JP2003259111A - Image processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To correct the direction and tilt of an input image. <P>SOLUTION: The direction of the input image is decided (S102) to rotate the image to a normal direction (S105) and the tilt is decided (S106) to correct the tile of the image (S113). Here, if the decision making ends in failure at a step S102 (N at S103), tilt decision processing is performed first (S108) and after the tilt of the image is corrected, the direction of the image is decided again (S110). After the direction is decided, the image is rotated to the normal direction (S112) and outputted (S113). <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention relates to character recognition (OC).
R), an image processing method applicable to scan image browsing software, etc. for correcting the direction and inclination of an image, an image processing apparatus, a program for executing image processing, and a recording medium recording the program Is.

[0002]

2. Description of the Related Art In many cases, the accuracy of processing for determining the orientation of a document is degraded when the document is tilted. For example, in the document image inclination detection method of Japanese Patent No. 3187894, the original image is reduced in order to detect the document image inclination for detecting the inclination of the binarized document image (original image). The character string area is extracted, and the black pixel connected component in the character string area on the original image corresponding to the character string area, that is, for the circumscribing rectangle surrounding the character, creates a histogram of the height of each circumscribing rectangle, The height of the circumscribing rectangle having the highest frequency is used as a reference value, and a circumscribing rectangle that is larger than a predetermined threshold by the reference value (that is, higher) is converted to a reference value. The straight line with the smallest deviation from the center of the circumscribed rectangle is calculated using the least squares approximation method to obtain the inclination of the character string. In this detection method, in order to obtain the center point of a character, a conversion process is performed to remove the upper part of a rectangle whose height is larger than the reference. However, if the size of the characters is different, such as English letters, the direction of the document can be visually read, for example, the direction in which one character can be read naturally (here, the normal direction,
For convenience, this direction is facing north, and the other directions are facing east, west, and south.) Different from south facing, that is, upside down, the center point shifts, which adversely affects the overall tilt detection. become. In addition, the document tilt detection process often assumes that the document is in the normal orientation, and the document is in a direction other than the regular direction, that is, east, west,
Accuracy is often reduced when facing south.

Regarding the document orientation discrimination disclosed in, for example, Japanese Patent Laid-Open No. 2000-113103, certainty of character recognition (similarity, for example, character feature amount and dictionary pattern feature amount (Euclidean distance). , Mahalanobis distance, etc.) is used) to determine the orientation of the document. That is, the layout of the image is analyzed and the processing target area having the highest priority is selected. Specifically, for example, N rows are extracted from the selected area, and the first M rows are extracted. Recognition processing is performed on each character in four directions of up, down, left, and right at 90 ° intervals, and a character having a higher similarity (smaller distance) has a higher certainty factor. The direction determining means determines that the character has directionality when only the maximum certainty factor in the four directions exceeds the threshold value. When it is determined that there is directionality, it is counted in the direction frequency histogram created in the data storage unit to create a direction frequency histogram, and the direction with the maximum frequency value that is greater than or equal to the threshold value is determined as the image direction. To do. However, in this case, when it is tilted, it is more difficult to cut out characters and extract features than when it is not tilted, and the confidence factor information tends to be smaller than in the normal state, so the rate of undecidability tends to increase slightly. There is. In such cases,
An image that is tilted and whose document orientation is not the proper orientation may not have the expected accuracy regardless of which is processed first.

[0004]

SUMMARY OF THE INVENTION The present invention has been made to solve the problems of the prior art as described above.
The first purpose is to correct an image that is not in a normal direction and is tilted with high accuracy so that, for example, when reading a character by OCR or when a user browses a document image, it is not unnatural and uncomfortable. Is to A second object of the present invention is to prevent deterioration of image quality due to correction of image direction and inclination. A third object of the present invention is to accurately correct the direction and inclination of a multi-valued image without degrading the image. A fourth object of the present invention is to make it possible to easily execute the image direction and inclination correction processing by an arbitrary computer.

[0005]

According to a first aspect of the present invention, the direction of an input binary image is detected, the image is rotated in a normal direction, and the inclination of the image rotated in the normal direction is detected. Then, the image processing method is characterized by including each step of correcting the inclination according to the detected inclination.

