of similarity is completed for all the cutout areas as described above, the process proceeds to step S406, where the cutout areas of the continuous cutout areas are half-width characters or full-width characters. It is determined whether or not it constitutes.

More specifically, when the characters are cut out, there is only information on the size and pitch of each character as shown in FIG. Therefore, there is a part where it cannot be determined whether it is a sequence of half-width characters such as "1234" or a sequence of full-width characters that can be separated into a half-size cutout area such as "Ihaho". Therefore, in order to determine whether the half-width character is a part of the full-width character, it is necessary to adopt the best one from the results of recognizing the half-width character at each angle and recognizing multiple character rectangles with one character. .

In step S406, a continuous half-size region is docked among the cut out regions,
One cutout area. Here, the half-size area can be determined by the aspect ratio of the cut out area. Then, in step S407, the new cutout area formed by docking is scanned in the designated direction (here, scanning in the direction of 0 degree) to extract the feature amount, and the similarity is calculated. As a result, when the similarity calculated in the full-width size area after docking is greater than the similarity calculated in the half-width size area before docking, the cutout area is determined to be full-width characters. To do. Then, the cumulative similarity and the count value of the number of characters obtained in the previous steps S401 to S405 are updated based on the docking result of the cutout area (step S4).
08, S409).

FIG. 8 is a view for explaining an example of docking a half-size cutout area. When three half-size areas are lined up as shown in the figure ((A) of FIG. 8), two docking methods are possible ((B) and (C) of FIG. 8).
In this case, the sum of the similarities, b + c + d, f + d, and b +
The cut-out state having the maximum value of g is adopted.

The above processing is executed for all dockable portions (step S410) to obtain the maximum similarity. Then, the maximum similarity is divided by the character number count value at this time to calculate the average similarity (step S411). The reason why the average similarity is used here is that when the above process is performed in each of the scanning directions of 0, 90, 180, and 270 degrees, the number of characters at each angle varies due to the docking process and the like.

In the above feature amount extraction, as shown in FIG. 5, the direction in which the image of the cutout area is scanned is changed to obtain the feature amount equivalent to rotating the image. However, the method of acquiring the feature amount corresponding to the rotated image is not limited to this. Here, another method for acquiring the feature amount corresponding to the rotated image will be described with reference to FIGS.
This will be described using 1.

FIG. 9 is a flow chart for explaining another method for acquiring the feature quantity corresponding to the rotated image. FIG. 10 is a diagram showing a block division of a cutout region and a direction component for obtaining a feature amount by another method. FIG. 11 is a diagram showing a storage state of data obtained by acquiring a feature amount by another method.

According to the method described here, the feature amount once extracted is used as a reference to convert the feature amount to extract the feature amount of the rotated character image. First, step S
In 421, the character image (cutout area) is divided into blocks 1 to 16 as shown in FIG. And
In step S422, with respect to the contour image included in each block, the appearance amounts of the direction components from the direction 1 to the direction 4 shown in FIG. 10 are counted and set as the feature amount. The feature amount for each direction component of each block thus obtained is stored in the RAM as shown in FIG.
Store in 13.

After extracting the 0-degree feature, 90, 18
The 0,270 degree feature amount can be generated from this 0 degree feature amount. More specifically, for example, for the 90-degree feature amount, the feature amounts of the directions 1 and 3 and the directions 2 and 4 are exchanged, and the data of block 1 is called block 4 and the data of block 2 is called block 8. You can generate it by moving it like this. The 180-degree and 270-degree feature amounts can also be obtained by exchanging the feature amounts of the blocks and the direction components. This method is faster than the method shown in FIG. 5 because it can be performed only by exchanging data without accessing the image.

Whichever method is used, it is not necessary to rotate the image itself, and the feature quantity for each angle can be extracted at high speed.

When the average degree of similarity in each direction is obtained as described above, the result is determined in step S5.
The determination process will be described in more detail. FIG. 12 is a flow chart showing the direction determining procedure in this embodiment.

First, in step S501, the highest average similarity is extracted. Then, in step S502, it is determined whether or not this average similarity is larger than a predetermined value. The direction having the highest average similarity is the correct direction, but as a problem, there is a possibility that it is not the character region when the character region is extracted. Also,
Even in the case of a character area, if the recognition means used to determine the direction of a halftone dot is a type of character area that is not compatible, the direction cannot be correctly determined. Therefore, if the average similarity in the first direction is lower than the predetermined value, step S50
From step 2, the process proceeds to step S508, and another character area is used for the determination.

