JPH07334620A - Image processor - Google Patents

Abstract

PURPOSE:To improve the picture quality correcting function of an image processor by correcting the partial runnings or distortions of characters. CONSTITUTION:When a character is detected by a character detecting part 51, a difference deciding part 58 decides the differences from other characters. If the running parts of characters are detected, a character recognizing part 52 recognizes the character and decides whether the character can be read or not based on a ratio set between the total number of character strings having the running parts and the number of characters recognized by the part 52. If the character can not be read, the characters having the running parts are corrected by a character correcting part 54 and the image data are synthesized by a synthesizing part 57. If no character is detected, a picture quality correction mode is canceled and a copying operation is carried out. In such a way, since the running parts of characters of an original image are detected and corrected, a copied image that is substantially available can be acquired.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus such as a digital copying machine, a facsimile, a scanner / printer.

[0002]

2. Description of the Related Art In recent years, electrophotographic copying machines have become popular,
Digital copying machines have also been put into practical use, and it has become possible to easily obtain copied images of various originals. When copying is performed using these copying machines, as the original, a document that is printed out or copied using a facsimile or a copying machine is often used. However, a facsimile original or an original created by making multiple copies from a parent original may have deteriorated image quality of the original image itself. Therefore, if such a deteriorated original is copied, it is naturally copied. A part that is difficult to read is generated in the image, and a substantially unusable copy image is generated. Therefore, there is a need for a method for correcting characters that are difficult to read.

As conventional techniques related to this correction, for example, Japanese Patent Laid-Open Nos. 3-276380 and 4-329.
83 and Japanese Patent Laid-Open No. 4-112284.

The technique disclosed in Japanese Patent Laid-Open No. 3-276380 detects the line width of a character and makes the character line below the first predetermined width thick and the character line above the second predetermined width thin. By the processing, it is possible to prevent the character lines from being blurred or crushed.

The technique described in Japanese Patent Laid-Open No. 4-32983 mentioned above is intended to improve the reading accuracy by correcting the character data by combining the broken portions due to the blur with the average value of the width of the line segment element of the character. Is.

The technique disclosed in Japanese Unexamined Patent Publication No. 4-112284 detects a pair of contour points having a small angle between the contour points and connects the detected pair of contour points to each other to form a faint character. On the other hand, it enables recognition with high accuracy.

[0007]

By the way, when an original is copied by using a copying machine, a facsimile, a handy scanner, etc., a part of the characters of the copied image is flowed, overlapped, missing, or distorted. , "Flow part",
"Overlap part", "missing part", "distortion" etc. may appear. 4 and 1 which are explanatory views of the present invention.
2, the flow section will be described with reference to FIGS.
Means that some characters are columns (rows) as shown in Fig. 4 (b).
The "overlapping part" refers to the feature that extends in the direction shown in FIG.
As shown in FIG. 2 (b), it is a feature that a part of characters are overlapped in the column (row) direction and are discontinuous. Further, the “missing portion” means a feature that characters are partially divided and missing, resulting in discontinuity, as shown in FIG. 15B, and the “distortion” is shown in FIG. 19B. Characters that are distorted due to twisting, as shown.

However, although the above-mentioned correction method of the prior art is designed to correct the blurring or crushing of characters, some of the characters as described above may flow, overlap, drop, or become distorted. It's not supposed to fix crap. Therefore, there is a risk that a substantially unusable copy image may be generated that includes characters with defective features such as "flow part", "overlap part", "missing part", and "distortion". There is.

Therefore, according to the present invention, even if a part of a character flows, overlaps, is missing, or is distorted, the character of the copied image becomes unreadable by correcting the part. It is an object of the present invention to provide an image processing apparatus capable of avoiding such a situation and having a further excellent image quality correction function.

[0010]

In order to solve the above-mentioned problems, an image processing apparatus according to the present invention comprises a character detection means for detecting a character from an input image read by an image reading means, The character detected by the character detection means is different from other characters and is discriminated from the different discriminating means, and the characteristic discriminating means discriminates the character from being discriminated. A feature detecting means for extracting points, a correcting means for correcting the feature points detected by the feature detecting means, and a synthesizing means for synthesizing the characters corrected by the correcting means with the input image. It is characterized by having.

In order to solve the above problems, the image processing apparatus according to the second aspect of the present invention is the image processing apparatus according to the first aspect, wherein the difference determining means determines whether or not the character size is different. However, when the height or width of the character is larger than that of the other characters, the feature detecting means extracts a flow part where the same data continues or a missing part where the character is divided and separated as a feature point. Further, when the height or width of the character is smaller than that of the other characters, the overlapping portion where a part of the character overlaps is extracted as a feature point, and the correction means extracts the flow portion that is the extracted feature point. It is characterized in that deletion is performed to attract a portion displaced due to deletion, or interpolation is performed for a missing portion that is a characteristic point, or correction is performed to separate an overlap that is a characteristic point and interpolate a separated portion.

In order to solve the above-mentioned problems, an image processing apparatus according to a third aspect of the present invention is the image processing apparatus according to the first aspect, wherein the differentiating means is a straight line in which rows or columns of characters are twisted. It is determined whether or not there is an unintended portion, and the feature detection means extracts a portion in which the character is twisted and distorted in a portion which is not linear as the row or column of the character is twisted, and is extracted as the feature point, The feature is that the distortion, which is the extracted feature point, is smoothly corrected.

In order to solve the above-mentioned problems, an image processing apparatus according to a fourth aspect of the present invention is the image processing apparatus according to the above-mentioned first aspect, in which characters whose feature points are detected by the feature detecting means are read. It is characterized in that it is provided with a read determination means for determining whether or not it is possible, and when the read determination means determines that it is unreadable, the correction means carries out a correction.

[0014]

According to the structure of the present invention, first, the character detecting means detects a character from the input image read by the image reading means, and the differentiating means determines another character from the detected character. Different from and determine the condition. As the determination of the difference, for example, a configuration for determining whether or not the height or width of the detected character is different from other characters based on the character size (a configuration according to claim 2), or a line or column of the character is twisted. There is a configuration (configuration according to claim 3) for determining whether or not there is a non-linear portion.

When the difference discriminating means discriminates the difference, the characteristic detecting means discriminates the character discriminated by the difference discriminating means to extract the defective characteristic points of the character having the difference. To do. For example, if the height or width of the detected character is different from other characters and it is larger than other characters, the flow part where the same data continues, or the missing part where the character is divided and missing is extracted, When it is smaller than the other characters, the overlapping portion where a part of the characters overlap is extracted (configuration according to claim 2). On the other hand, when there is a portion where the rows or columns of the characters are twisted and not linear, the portion where the characters are twisted and distorted is extracted from that portion (configuration according to claim 3).