According to a second aspect of the present invention, in the image processing method according to the first aspect, when the direction of the input binary image cannot be detected, the inclination of the image is first detected and the detected inclination is detected. The image processing method is characterized by including each step of correcting the inclination of the image, detecting the image direction for the second time, and rotating the image in the normal direction.

According to a third aspect of the present invention, in the image processing method according to the second aspect, the image is rotated in the normal direction based on the detection of the image direction of the second time, and thereafter the inclination detection of the image of the second time is performed. The image processing method is characterized by including each step of correcting the inclination of the image according to the detected inclination.

According to a fourth aspect of the present invention, a binary image for image direction and inclination correction processing is formed from the input image, and the binary image for correction processing is the image described in any one of the first to third aspects. Rotation in a normal direction and inclination correction are performed based on a processing method, the correction data is stored in a memory, and the direction and the inclination of the input image are corrected based on the correction data. Image processing method.

According to a fifth aspect of the present invention, a first binary image for detecting the image direction and a second binary image for detecting the inclination of the image are formed from the multi-valued image. Based on the image processing method described in any one of (1) to (3), the image of the first binary image is rotated in the normal direction and the inclination of the second binary image is corrected, and each correction data is stored in the memory. Then, the image processing method is characterized by including each step of performing rotation and tilt correction of the multi-valued image in the normal direction based on the correction data.

A sixth aspect of the present invention is a program for executing processing of each step in the image processing method according to any one of the first to fifth aspects.

A seventh aspect of the present invention is a computer-readable recording medium in which the program according to the sixth aspect is recorded.

According to an eighth aspect of the present invention, an image input means, a means for storing the program according to the sixth aspect, and an image processing control for executing rotation and inclination correction processing of an input image in a normal direction based on the program. And an image processing device.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments illustrating the present invention will be described. FIG. 1 is a block diagram showing the configuration of an embodiment of an image processing apparatus according to the present invention. The illustrated image forming apparatus is configured to be executed by digital processing by a microprocessor or the like. In the figure, a scanner 101 optically reads a document image original and converts it into image data which is an electric signal. CPU1 of computer
Reference numeral 02 denotes a CD-ROM / FD drive 1 from a program storage ROM / RAM 107 in which a program for executing the processing of the present invention is stored, or from an arbitrary recording medium 100 such as a CD-ROM / FD (flexible disk).
In step 05, the program is read out, and the work area RAM 106 executes the direction determination processing, the inclination angle determination processing, the image rotation processing, and the like for the document image data. The processed document image data is stored in the memory 103 according to an instruction from the CPU 102, displayed on the display 104 as necessary, or printed out by the printing device 108.

Next, description will be made on the creation of an image in the normal direction without inclination when the input image is not in the normal direction but in a direction different from the normal direction, or when the input image is input in a state in which there may be inclination. . (Invention of Claims 1 and 2) In the embodiments of the inventions of Claims 1 and 2, first, the document direction determination processing is performed (note that the document direction determination is performed by a known method. ). Then, the image is rotated in the direction opposite to the inclination, that is, corrected to generate an image without inclination. If the first document orientation determination fails, the inclination (inclination angle) is detected as the next step (step), and the document orientation is determined again after the inclination is corrected. As a result, in the second processing, the document orientation can be determined in a state where the image is not tilted, so that the orientation of the document can be successfully determined even in the image in which the first identification has failed. FIG. 2 is a flow chart for explaining this processing. An image is input (S101), the direction of the image is discriminated by an appropriately known method (S102), and the discrimination is successful (S103,
Y), if it is the normal direction (S104, Y),
If it is not in the normal direction (S104, N), the image is rotated in the normal direction (S105), and then the inclination determination process is also performed by an appropriate method that has already been known (S10).
6). The image is rotated and output in order to correct the finally determined inclination (S113). In the above process,
When the determination fails in step S103 (N in S103), the inclination determination process is performed in that state (S103).
108), according to the determination result, the image is rotated to correct the inclination of the image (S109). Next, the direction of the image is discriminated again in the state in which the inclination of the image is corrected (S11).
0), if the direction of the image as a result of the discrimination is the normal direction (S1
11, Y) If it is not in the normal direction as it is (S11
1, N), the image is rotated in the normal direction (S11
2) Output (S113). As described above, an image in the normal direction without inclination can be obtained. Hereinafter, the same processing will be described with respect to another embodiment.