On the other hand, even if recognition processing is performed from any of the four directions such as "1 + 100 = 101", there is a character having a high degree of similarity, and the difference in the degree of similarity between the directions is small. In addition, in this example, the image rotated 270 degrees counterclockwise is number 1,
It is not unusual for the 0 part to be recognized as one of the Chinese numerals, and it is not possible to judge from which direction it is correct even if knowledge processing such as syntax analysis is performed (see FIG. 13). Since it is inappropriate to use such a character area for determining the document direction, another character area must be used for determining the direction. Therefore, in step S503, the average similarity of the second rank is obtained, and the difference between the average similarity of the first rank and the average similarity of the second rank is calculated. If the difference in average similarity between the first and second positions is small, the character region is not used for the direction determination, and another character region is used so as to use another character region.
Go to.

Further, although it has been described above that the count value of the number of characters is different in each of the four directions, if the final number of characters does not exceed the predetermined number, step S3 is performed.
Then, the process returns to and the cutout of the remaining characters in the character area being processed is added (steps S505 and S506). Further, when the character segmentation of the character area being processed is completed, the processing is executed for the other character areas extracted in step S2, as described with reference to the flowchart of FIG.

When the above conditions are satisfied, the direction having the maximum average similarity is set as the direction of the image and the result is output (step S507).

As described above, according to this embodiment, the orientation of the original is determined by determining the direction of the characters, and the image is rotated if necessary based on this to perform electronic filing. Therefore, even if the original is read in the wrong direction, it is automatically corrected to the correct direction and saved, so that the operability is dramatically improved.

Further, in extracting the feature amount obtained by rotating the image in the clipped area, it is not necessary to rotate the image itself, so that the processing speed is improved.

Further, for full-width / half-width characters, the cutout areas are combined so that the degree of similarity is maximized.
The degree of similarity can be calculated accurately.

In the above embodiment, the area joining is tried for all the continuous cut-out areas having the half-width size. However, the half-size area having the similarity less than a predetermined value is set as the joining target. Good.

Further, according to the above-mentioned embodiment, when the difference between the average similarity of the first rank and the average similarity of the second rank is less than a predetermined value, it is determined that the orientation of the document cannot be identified. The process of determining the direction is redone using another character area. For this reason, it is possible to avoid determining the direction by using numbers or the like whose vertical and horizontal identification is difficult, and it is possible to accurately determine the direction of the original.

When a figure is described in the original document, the characters in the figure may be written in a direction different from the correct document direction. For example, as shown in FIG. 14, the numbers representing the dimensions are written in a direction different from the correct document direction. Therefore, in the extraction of the character area in step S2 described above, when characters in the drawing or characters close to the drawing are also extracted, the character area in the drawing is not used at all in the determination and is ignored, or the determination processing is prioritized. Try not to use it as low as possible.

Further, in the case of a document in which only characters in the drawing or characters close to the drawing exist, if there are a plurality of them, various judgment results are majority-determined to determine the direction of the image.

Similarly, the letters in the table are likely to be numbers. The numbers are difficult to distinguish even if the recognition process is performed by rotating the numbers, such as “1, 6, 8, 9, 0”. Therefore, for the character areas in the table, the priority of the determination process is set to a low level so that it is not used as much as possible.

As a method for determining a character area, a figure, and a table from image data obtained by reading an original, for example, a cluster of black pixels independent of the image data is taken out, and the size and arrangement of the cluster are A method based on internal and external relationships is known. As a method of extracting the black pixel, there are a labeling method and a method of performing contour tracking.

Further, in step S2, the character areas may be sorted according to the size and used for determining the direction from the larger character area. Alternatively, when the number of characters in each character region is known when extracting the character regions in step S2, the character regions may be sorted in descending order of the number of characters and used for direction determination from the character region having the largest number of characters. By doing so, it is highly possible that a necessary number of cut-out regions can be obtained from the first character region, so that the processing becomes faster.