When the defective feature point is extracted in this way, the correction means corrects the extracted feature point. That is,
When the feature point is the flow part, the flow part is deleted to attract the shifted part, when the feature point is the missing part, the missing part is interpolated, and when the feature point is the overlapping part, the overlap Is separated and the part is interpolated (correction). If the characteristic point is a distortion, the distortion is corrected smoothly (the structure according to claim 3). Then, the synthesizing unit synthesizes the character corrected in this way with the input image.

Therefore, even if the manuscript has a defective feature point,
For example, even if a character including characters such as “flow part”, “overlap part”, “missing part”, and “distortion” is used,
This makes it possible to obtain a substantially usable copy image in which the defective feature points of the characters are corrected.

Further, according to a fourth aspect of the present invention, there is provided reading determination means for determining whether or not the character whose characteristic point is detected by the characteristic detection means is readable.
By performing the correction by the correction means only when it is determined by the reading determination means that the reading is impossible, there are some cases where the recognition is changed at the characteristic points caused by the character and there is no change. However, since it can be averaged and judged comprehensively, unnecessary correction processing is not performed even though it is readable, and the processing time is increased due to unnecessary processing and the cost is reduced. The increase can be prevented.

[0019]

【Example】

[Embodiment 1] One embodiment of the present invention will be described with reference to FIGS.
The description will be made based on 1.

As shown in FIG. 2, the digital copying machine as the image processing apparatus of this embodiment is provided with a document placing table 27 made of hard transparent glass at the upper end of the copying machine main body 26.

Below the document table 27, the lamp unit 1, the mirrors 2, 3, 4, the lens unit 5, the CCD (C
A scanner 22 which is a document reading unit having a large coupled device sensor 6 and the like is provided. In the scanner 22, an image of an original document (not shown) placed on the original document table 27 is illuminated by the lamp unit 1 and its reflected light is reflected by the mirrors 2, 3, 4 and the lens unit 5. The light is introduced as an electric signal by being guided to the light receiving surface of the CCD sensor 6 via.

A laser driver unit 7 is provided below the scanner 22. The original data taken in by the CCD sensor 6 is processed by an image processing unit (not shown) that performs image processing for copying, and then sent to the laser driver unit 7, which then receives the input original data. In response to this, laser light is emitted from, for example, a semiconductor laser inside.

The laser light emitted from the laser driver unit 7 is reflected by the mirrors 8 and 9 arranged on the optical path, and is irradiated on the photosensitive drum 10 rotating in the direction of arrow B, so that the photosensitive drum 10 is rotated. An electrostatic latent image is formed on. A charger 16 is provided around the photoconductor drum 10 to uniformly charge the surface of the photoconductor drum 10 to a predetermined potential prior to the exposure with the laser light. further,
A developing device 28 that supplies toner to the electrostatic latent image on the photoconductor drum 10 to develop the electrostatic latent image on the photoconductor drum 10 from the charger 16 in the rotating direction of the photoconductor drum 10, and the toner image on the photoconductor drum 10 is temporarily transferred. A transfer belt 17 to be transferred, a cleaning device 21 for removing the toner remaining on the surface of the photoconductor drum 10, a discharge lamp 15 for removing the residual potential of the photoconductor drum 10 prior to the next charging, and the like are arranged in this order. ing.

The developing device 28 is a black developing tank 1.
1, a yellow developing tank 12, a magenta developing tank 13, and a cyan developing tank 14, and the developing tanks 11 to 14 respectively store toners of the corresponding colors. The transfer belt 17 is endless and is provided so as to be movable in the arrow direction. A part of the transfer belt 17 is pressed against the photosensitive drum 10, so that the toner image on the photosensitive drum 10 is transferred.

A registration roller 19, which feeds a sheet to the transfer belt 17 at a predetermined timing, a sheet feeding cassette 20, and a manual sheet feeding section 23 are provided on the sheet feeding side of the transfer belt 17. These paper feed cassettes 20
In the vicinity of the manual paper feed unit 23, a paper feed roller 24 for carrying the paper, a carry roller 25, and the like are provided. Below the transfer belt 17, there is provided a transfer roller 18 that presses the paper sent from the registration roller 19 against the transfer belt 17 and transfers the toner image on the transfer belt 17 onto the paper.

On the paper output side of the transfer belt 17, a conveyor belt 30 for conveying the paper after the toner transfer, a fixing device 31 for heating and fixing the toner image on the paper, and a discharge roller 32 for discharging the paper after the fixing to the outside of the machine. Is provided.

In the above arrangement, color copy (3 col
or copy) is performed by the following operation procedure. First, when the charger 16 uniformly charges the surface of the photosensitive drum 10,
The scanner 22 performs the first scan.
At this time, the image processing unit first generates yellow data based on each data of the red component (R), the green component (G), and the blue component (B) of the original taken in by the CCD sensor 6. This yellow data is output as laser light from the laser driver unit 7, and this laser light exposes the surface of the photosensitive drum 10 to form a yellow electrostatic latent image on this exposed portion. Next, the yellow toner from the yellow developing tank 12 is supplied to the electrostatic latent image in the image area to form a toner image of the same color.

Next, the yellow toner image is transferred onto the transfer belt 17 which is pressed against the photosensitive drum 10. At this time, a part of the toner that does not contribute to the transfer remains on the surface of the photosensitive drum 10, but the cleaning device 21 scrapes off the remaining toner. Further, the charge eliminating lamp 15 removes the residual electric charge on the surface of the photoconductor drum 10.

When the above process is completed, the charger 16 charges the surface of the photosensitive drum 10 uniformly again, and the scanner 2
2 performs the second scan. From the original data thus obtained, magenta data is obtained by the image processing section, and the laser driver unit 7 irradiates the photosensitive drum 10 with laser light to form a magenta electrostatic latent image. Then, magenta toner is supplied from the magenta developing tank 13 and a toner image of the same color is formed on the photosensitive drum 10. Then, this magenta toner image is transferred to the transfer belt 17.
Then, the images are transferred onto the yellow toner image and the images are superimposed. After that, the cleaning device 21 and the static elimination lamp 1
When the same processing as the above described by 5 is performed, the charger 16
The electrostatic latent image of cyan is formed on the photosensitive drum 10 by uniformly charging the photosensitive drum 10 and performing the third scanning by the scanner 22 and irradiating cyan data as laser light. . Then, cyan toner is supplied from the cyan developing tank 14 to the photosensitive drum 10 to form a toner image of the same color, and this cyan toner image is transferred onto the magenta toner image on the transfer belt 17 to finally obtain the image. The images are superposed.