(Invention of Claim 3) An embodiment of the invention of Claim 3 will be described. When the input image is input in a state where the document direction is not limited to the normal direction and there may be a tilt, the same processing as that of the embodiments of claims 1 and 2 is performed. Here, in the present embodiment, when the document orientation determination processing for the second time is successful, the orientation of the image is rotated to the regular orientation. By this, the orientation of the image is changed to the orientation of the image ( The inclination angle) can be accurately detected. So, for this normal orientation image, 2
If the angle for which the correction is necessary can be detected by performing the inclination detection process for the second time, the correction is performed at that angle. Figure 3
6 is a flowchart for explaining the processing according to the exemplary embodiment of claim 3. In FIG. 3, input an image (S
201), the direction determination processing is performed on the image by an appropriate method that has been conventionally known (S202), the determination is successful (S203, Y), and if the image is in the normal direction (S204, Y). As it is, if it is not in the normal direction (S204, N), the image is rotated in the normal direction (S).
205). Next, the inclination is discriminated by an appropriate method that has already been known (S206). The image is rotated to correct the finally determined inclination (S207), and the image is output (215). In the above processing, when the determination of the image orientation fails in step S203 (S20
3, N), tilt determination is performed in that state (S208), and rotation is performed to correct the tilt (tilt angle) of the image according to the determination result (S209). In this way, the orientation of the image is determined again with the inclination of the image corrected (S210),
As a result of the determination, if the direction of the image is the normal direction (S211,
Y) If it is not in the normal direction as it is (S211,
N), the image is rotated in the normal direction (S212), the image inclination determination process is performed again (S213), and the image is rotated and corrected based on the determination result (S214).
It is output (S215). By processing as above,
Both the orientation and the angle of the image can be recognized more accurately than the processing of the embodiments of the invention according to claims 1 and 2.

(Invention of Claim 4) An embodiment of the invention of Claim 4 will be described with reference to Examples 1 and 2. (Embodiment 1) In this embodiment, prior to image processing, the input image is first saved as an original image, and the actual processing is the same as in Embodiment 2 for the duplicate image, and the image direction is changed. , The tilt angle is detected and corrected. And all the corrected data (angle data)
Is stored in memory. For example, when the first direction determination fails, that is, in this case, the image cannot be rotated in the normal direction, so the correction angle A1 = 0 by the rotation.
Degree, the first image inclination angle of 1.3 degrees (correction angle B1 =-
1.3 degrees) Second direction determination East direction (direction of document when normal direction is north for convenience, correction angle A2 = -9)
0 degree) The second image tilt angle 0.8 degree (correction angle B2 =
-0.8 degrees). Here, in the entire correction, the first created duplicate image is corrected by -90 degrees in 90 units and by -2.1 degrees in fine units. Therefore, based on this correction data (angle data), the stored original image is corrected, and thereby the final image in which the target direction and inclination are corrected can be obtained. The reason for performing the above processing is that it is unavoidable that the image rotation is accompanied by a quantization error. Therefore, if the rotation of the minute angle is performed multiple times, the image quality obviously deteriorates, and the correction of the minute angle is performed at high speed. This is because, depending on the image processing software, correction may be performed for each byte, and in this case, there is a problem that the image is further distorted, so that it is avoided.

Here, the reason why the angle is divided into 90-degree unit and not 90-degree unit is that the image correction in 90-degree unit can be implemented by an algorithm different from the correction of the minute angle. Is rotated 180 degrees, all that is required is to reverse the top, bottom, left, and right. Therefore, when looking at the X and Y coordinates, a dot (X, Y) is converted to X = -x, Y = -y. It is only necessary to calculate, but in the case of a small angle, a certain dot (X, Y) is calculated as X = xcosθ + ysinθ, Y = ycos
This is because the calculation is performed by converting to θ−xsin θ, and therefore the calculation processing in 90 ° units can be performed at a higher speed than the calculation for correction of a minute angle. It should be noted that if there is a method in which the image rotation processing capable of correcting the image of −92.1 degrees of all the sums at one time is performed once, it is more efficient.