Further, in step S2 and step S3, an area of a certain size or more may be searched for in the size of each character area, and the area found may be used for the direction determination. When the last character area is processed, the character area skipped back to the beginning is used for the direction determination. Since processing such as sorting is unnecessary, the processing becomes faster when the number of extracted character areas is large.

Further, in the character cut-out in step S3, a certain number or more of characters are cut out except for the characters that cannot be determined to be half-width characters or full-width characters, and 0, 90, 180, 270 degrees of each character are cut out. May be recognized to calculate the degree of similarity. Since only half-width / full-width characters can be determined for determination, the number of characters at each angle does not change and the number of characters can be easily managed. Further, since the direction determination processing is performed only on the portion where the character cutting out accuracy is high, the result determined from that portion is also highly reliable. Also, the same effect can be obtained by using only the full-width cutout region.

Further, in a system in which the orientation of the original document is limited to either portrait orientation or landscape orientation, 0 degree or 180 degrees.
It only needs to be judged a few times.

In the above embodiment, the application to the electronic filing apparatus has been described, but the application to other apparatuses is not limited. For example, if it is applied to a digital copying machine, even if document pages in the wrong direction are mixed, it is possible to automatically rotate the image in the correct direction and form an image on the recording paper. There is no need to modify it by operation, improving operability.

In the above embodiment, the orientation of the image is determined after reading one page of the document, but the invention is not limited to this. For example, the original may be read in units of parts (by band processing) and the processing for direction determination may be executed.
When the determination can be made, the processing for each band is terminated, and the processing for direction correction is started. As a result, the direction can be determined while scanning the document in the reading device, so that the direction can be determined more quickly. It should be noted that the characters spanning the bands are combined with the image of the previous band.

The present invention may be applied to a system composed of a plurality of devices or an apparatus composed of a single device. Further, it goes without saying that the present invention can be applied to the case where it is achieved by supplying a program to a system or an apparatus.

[0078]

As described above, according to the present invention,
It is possible to determine the direction of the image from the read image data.

Further, according to the present invention, since the degree of similarity between a plurality of character images is used to determine the direction, it is possible to reduce the influence of a failure in cutting or recognizing a character and to make a more accurate determination.

Further, according to the present invention, when recognizing a character image from each direction, it is not necessary to rotate the image itself, so that the speed is high.

Further, according to the present invention, since the recognition processing is not performed for all the characters in the document, the processing can be performed at high speed.

Further, according to the present invention, since full-width characters / half-width characters are determined for each of the four directions, unnecessary patterns that reduce the degree of recognition are removed, and the accuracy of direction determination is further improved. be able to.

Further, according to another configuration of the present invention, since only the characters estimated to be full-width characters are used for the determination, full-width characters /
The determination based on the degree of recognition of half-width characters becomes unnecessary. Also, 4
Since the number of characters in the direction is the same, it is not necessary to calculate the average value of the similarities, and the direction of the image can be determined in total.

[0084]

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an electronic filing device in an example.

FIG. 2 is a flowchart showing an operation procedure of the electronic filing device in this embodiment.

FIG. 3 is a flowchart showing a procedure of character cutout processing (step S3) in the present embodiment.

FIG. 4 is a flowchart showing a procedure of four-direction recognition processing in this embodiment.

FIG. 5 is a diagram illustrating a scanning direction for feature amount extraction according to the present exemplary embodiment.

FIG. 6 is a flowchart showing a procedure of calculating an average similarity in a scanning direction according to the present embodiment.

FIG. 7 is a diagram illustrating a state in which a character is cut out from a character area.

FIG. 8 is a diagram illustrating an example of docking a cut-out region having a half-width size.

FIG. 9 is a flowchart illustrating another method of acquiring a feature amount corresponding to a rotated image.

FIG. 10 is a diagram showing a block division of a cutout region and a direction component for obtaining a feature amount by another method.

FIG. 11 is a diagram showing a storage state of data obtained by acquiring a feature amount by another method.

FIG. 12 is a flowchart showing a direction determining procedure in the present embodiment.

FIG. 13 is a diagram illustrating an example of a character string in which it is difficult to identify rotation.

FIG. 14 is a diagram showing an example of directions of characters in a document described in a document.