After that, the toner image on the transfer belt 17 on which the images are superposed is transferred onto the sheet by the transfer roller 18, and the toner image is heated and fixed by the fixing device 31, and then the sheet is ejected from the apparatus by the discharge roller 32. Is discharged.

The above-mentioned process is a process for three-color, and in the case of a four-color process,
The process using the black toner stored in the black developing tank 11 is added to the above process. On the other hand, black-and-white copying is performed by supplying black toner from the black developing tank 11 to the electrostatic latent image on the photoconductor drum 10 and transferring the toner image onto a sheet via the transfer belt 17.

By the way, when a manuscript is copied by using a copying machine, the characters of the manuscript should be normal characters as shown in FIG. 4 (a), but those shown in FIG. 4 (b). ,
In some cases, a part of the character has a "flow part" extending in the column (row) direction. Documents having such characters with a "flow part" are often facsimiles, manuscripts that move on a document table, or documents created using a handy scanner.

That is, as shown in FIG.
The document 02 is configured so that the document P is sent by a pair of conveying rollers 100 and 100, and data is read by a reading unit 101 such as a fixed CCD. Therefore, if the document P slips during document transport or if the transport rollers 100, 100 partially idle, the document P is not fed, and the reading unit 101 repeatedly scans the same position, resulting in a "flow section". .

Further, as shown in FIG. 6, the manuscript table moving type copying machine 103 includes a manuscript table 1 provided on the upper surface of the main body 107.
04 is reciprocated by the driving roller 105 and the driving gear 106 in the main body 107, and the reading of data is performed by the fixed reading unit 108 as in the case of the facsimile 102. Therefore, when a document pressing cover (not shown) is manually pressed during scanning of a document, if a force is applied in the opposite direction of scanning or the document table 104 hits an obstacle, the drive roller 105 and the drive gear 106 will idle. Then, the document table 104 does not move, and the reading unit 108 repeatedly scans the same position, resulting in a "flow section".

Further, as shown in FIG. 7, the handy scanner 109 is provided with a reading section 110 composed of a CCD inside, and a scanning roller 11 capable of scanning the original P.
The reading unit 110 reads the data while scanning the original P. Therefore, the scanning roller 111 hits an obstacle during reading or partially idles to cause CC
If the D unit 112 is not sent, the same position will be scanned and a "flow part" will occur.

Therefore, the image processing control unit 40 shown in FIG. 1 provided in the digital copying machine of this embodiment corrects the characters having such "flowing portions" to correct the image quality. The function for carrying out is provided.

In the image processing control section 40, a plurality of control sections such as a document reading unit 41 and a laser driver unit 7 are controlled by a main CPU (Central Processing Unit) 43. ROM (Read Only Memor
y) 44 is an area for storing a control program,
A RAM (Random Access Memory) 45 is used as a working area for storing parameters peculiar to the copying machine and executing programs.

The document reading unit 41 includes the scanner 22 and reads the image of the document set on the document table 27 as described above. The read input image is subjected to image processing such as changing the image density and the image magnification according to the user's request in the image processing unit 46. The input image processed by the image processing unit 46 is sent to the laser driver unit 7. Then, it is output as laser light controlled by the laser driver unit 7. The operation panel control unit 47, as shown in FIG.
A control unit that controls an operation panel 48 provided on the upper surface of the copying machine body 26. The operation panel 48 includes a ten-key pad 49 for inputting the number of copies and the like, an image quality correction mode setting key 53, and a liquid crystal display panel 50 for displaying the number of copies and the like and displaying a message to the operator. .

Further, in order to carry out the above-mentioned image quality correction mode, the character detecting unit 51 as the character detecting unit shown in FIG. 1, the character recognizing unit 52, the different determining unit 58 as the different discriminating unit, and the characteristic detecting unit. A feature detection unit 56, a read determination unit 55 as a read determination unit, a character correction processing unit 54 as a correction unit, a combination processing unit 57 as a combination unit, and the like are provided.

The character detecting section 51 detects whether or not there is a character from the image data of the input image processed by the image processing section 46. There are two methods for character detection: one is to extract the characteristics of characters and the other is to use character recognition. The former is a unit of graphics that connects image data due to features such as "characters are small, independent, and are arranged in a certain line". It is determined based on whether the figure matches the character size.

The difference discriminating section 58 discriminates the character detected by the character detecting section 51 from the other characters. To determine the difference, first, the outer shape rectangle size of the character detected by the character detecting unit 51 is detected, and the matrix of the document is set based on the arrangement of the outer shape rectangle sizes. And then detect the width of each matrix, the longer rows compared to others,
Alternatively, it is determined by determining whether or not there is a long line.

The feature detecting section 56 is provided with the difference determining section 58.
From the character data determined to be different from each other, the "flow part" where the same image data continues is detected as a feature point.

The character recognizing section 52 performs character recognition according to the character recognition method. The reading determination unit 55 determines, by using character recognition, whether or not the character for which the "flow part" is detected by the feature detection unit 56 can be read. Whether or not it is readable is determined by the ratio of the total number of characters in the character string in which the "flow part" is detected and the number of characters recognized by the character recognition unit 52. The purpose of this is to make a comprehensive judgment by averaging the "flow part" that occurs depending on the character, and the one that does not change.

The character correction processing unit 54 performs a correction process on a character having a "flow part" which is determined to be unreadable. A specific correction method will be described later.

The combining processing unit 57 combines the characters corrected by the character correction processing unit 54 with the image data.

Next, the processing procedure when the image quality correction mode is set in the digital copying machine having the above configuration will be described with reference to FIGS. 8 and 9 with reference to the flowchart of FIG.

In the image quality correction here, a character is detected from the input image, a "flow part" is detected from the character, it is determined whether the character is readable, and if it is determined that the character cannot be read, " A character having a "flow part" is corrected and the corrected character is output.

First, the image quality correction mode is set by the image quality correction mode setting key 53 of the operation panel 48 (S
1). Next, the document is scanned by the scanner 22 and the image data of the input image is stored in the RAM 45 (S2).