FIG. 4 is a flowchart showing the processing in the first embodiment described above. Enter the image (S30
1), a copy image is created for the image (S302),
In addition, the original image is saved as an original image. The direction of this copied image is discriminated by a method known in the art (S303), and the discrimination succeeds (S30).
4, Y), if the image direction is the normal direction (S305,
Y) If the image is not in the normal direction as it is (S305, N), the image is rotated in the normal direction (S306), and then the inclination determination process is performed by a conventionally known method (S307). The correction data (angle data) corrected in the above process is stored in the memory, and the stored original image is rotated based on the correction data stored in the memory (S308). It is output (S316).
In the above processing, when the determination fails in step S304 (N in S304), the inclination determination processing is performed in that state (S309), and the image is rotated to correct the inclination of the copied image according to the determination result (S310). In this way, the image orientation is determined again with the inclination of the copied image corrected (S311). If the orientation of the image is a normal orientation as a result of the determination (S312, Y), the orientation is not correct (Y). (S312, N), the image is rotated in the normal direction (S313), and the tilt determination is performed again (S31).
4). The correction data (angle data) corrected in the above process is stored in the memory, the original image is rotated based on the correction data stored in this memory (S315), and the corrected image is output. Yes (S31
6).

(Embodiment 2) Embodiment 2 is an operation when a multi-valued image (color or gray) is input. When a multi-valued image is input, the document direction,
It is necessary to perform conversion into a binary image that can be processed by skew (tilt) correction. Regarding the binarization method, for example, in Japanese Patent Laid-Open No. 9-233326, the brightness frequency is calculated for each block in a multi-valued image, the brightness frequency optimum region in the brightness frequency is specified, and the brightness frequency is optimized. The average brightness value of the area is calculated, the binarization threshold value of the block unit is calculated based on the average value, the area that complements the binarization threshold value of the block unit is set, and the block unit is set for the set area. Although a method such as a method of calculating a binarization threshold value for each pixel and binarizing the pixel value by complementing the binarization threshold value is disclosed, a multi-valued image is appropriately used by using such a known method. From which a binary image can be formed. In the present example, the image inclination and the image direction are detected for the obtained binary image by any of the methods of the embodiments of the invention according to claims 1 to 3. In order to reflect the result, correction angle data (information) is stored in the present embodiment as in the case of the embodiment of claim 3, and the angle is corrected for the original multi-valued image. Thus, skew correction of the color image and document direction correction are possible.

FIG. 5 is a flow chart showing the processing of the second embodiment, and the processing procedure of the fourth embodiment will be described below with reference to FIG. First, a multi-valued image is input (S401). Next, a binary image copy image is created for the image (S402).
The original image is stored as an original image. The direction of the copied image is discriminated by the conventional method, for example (S403). As a result, the discrimination is successful (S404, Y), and the direction of the image is changed. If it is in the normal direction (Y in S405), it is not in the normal direction (S4).
05, N), the binary image is rotated in the normal orientation (S406), and then the inclination determination process is performed by a method already known in the related art (S407). The correction data (angle data) corrected in the above process is stored in the memory, and the stored multivalued input image is rotated based on the correction data stored in the memory (S408). )
It is output (S416). In the above processing, when the image orientation determination fails in step S404 (N in S404), the inclination determination processing is performed in that state (S409), and rotation is performed to correct the inclination of the binary image according to the determination result. Yes (S410). In this way, the orientation of the image is discriminated again with the inclination of the binary image corrected (S411). If the orientation of the image as a result of the discrimination is the normal orientation (S412, Y), it is not the normal orientation as it is ( (S412, N), the binary image is rotated in the normal direction (S413), and the tilt determination is performed (S414). The correction data (angle data) corrected in the above process is stored in the memory, and the multivalued image is rotated based on the correction data stored in the memory (S415).
An image is output (S416).