[Explanation of symbols]

 10 Scanner 11 CPU 12 ROM 13 RAM 14 Interface 15 External Storage Device 16 Display 17 Bus 18 Printing Section

Claims (20)

[Claims] 1. A cutout unit for cutting out a character image from an image, a judgment unit for judging the direction of the character image cut out by the cutout unit, and a direction of the image based on the judgment result of the judgment unit. An image processing apparatus comprising: a determining unit. 2. The determination means determines that the character image cut out by the cutout means is 0 degree, 90 degrees, 180 degrees, 27 degrees.
The image processing apparatus according to claim 1, wherein it is determined which one of the four directions of 0 degrees. 3. A cutting-out means for cutting out a plurality of character images from an image, and each character image cut out by the cutting-out means,
And a determining unit that determines the direction of the image to be one of the plurality of directions based on the similarity calculated by the similarity calculating unit. Image processing device. 4. The plurality of directions are 0 °, 90 °, 18
The image processing apparatus according to claim 3, wherein the image processing apparatus has four directions of 0 degree and 270 degrees. 5. The image processing apparatus according to claim 3, wherein when the predetermined number of character images are cut out by the cutout unit, the cutout process is ended. 6. A plurality of clipped character images are combined to form a new character image, and when the similarity is improved by the new character image, the new character image is obtained by the calculation means based on the similarity. The image processing apparatus according to claim 3, further comprising an updating unit that updates the degree of similarity. 7. The updating means, when character images of half-width character size are continuous, combines these to form a character image of full-width size, and when the similarity is improved, the calculation is performed based on the similarity. The image processing apparatus according to claim 6, wherein the similarity obtained by the means is updated. 8. The image processing apparatus according to claim 7, wherein the updating unit performs combining to a full-size character when the similarity of the half-size character image is smaller than a predetermined value. 9. The calculating means calculates the similarity of each direction for a character image that is determined to be a full-width character, and the determining means determines the similarity of each direction of the similarity calculated by the calculating means. The image processing apparatus according to claim 3, wherein the direction of the image is determined based on a total amount. 10. When the number of characters used for the direction determination in the calculation means is less than a predetermined number, the character image is cut out by the cutout means, and the calculation means and the determination means are repeatedly executed. The image processing apparatus according to claim 3, further comprising a control unit. 11. In the similarity in each direction calculated by the calculating means, if the difference between the first-rank similarity and the second-rank similarity is less than a predetermined value, the clipping means separates another character area. The image processing apparatus according to claim 3, further comprising a control unit that cuts out a character image from the image, and repeatedly executes the calculation unit and the determination unit. 12. In the similarity in each direction calculated by the calculating means, if the first-order similarity is less than a predetermined value, the cutting means cuts out a character image from another character area, The image processing apparatus according to claim 3, further comprising: a control unit that repeatedly executes the calculation unit and the determination unit. 13. The method according to claim 3, wherein when the image has a figure or table area, the character area in the figure or table has a lower priority for cutting out the character image by the cutting means. Image processing device. 14. The plurality of directions are 0 ° or 18
The image processing apparatus according to claim 3, wherein the image processing apparatus has two directions of 0 degree. 15. The image processing according to claim 3, further comprising an output unit that rotates and outputs the image as needed based on the direction of the image determined by the determination unit. apparatus. 16. The image processing according to claim 3, further comprising an output unit that rotates and outputs the image as needed based on the direction of the image determined by the determination unit. apparatus. 17. The calculation means calculates the similarity corresponding to the plurality of directions by changing the scanning direction of the clipped character image and calculating the similarity. The image processing device according to item 3. 18. The calculating means divides the cut-out character image into a predetermined number of blocks, scans in a predetermined direction, counts the direction component of the image for each block, and obtains each block obtained. The count value for each direction component is stored as feature data, and the feature data in a direction other than the predetermined direction is calculated by
The image processing apparatus according to claim 3, wherein the count value is exchanged in the block unit and the direction component unit, and the similarity is calculated based on the obtained feature data in each direction. 19. A cutting out step of cutting out a character image from an image, a judging step of judging the direction of the character image cut out in the cutting step, and a direction of the image based on the judgment result of the judging step. An image processing method comprising: a determining step. 20. A cutting step of cutting out a plurality of character images from an image, and each character image cut out in the cutting step,
A calculation step of obtaining a similarity for a plurality of directions; and a determination step of determining the direction of the image as one of the plurality of directions based on the similarity calculated in the similarity calculation step. Image processing method.