Next, the character detection section 51 detects characters whether or not there are characters in the input image (S3). It is possible to detect whether or not there is a character in the input image by a method of extracting a characteristic of the character or a method of using character recognition. The method of extracting the character features described above is performed by separating the input image into connected graphic units according to the features such as "small characters", "characters are isolated", and "lined up in a line". This is a method of detecting whether or not the figure matches the character size, and then comparing the separated figure units with the numeral pattern to detect the character.

When a character is detected in S3, the difference discriminating unit 58 discriminates the degree of difference from other characters.
First, as shown in FIG. 9, the outline rectangle size of the character is detected (S4), and then the matrix of the document is set based on the arrangement of the outline rectangle sizes (S5). Then, the width of each matrix is detected (S6), and it is determined whether or not there is a long row or column as compared with others (S7). If it is determined that the same data continues in the longitudinal direction. It is determined whether or not (S8). Here, if it is judged to continue, "flow part"
Has been detected.

When the "flow part" is detected, the character recognition part 5
Characters are recognized in 2 (S9). Then, based on the ratio of the total number of character strings in which the "flow part" is detected and the number of characters recognized by the character recognition unit 52, the read determination unit 55 determines whether or not the data can be read (S10).

If it is determined in S10 that the character cannot be read, the character having the "flow part" is corrected by the character correction processing unit 54. That is, as shown in FIG. 8, the shaded area having the “flow part” is deleted so that the part where the same data continues becomes the same as other rows or columns, and the position is displaced by the “flow part”. The moved portion is moved to combine and correct the image data (S11 and S12). And
The corrected character data is combined with the image data by the combining processing unit 57.

Thereafter, when a copy button (not shown) on the operation panel 48 is pressed (S14), a copying operation is performed (S15), and the process is completed.

On the other hand, no character is detected in S3, it is determined in S7 that there is no long line or column, there is no portion where the same data continues in S8, or the reading is performed in S10. If it is determined that it is possible, the image quality correction mode is released (S13), and the normal copying operation is performed (S14 and S15).

As described above, in the digital copying machine of the present embodiment, the "flow part" of the characters of the original image is detected and the correction processing is performed, so that even if the original has the "flow part". Even if a character including the above-mentioned character is used, it is possible to obtain a substantially usable copy image in which the character having the “flow portion” is corrected.

That is, for example, in the case of the facsimile 102, when the document P is transported, the document P slips, or the transport rollers 100, 100 for transporting the document P partially idle, so that the document P is not fed and the reading unit is not fed. Since 101 may repeatedly scan the same position, a "flow portion" may be generated in the character. Due to this, it is possible to correct the image quality deterioration due to the abnormality of the conveyance rollers 100 of the facsimile 102. Is possible (see FIG. 5).

In the case of the movable copyboard 103 of the document table, during scanning of the document, a force is applied in the opposite direction of the scanning, or the document table 104 hits an obstacle, and the reading unit 10
8 may repeatedly scan the same position, resulting in a "flow portion" in the character. Due to this, the image quality is deteriorated due to such an abnormality during document scanning of the platen moving type copying machine 103. Can be corrected (see FIG. 6).

When the document P is read by using the handy scanner 109, the scanning roller 111 which sends the CCD unit 112 hits an obstacle, or the CCD unit 112 is not fed due to partial idling and the reading unit 110 is not sent. Although the same position may be scanned, a "flow part" is generated in the character, but by using the configuration of the present invention, the deterioration of the image quality due to the abnormality during the scanning of the handy scanner 109 is corrected. It becomes possible.

Further, in order to output data read by a facsimile or a handy scanner as an image, for example, FIG.
As shown in, a handy scanner or a facsimile is connected to the copying machine main body 26 'through a communication line 119,
Image data read by a handy scanner or a facsimile is transferred to the level conversion unit 120 and the serial i / F 121.
In the case of a configuration in which the same data is sent to the image processing unit 46 via the, the same data may be duplicated due to noise interference in communication. Then, a "flow part" is generated in the character, which makes it possible to correct the deterioration in image quality due to such a communication noise obstacle. Further, even if the original read by the handy scanner or facsimile includes characters with "flow part", it is of course possible to correct it to obtain a practically usable copy image.

Further, in the case of the present embodiment, in S10, the reading determination unit 55 determines the characteristics based on the ratio between the total number of characters of the character string in which the "flow part" is detected and the number of characters recognized by the character recognition unit 52. The character correction processing unit 54 determines whether the character whose "flow part" is detected by the detection unit 56 is readable, and only when the read determination unit 55 determines that the character is unreadable. It is supposed to carry out the modification. Therefore, depending on the character, there are some things that change the recognition in the generated "flow part" and some that do not change, but since it can be averaged and comprehensively judged, unnecessary correction processing is applied. It is possible to prevent an increase in processing time due to useless processing and an increase in cost.

[Embodiment 2] Another embodiment of the present invention is shown in FIG.
The following is a description with reference to FIGS. 12 to 14. For convenience of explanation, members having the same functions as the members shown in the drawings of the first embodiment will be designated by the same reference numerals, and the description thereof will be omitted.

The image processing control unit 40 shown in FIG. 1 of the digital copying machine of this embodiment has a function for carrying out an image quality correction mode for correcting the "overlap portion" of the characters in the document to correct the image quality. It is equipped.

The "overlapping part" is supposed to be a normal character as shown in FIG. 12 (a), but as shown in FIG. 12 (b), a part of the character is in the column (row) direction. Characteristic part that overlaps with and has discontinuity. Such an "overlap"
Are often printed out from a facsimile, originals copied by a platen moving type copying machine, and originals scanned by a handy scanner.

In the case of a facsimile, the transfer paper is carried by the carrying roller, and the data is written on the transfer paper using the fixed writing section. Therefore, when writing data, if the transfer paper slips or the transport roller partially idles and the transfer paper is not fed, the writing unit may repeatedly write data at the same position. Occurs. That is, it is a state in which the character is shortened in the row (column) direction than the normal character and recorded.

Further, the above-mentioned original table moving type copying machine 10
In the case of No. 3, when the document pressing cover is manually pressed during document scanning, a force is applied in the scanning direction, or the document table 104 hits an obstacle, and an excessive force is applied to the document table 104 to drive the drive roller 105. Also, the drive gear 106 may idle, and the reading unit 108 may skip a certain position of the document to perform scanning (see FIG. 6).