(Invention of Claim 5) In the present embodiment, it is considered that the optimum binary image for detecting the skew and the document direction are different when performing the binarization. is there. For example, in order to determine the document direction, it is effective to use character recognition.
In the processing method disclosed in Japanese Unexamined Patent Application Publication No. 00-113103, binarization that increases the accuracy of character recognition is optimal. For example, when a 200 dpi color image is input, it has been known that in Japanese or the like, the recognition accuracy is improved by enlarging the image to 400 dpi and performing character recognition. Therefore, in order to discriminate the document orientation, binarization for enlarging the image is desirable. Further, as far as image tilt determination is concerned, for example, unlike the document image tilt detection method disclosed in the specification of Japanese Patent No. 3187894 mentioned above, characters are not recognized,
Many of them utilize the character and a possibly circumscribed rectangle, and these known detection methods can be used. Further, for example, the method shown in Japanese Patent Laid-Open No. 7-105310, which focuses on ruled line information, can be used. In this method, for example, as shown in FIG. 6, when the black runs in the main scanning direction are focused on the ruled lines in the main scanning direction, the start points of the two long black runs LH1 and LH2 to be integrated in the main scanning direction are integrated. Alternatively, the X coordinate of the end point is compared. As shown in FIG. 6A, the relationship between the X coordinate (xs1) of the starting point of the upper black run LH1 and the X coordinate (xs2) of the starting point of the lower black run LH2 is xs1 <xs.
If it is 2, it is determined to be “downward right”, and if the X coordinate (xe1) of the end point of the black run LH1 and the X coordinate (xe2) of the end point of the black run LH2 are xe1> xe2 as shown in FIG. 6B. It is determined to be "rightward". If none of the above, it is determined that there is no inclination. This determination is performed for a plurality of pairs of integrated black runs, and the final determination result is determined by, for example, a majority decision of the determination results of each pair. Alternatively, the number of times of discrimination of each of the downward sloping and the upward sloping is counted, and the inclination direction which has reached the predetermined value first is used as the final discrimination result. In addition, the inclination angle θ is, for example,
The thickness of the ruled line = the number of black pixels / the width of the ruled line is obtained from the width of the ruled line and the black pixel of interest, and from the height of the ruled line of interest and the thickness of the ruled line, the deviation of the ruled line = the height of the ruled line−the ruled line The thickness is obtained, and the inclination angle θ is obtained from tan θ = ruled line deviation / ruled line width.

Here, in detecting the image inclination, it is better not to enlarge the image but to reduce the processing amount by conversely reducing it, or to perform edge emphasis to perform binarization so that the ruled line appears clearly. Higher precision detection can be expected. In the embodiment of the invention according to claim 5, in consideration of this point, an image suitable for each process is created, and the image direction and the image inclination are detected by applying the images to the original multi-valued image. By reflecting it, it becomes possible to create an image in which the image direction and the image inclination of the original image are eliminated. FIG. 6 shows a flowchart for explaining the process of this embodiment. A multi-valued input image is prepared (S501), a binary image for discriminating the direction of the image is created as a copy image (S502), and the original multi-valued input image is stored as an original image.
The direction discrimination processing is performed on the copied binary image by a conventionally known method (S503). As a result, the discrimination is successful (S504, Y), and if the image direction is the normal orientation (S505, Y). As it is, if it is not the normal direction (S505, N), the binary image is rotated in the normal direction (S506), and then the binarization process suitable for the tilt determination,
That is, binarization for inclination determination is performed (S507), and inclination determination processing is performed by a method already known in the related art (S508). The correction data (angle data) corrected in the above process is stored in the memory, and the stored multivalued input image is rotated based on the correction data stored in this memory (S509). ) Output (S519). In the above process, if the determination fails in step S504 (N in S504), the binarization process for tilt determination is performed (510), and the tilt determination process is performed in that state (S5).
11). According to the result of the determination, rotation processing is performed to correct the inclination of the binary image (S512). In this way, in the state where the inclination of the binary image is corrected, the direction discrimination binarization process is performed again to discriminate the direction of the image (S513), and the direction discrimination process is performed (514). As a result of the discrimination, if the direction of the image is the normal direction (S515, Y), and if it is not the normal direction (S515, N), the binary image is rotated to the normal direction (S516) and the inclination judgment is performed (S516). S517). The correction data (angle data) corrected in the above processing is stored in the memory, and the multivalued image rotation processing is performed based on the correction data stored in this memory (S51).
8) The image is output (S519).