In the case of the handy scanner 109, the scanning roller 111 is partially locked and C
The CD unit 112 may be sent and may be skipped at an arbitrary position for scanning, which results (see FIG. 7).

Therefore, the difference discriminating unit 58 of the image processing control unit 40 first detects the outer rectangle size of the character detected by the character detecting unit 51, and sets the matrix of the document from the arrangement of the outer rectangle sizes. To do. Then, the width of each matrix is detected, and it is determined whether or not there is a short column or a short column in comparison with the other, and the different condition is determined.

The feature detecting section 56 is provided with the difference determining section 58.
From the character data that is determined to be different, the feature point is to detect the part where the pixel data is discontinuous with an arbitrary line or pixel row as a boundary, and to detect the "overlap part". There is.

The reading determination section 55 is the feature detection section 5 described above.
In the same manner as in the first embodiment, the character recognition is used to determine whether or not the character in which the "overlapping part" is detected can be read.

The character correction processing section 54 includes a reading determination section 5
At 5, the characters having the "overlapping part" which are determined to be unreadable are subjected to a correction process as described later.

The processing procedure when the image quality correction mode of the digital copying machine of this embodiment is set will be described with reference to the flowchart of FIG. 14 and with reference to FIG.

In the image quality correction here, a character is detected from an original image, an "overlapping part" is detected from the character, it is determined whether or not the character is readable, and if it is determined that the character cannot be read, " A character having an “overlap portion” is corrected, and the corrected character is output.

First, the image quality correction mode is set by the image quality correction mode setting key 53 on the operation panel 48 (S
1). Next, the document is scanned by the scanner 22 and the image data of the input image is stored in the RAM 45 (S2).

Next, the character detection section 51 detects characters whether or not there are characters in the input image (S3). It is possible to detect whether or not there is a character in the input image by a method of extracting a characteristic of the character or a method of using character recognition.

When a character is detected in S3, the outer rectangle size of the character is detected (S4), and then the original matrix is set from the arrangement of the outer rectangle sizes (S).
5). Then, the width of each matrix is detected (S6), and it is determined whether or not there is a short row or column as compared with others (S1).
7) If it is determined that there is, it is determined whether or not there is a portion where the image data is discontinuous at an arbitrary line or a pixel column in that portion (S18). Here, if it is determined that there is an “overlapping portion”, it means that the “overlapping portion” has been detected.

When the "overlap portion" is detected, the character recognition unit 52 recognizes the character (S9). Then, based on the ratio between the total number of characters of the character string in which the "overlapping part" is detected and the number of characters recognized by the character recognition unit 52, the reading determination unit 55 determines whether the reading is possible (S10).

When it is determined in S10 that the character cannot be read, the character having the "overlap portion" is corrected by the character correction processing unit 54. That is, as shown in FIG. 13, a portion where the image data is discontinuous with an arbitrary line as a boundary is separated so as to have the same height as other portions, and a missing portion is corrected so that the image data is smoothly continuous ( It is interpolated) and corrected (S19, S20). Then, the corrected character data is combined with the image data by the combining processing unit 57.

Thereafter, when a copy button (not shown) on the operation panel 48 is pressed (S14), a copying operation is performed (S15), and the process is completed.

On the other hand, no character is detected in S3, it is determined in S17 that there is no short row or column, and in S18, the image data is discontinuous at any line or pixel column. It is judged that there is no part that becomes,
Alternatively, when it is determined in S10 that the image is readable, the image quality correction mode is canceled (S13), and the normal copying operation is performed (S14 and S15).

As described above, in the digital copying machine of this embodiment, the "overlapping portion" of the characters of the original image is detected and the correction processing is performed, so that even if the original has an "overlapping portion". Even if a character including the above-mentioned character is used, it is possible to obtain a substantially usable copy image in which the character having the “overlap portion” is corrected.

That is, in the case of a facsimile, for example, when the transfer sheet is conveyed, the transfer sheet slips, or the conveying roller for conveying the transfer sheet partially idles, so that the transfer sheet is not sent and the data is transferred to the same position. Writing may occur, and an "overlapping portion" may occur in the character, which makes it possible to correct the image quality deterioration due to such an abnormality of the conveyance roller of the facsimile.

Further, in the case of the original table moving type copying machine 103, a force is applied in the scanning direction during the original scanning, and the original table 1 is moved.
04 may hit an obstacle or force may be applied to skip the position where the reading unit 108 is present, and may cause an “overlapping part” to occur in the character. It is possible to correct the deterioration in image quality due to an abnormality during document scanning of the copying machine 103 (see FIG. 6).

When the manuscript P is read by using the handy scanner 109, the scanning roller 111 sending the CCD unit 112 may partially idle to skip an arbitrary part on the manuscript and scan it. However, this makes it possible to correct the deterioration in image quality due to such an abnormality during scanning of the handy scanner 109 (see FIG. 7).

Further, in order to output the data read by the facsimile or the handy scanner as an image, the copying machine main body 2
6'is connected to a handy scanner or a facsimile via a communication line 119, and the image data read by the handy scanner or the facsimile is transferred to the level conversion unit 12.
0, in the case of a configuration in which the data is sent to the image processing unit 46 via the serial i / F 121, arbitrary data may be lost due to a communication noise failure. Then, the letters "overlap"
However, this makes it possible to correct the deterioration in image quality due to such communication noise interference. Further, even when a document read by a handy scanner or a facsimile includes a character having an "overlap portion", it is of course possible to correct it to obtain a practically usable copy image (see FIG. 11).

Further, as in the first embodiment, S10
At the reading determination unit 55, the feature detection unit 56 detects the “overlap portion” from the ratio between the total number of characters of the character string in which the “overlap portion” is detected and the number of characters recognized by the character recognition unit 52. It is determined whether or not the read character is readable, and only when the read determination unit 55 determines that the read character is unreadable, the character correction processing unit 54 performs the correction. Therefore, depending on the characters, there may be some "overlapping parts" that change recognition and others that do not change, but since they can be averaged to make a comprehensive judgment, unnecessary correction processing will be applied. It is possible to prevent an increase in processing time due to useless processing and an increase in cost.

[Embodiment 3] Another embodiment of the present invention is shown in FIG.
The description will be made with reference to FIGS. 15 to 18. For convenience of explanation, members having the same functions as the members shown in the drawings of the first embodiment will be designated by the same reference numerals, and the description thereof will be omitted.

The image processing controller 40 shown in FIG. 1 of the digital copying machine of the present embodiment has a function for carrying out an image quality correction mode for correcting the "missing part" in the characters in the document to correct the image quality. It is equipped.

The "missing portion" is supposed to be a normal character as shown in FIG. 15 (a), but as shown in FIG. 15 (b), the character is partially divided and separated. , A discontinuous characteristic part. Such a "missing part" is
This is likely to occur in a facsimile machine or a copying machine in which transfer paper is conveyed by a conveyance roller. That is, when the transfer sheet S is conveyed by the conveying rollers, wrinkles may occur on the transfer sheet S during conveyance as shown in FIG. 18. As a result of characters being transferred onto the wrinkles, when the wrinkles are extended. The character will be divided into, and a "missing part" will occur.

Therefore, the difference determination unit 58 of the image processing control unit 40 first detects the outer shape rectangle size of the character detected by the character detection unit 51, and sets the matrix of the document based on the arrangement of the outer shape rectangle sizes. To do. Then, the width of each matrix is detected, and it is determined whether or not there is a long column or a long column as compared with the others, and the difference is determined.

The feature detecting section 56 has a different discriminating section 58.
From the character data determined to be different from each other, a "missing portion" is detected as a feature point by detecting a portion where the image data is lost and becomes discontinuous.

The reading determination section 55 is the feature detection section 5 described above.
In the same manner as in the first embodiment, the character recognition is used to determine whether or not the character in which the "missing part" is detected can be read.

The character correction processing section 54 includes a reading determination section 5
At 5, the character having the "missing part" which is determined to be unreadable is subjected to a correction process as described later.

The processing procedure when the image quality correction mode of the digital copying machine of this embodiment is set will be described with reference to FIG. 16 while referring to the flowchart of FIG.

The image quality correction here is performed by detecting a character from an original image, detecting a "missing portion" from the character in the image, determining whether the character is readable, and determining that the character cannot be read. , A character having a “missing part” is corrected, and the corrected character is output.

First, the image quality correction mode is set by the image quality correction mode setting key 53 on the operation panel 48 (S
1). Next, the document is scanned by the scanner 22 and the image data of the input image is stored in the RAM 45 (S2).

Then, character detection is performed by the character detection unit 51 as to whether or not there are characters in the input image (S3). The detection of whether or not there is a character in the input image can be performed by a method of extracting the characteristic of the character or a method of using character recognition.

When a character is detected in S3, the outer rectangle size of the character is detected (S4), and then the original matrix is set from the arrangement of the outer rectangle sizes (S).
5). Then, the width of each matrix is detected (S6), and it is determined whether or not there is a long row or column as compared with others (S7),
If it is determined that there is a portion where the image data is missing and discontinuous in that portion, it is determined (S).
21). Here, if it is determined that there is a “missing portion”, it means that the “missing portion” has been detected.

When the "missing part" is detected, the character recognition part 5
Characters are recognized in 2 (S9). Then, based on the ratio between the total number of characters in the character string in which the "missing part" is detected and the number of characters recognized by the character recognition unit 52, the read determination unit 55 determines whether or not it is readable (S10).

If it is determined in S10 that the character cannot be read, the character having the "missing portion" is corrected by the character correction processing unit 54. That is, as shown in FIG.
The missing and discontinuous portions of the image data are attracted so as to have the same height as the others, and are corrected (interpolated) and corrected to be smoothly continuous (S22 and S23). And
The corrected character data is combined with the image data by the combining processing unit 57.

Thereafter, when a copy button (not shown) on the operation panel 48 is pressed (S14), a copying operation is performed (S15), and the process is completed.

On the other hand, no character is detected in S3, it is determined in S7 that there is no long line or column, and in S21 that there is no discontinuous portion due to missing image data. Alternatively, if it is determined in S10 that the image is readable, the image quality correction mode is canceled (S13), and a normal copying operation is performed (S14.
S15).

As described above, in the digital copying machine of the present embodiment, the "missing portion" of the character of the original image is detected and the correction processing is performed, so that even if the original is
Even if a character including a "missing part" is used, a substantially usable copy image in which the character having the "missing part" is corrected can be obtained.

That is, for example, in a facsimile machine or a copying machine in which a transfer roller is used to transfer a transfer sheet, wrinkles are likely to occur during transfer sheet transfer, and characters are output on the wrinkles. , "Missing part" in the characters of the copied image
However, this makes it possible to correct the deterioration in image quality due to such wrinkles on the transfer sheet due to the transport roller.

In order to output an image of data read by a facsimile or a handy scanner, for example, a handy scanner or a facsimile is connected to the copying machine main body 26 'via a communication line 119 and read by the handy scanner or the facsimile. The image processing unit 4 converts the image data through the level conversion unit 120 and the serial i / F 121.
Even in the case of the configuration for sending to 6, the original image read by the handy scanner or facsimile includes characters with a "missing portion", of course, it can be corrected to obtain a practically usable copy image ( (See FIG. 11).

Further, as in the first embodiment, S10
At the reading determination unit 55, the feature detection unit 56 detects the “missing portion” from the ratio between the total number of characters of the character string in which the “missing portion” is detected and the number of characters recognized by the character recognition unit 52. It is determined whether or not the read character is readable, and only when the read determination unit 55 determines that the read character is unreadable, the character correction processing unit 54 performs the correction. Therefore, depending on the character, there are some "missing parts" that occur and some of which have different recognition,
Since it can be averaged and judged comprehensively,
There is no need for unnecessary correction processing,
It is possible to prevent an increase in processing time due to useless processing and an increase in cost.

[Embodiment 4] Another embodiment of the present invention is shown in FIG.
The following is a description with reference to FIGS. 19 to 22. For convenience of explanation, members having the same functions as the members shown in the drawings of the first embodiment will be designated by the same reference numerals, and the description thereof will be omitted.

The image processing controller 40 shown in FIG. 1 of the digital copying machine of the present embodiment has a function for carrying out an image quality correction mode for correcting "distortion" in characters in a document to correct the image quality. Has been.

The "distortion" means a characteristic portion where a character that should normally be a normal character as shown in FIG. 19 (a) is distorted by twisting as shown in FIG. 19 (b). . Such a "distortion" is as shown in FIG.
This is likely to occur when manipulating a document with the handy scanner 119. In other words, when the balance of the force applied to the CCD unit 112 is lost, the curve R may be drawn and twisted to scan. Then, the characters are twisted and “distorted”.
Will occur.

Therefore, the difference discriminating unit 58 of the image processing control unit 40 first detects the outer shape rectangle size of the character detected by the character detecting unit 51, and sets the matrix of the original from the arrangement of the outer shape rectangle sizes. To do. Then, it is determined whether or not each matrix is twisted and there is a non-linear portion, and the different states are determined.

The feature detecting section 56 includes the difference determining section 58.
The "distortion", which is the portion where the image data is distorted, is detected as a feature point from the character data determined to be different. That is, in the case of "distortion", a point at which the curvature of the curve changes by a certain value or more, a so-called inflection point, occurs in a part of the curve, and "distortion" can be detected depending on the presence or absence of this inflection point. Although there is an inflection point in a character, such an inflection point has a pattern stored in advance in order to recognize the character, and an inflection point which is not in the pattern of the unrecognizable character may be detected.

The read determination unit 55 is the feature detection unit 5 described above.
In the same manner as in the first embodiment, it is determined whether or not the character in which "distortion" is detected in 6 can be read by using character recognition.

The character correction processing unit 54 includes a read determination unit 5
At 5, the character having the “distortion” determined to be unreadable is subjected to a correction process as described later.

The processing procedure when the image quality correction mode of the digital copying machine of this embodiment is set will be described with reference to FIG. 20 while referring to the flowchart of FIG.

The image quality correction here is performed by detecting a character from the original image, detecting "distortion" from the character, determining whether the character is readable, and determining "readable" when the character is not readable. Is corrected, and the corrected character is output.

First, the image quality correction mode is set by the image quality correction mode setting key 53 on the operation panel 48 (S
1). Next, the document is scanned by the scanner 22 and the image data of the input image is stored in the RAM 45 (S2).

Then, character detection is performed by the character detection unit 51 as to whether or not there are characters in the input image (S3). The detection of whether or not there is a character in the input image can be performed by a method of extracting the characteristic of the character or a method of using character recognition.

When a character is detected in S3, the outer rectangle size of the character is detected (S4), and then the original matrix is set from the arrangement of the outer rectangle sizes (S4).
5). Then, the twisted non-linear portion of each matrix is detected (S24), and when the twisted portion is detected, the portion in which the character data is distorted is detected (S24).
25). This means that "distortion" has been detected.

When "distortion" is detected, the character recognition unit 5
Characters are recognized in 2 (S9). Then, based on the ratio between the total number of characters in the character string in which "distortion" is detected and the number of characters recognized by the character recognition unit 52, the read determination unit 55 determines whether or not it is readable (S10).

If it is determined in S10 that the character cannot be read, the character having the "distortion" is corrected by the character correction processing unit 54. That is, as shown in FIG. 20, the distorted portions are rotated about arbitrary points (A) and (B) so that the distorted portions are lined up linearly like the others, and the character data of the distorted portions are smoothed. It is corrected and corrected so as to be linearly continuous (S26 and S27). Then, the corrected character data is combined with the image data by the combining processing unit 57.

After that, when a copy button (not shown) on the operation panel 48 is pressed (S14), a copying operation is performed (S15), and the process ends.

On the other hand, no character is detected in S3, the twisted portion of the matrix is not detected in S24, and the distorted portion is not detected in S25.
Alternatively, when it is determined in S10 that the image is readable, the image quality correction mode is canceled (S13), and the normal copying operation is performed (S14 and S15).

As described above, in the digital copying machine of the present embodiment, the character "distortion" of the original image is detected and the correction processing is performed. Even if something containing
It is possible to obtain a substantially usable copy image in which characters having "distortion" are corrected.

That is, for example, when scanning a document with a handy scanner to read data, the balance of the force for moving the scanner is lost, or the scanner is twisted and scanned on the document due to partial idling of the left and right scanning rollers. In that case, "distortion" occurs in the characters, which makes it possible to correct the deterioration in image quality due to such an abnormality during scanning of the handy scanner.

In order to output an image of data read by a facsimile or a handy scanner, for example, a handy scanner or a facsimile is connected to the copying machine main body 26 'via a communication line 119 and read by the handy scanner or the facsimile. The image processing unit 4 converts the image data through the level conversion unit 120 and the serial i / F 121.
Even in the case of the configuration of sending to 6, the case where the original read by the handy scanner or the facsimile contains characters with "distortion" and the case where the handy scanner is twisted during scanning, of course, it is corrected and substantially corrected. It is possible to obtain a copy image that can be used physically (see FIG. 11).

Further, as in the first embodiment, in S10, the reading determination unit 55 calculates the ratio between the total number of characters in the character string in which "distortion" is detected and the number of characters recognized by the character recognition unit 52. The character detection processing unit 54 determines whether the character in which "distortion" is detected by the feature detection unit 56 is readable, and only when the read determination unit 55 determines that the character is unreadable. It is supposed to carry out the modification. Therefore, depending on the character, there are some "distortions" that occur and some of which have different perceptions.
Since it can be averaged and judged comprehensively,
There is no need for unnecessary correction processing,
It is possible to prevent an increase in processing time due to useless processing and an increase in cost.

In the description of this embodiment, the "flow part", the "overlap part", the "missing part", and the "distortion" and the characteristic parts to be corrected are separately described. It is of course possible to equip one image processing device with the configurations of Examples 1 to 4, and any one of the "flow part", "overlap part", "missing part", and "distortion" can be provided. It is possible to obtain a usable copy image without forming an unusable copy image even on a document including characters having a characteristic portion, and to obtain an image processing apparatus having an excellent image quality correction function. it can.

[0127]

According to the image processing apparatus of the invention described in claim 1,
As described above, the character detection unit that detects a character from the input image read by the image reading unit and the difference determination that determines the difference between the character detected by the character detection unit and another character Means and the feature detecting means for extracting the defective feature points of the different characters by determining the difference between the characters by the difference determining means, and correcting the feature points detected by the feature detecting means. The configuration includes a correction unit that performs the correction and a combining unit that combines the characters corrected by the correction unit with the input image.

As described above, the image processing apparatus according to the second aspect of the present invention is the image processing apparatus according to the first aspect, wherein the difference determining means determines whether or not the character size is different, and When the height or width of the character is larger than that of other characters, the detecting means extracts a flow portion where the same data continues or a missing portion where the character is divided and separated as a feature point, and When the height or width of the character is smaller than other characters, the overlapping part where a part of the character overlaps is extracted as a feature point, and the correction means deletes and deletes the flow part that is the extracted feature point. The correction is performed by attracting a portion displaced by, or interpolating a missing portion which is a characteristic point, or separating an overlap which is a characteristic point and interpolating the separated portion.

As described above, the image processing apparatus according to the third aspect of the present invention is the image processing apparatus according to the first aspect, wherein the difference determining means has a non-linear portion where the rows or columns of characters are twisted. It is determined whether or not there is, and the feature detecting means extracts a portion in which a character is twisted and distorted in a non-linear portion where a row or a column of the character is twisted as a feature point, and the correction means is extracted. It is a configuration that smoothly corrects distortion, which is a feature point.

Therefore, even if the manuscript contains characters having defective feature points, for example, "flow part", "overlap part", "missing part", "distortion", etc.,
As a result, it is possible to obtain a substantially usable copy image in which the defective feature points of the characters are corrected.

As described above, the image processing apparatus according to the present invention can read the character of which the characteristic point is detected by the characteristic detecting means in the image processing apparatus according to the first aspect. The reading determination means for determining whether or not the correction is performed is provided, and when the reading determination means determines that the reading is impossible, the correction means performs the correction.

Therefore, depending on the characters, there are some characteristic points that change recognition and others that do not change. However, since it can be averaged and comprehensively judged, it is not possible to read it even though it is readable. It is possible to prevent an unnecessary correction process from being performed, and it is possible to prevent an increase in processing time due to useless processing and an increase in cost.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an image processing control unit included in a digital copying machine common to each embodiment of the present invention.

FIG. 2 is a front view showing the structure of the digital copying machine.

FIG. 3 is an explanatory diagram of an operation panel included in the digital copying machine.

4A and 4B are explanatory diagrams showing characteristics of characters to be corrected by the digital copying machine according to the first embodiment. FIG. 4A is a normal character, and FIG. 4B is a “flow part” to be corrected. Indicates the character that was written.

FIG. 5 is an explanatory diagram for explaining reading of an original document of a facsimile for showing a cause of occurrence of a “flow part” in a character.

FIG. 6 is an explanatory diagram for explaining reading of a document of a document table moving type copying machine, for showing the cause of occurrence of “flow part” and “overlap part” in characters.

FIG. 7 is an explanatory diagram for explaining reading of a document by a handy scanner, which is for showing the cause of occurrence of “flow part” and “overlap part” in characters.

FIG. 8 is an explanatory diagram showing a correction process of a character having a “flow part”.

FIG. 9 is an explanatory diagram showing a state of detecting a “flow part”.

FIG. 10 is a flowchart showing a processing procedure when the image quality correction mode is set in the digital copying machine of the first embodiment.

FIG. 11 is an explanatory diagram showing a digital copying machine, which is common to each embodiment of the present invention, in which an input image is sent from a handy scanner or a facsimile via a communication line.

12A and 12B are explanatory diagrams showing characteristics of a character to be corrected by the digital copying machine according to the second embodiment. FIG. 12A is a normal character, and FIG. 12B is an “overlapping part” to be corrected. Indicates the character that was written.

FIG. 13 is an explanatory diagram showing a correction process of a character having an “overlapping part”.

FIG. 14 is a flowchart illustrating a processing procedure when the image quality correction mode is set in the digital copying machine of the second embodiment.

15A and 15B are explanatory diagrams showing characteristics of characters to be corrected by the digital copying machine according to the third embodiment. FIG. 15A is a normal character, and FIG. 15B has a “missing part” to be corrected. Indicates the character that was written.

FIG. 16 is an explanatory diagram showing a process of correcting a character having a “missing part”.

FIG. 17 is a flowchart illustrating a processing procedure when the image quality correction mode is set in the digital copying machine of the third embodiment.

FIG. 18 is an explanatory diagram illustrating a cause of occurrence of a “missing part” in a character.

19A and 19B are explanatory diagrams showing characteristics of characters to be corrected by the digital copying machine according to the fourth embodiment. FIG. 19A is a normal character, and FIG. 19B has a "distortion" to be corrected. Indicates a character.

FIG. 20 is an explanatory diagram showing a correction process of a character having “distortion”.

FIG. 21 is a flowchart illustrating a processing procedure when the image quality correction mode is set in the digital copying machine of the fourth embodiment.

FIG. 22 is an explanatory diagram illustrating a cause of occurrence of “distortion” in a character.

[Explanation of symbols]

 Reference numeral 22 Scanner (image reading means) 40 Image processing control section 45 RAM 47 Operation panel control section 51 Character detection section (character detection section) 52 Character recognition section 54 Character correction processing section (correction section) 55 Reading determination section (read determination section) 56 Feature Detecting Section (Feature Detecting Means) 57 Compositing Processing Section (Combining Means) 58 Difference Discriminating Section (Difference Discriminating Means)

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Office reference number FI Technical display location G06K 9/36 H04N 1/40

Claims (4)

[Claims] 1. A character detection means for detecting a character from an input image read by an image reading means, and a difference discrimination for discriminating the difference between another character and the character detected by the character detection means. Means and a feature detecting means for extracting a defective feature point of a character having a difference by discriminating the difference between the characters by the difference discriminating means, and correcting the feature point detected by the feature detecting means. An image processing apparatus comprising: a correction unit that performs the correction and a combining unit that combines the character corrected by the correction unit with the input image. 2. The difference discriminating means discriminates whether or not the character sizes are different, and the feature detecting means, when the height or width of the character is larger than that of another character, the same data continues. A part or a missing part where the character is divided and separated is extracted as a feature point, and when the height or width of the character is smaller than other characters, the overlapping part where a part of the character overlaps is characterized. The points are extracted as points, and the correction means deletes the flow part that is the extracted feature point to attract the part that is displaced by the deletion, or interpolates the missing part that is the feature point, or the overlap that is the feature point. 2. The image processing apparatus according to claim 1, wherein the image processing apparatus is modified by performing separation so as to interpolate the separated portion. 3. The difference discriminating means discriminates whether or not a line or column of a character is twisted and is not linear, and the feature detecting means is a portion where the line or column of a character is twisted and not linear. 2. The image processing apparatus according to claim 1, wherein a portion in which the character is twisted and distorted is extracted as a feature point, and the correction unit smoothly corrects the distortion that is the extracted feature point. 4. A reading determination means for determining whether or not a character whose characteristic point is detected by the characteristic detection means is readable, and the reading determination means determines that the character cannot be read. Then, the image processing apparatus according to claim 1, wherein correction is performed by the correction means.