[0023]

The effect corresponding to claim 1; since it is possible to accurately correct an image which is not in the normal direction and is inclined, O
When reading characters with CR or when a user browses a document image, correct the direction and tilt of the image to improve literacy accuracy or obtain an image that is not unnatural and unnatural when viewed. You can Effect corresponding to claim 2 Even if the initial determination of the image direction fails, the image can be determined by correcting the inclination of the image and then determining the image direction again. Effect corresponding to claim 3; Correct inclination correction of an image can be performed by performing the inclination of the image again after the image determination. Effect corresponding to claim 4; A binary image for orientation and tilt correction processing of an original image created from an original image is created, image orientation and tilt correction processing is performed on the image, and the original data is used by using the correction data. Since the image is corrected, it is possible to obtain the image in which the direction and the inclination are corrected by preventing the image deterioration caused by the rotation of the minute angle and the like. Effect corresponding to claim 5; When forming a binary image from a multivalued image, a first binary image for correcting the orientation of the image and a second binary image for correcting the inclination of the image. And are created separately, the direction of the image is determined for the first binary image and correction is performed, and the inclination is determined for the second binary image and correction is performed, and multivalued with each correction value. Since the direction and the tilt of the image are corrected, the correction of the direction and the tilt of the multi-valued image can be performed with high accuracy without deterioration of the image. Effect corresponding to claim 6; By causing a computer to read the program for the processing of the present invention, the processing for correcting the direction and inclination of the image can be easily executed. Effect corresponding to claim 7: The program for executing the process of correcting the direction and the inclination of the image can be loaded into any computer to easily execute the process. Effect corresponding to claim 8: A program for processing of the present invention is read, the direction and inclination of an input image are corrected according to the processing procedure, and the effect corresponding to claims 1 to 5 can be realized. .

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an image processing apparatus according to an embodiment of the present invention.

FIG. 2 is a flow chart of a process for correcting the direction and inclination of an image according to the embodiment of the present invention.

FIG. 3 is a flowchart of a process for correcting the direction and inclination of an image according to another embodiment of the present invention.

FIG. 4 is a flowchart of a process for correcting the direction and inclination of an image according to another embodiment of the present invention.

FIG. 5 is a flow chart of a process for correcting the direction and inclination of an image according to another embodiment of the present invention.

FIG. 6 is a diagram for explaining an example of an image inclination determination method.

FIG. 7 is a flowchart of a process for correcting the direction and inclination of an image according to another embodiment of the present invention.

[Explanation of symbols]

100 ... Recording medium (CD-ROM), 101 ... Scanner, 102 ... CPU, 103 ... Memory, 104 ... Display, 105 ... CD-ROM / FD drive, 106
... work area RAM, 107 ... program storage ROM
/ RAM, 108 ... Printing device.

Continued front page    F term (reference) 5B057 AA11 BA02 CA02 CA06 CA08                       CA12 CA16 CB02 CB06 CB12                       CB16 CC01 CD03 CE12 DA17                       DB02 DB05 DB08 DC08 DC22                       DC36                 5C076 AA24 BA06                 5L096 AA07 FA67 MA03

Claims (8)

[Claims] 1. A direction of an input binary image is detected, the image is rotated in a normal direction, an inclination of the image rotated in the normal direction is detected, and the inclination is corrected according to the detected inclination. An image processing method comprising the steps of: 2. The image processing method according to claim 1, wherein when the direction of the input binary image cannot be detected, the inclination of the image is first detected, and the inclination of the image is detected according to the detected inclination. Is corrected, then the image direction is detected for the second time, and the image is rotated in the normal direction. 3. The image processing method according to claim 2, wherein the image is rotated in the normal direction based on the detection of the image direction of the second time, and thereafter the inclination of the image of the second time is detected, and the detected inclination is detected. Corrects the inclination of the image according to
An image processing method comprising each step. 4. A binary image for image direction and tilt correction processing is formed from an input image, and the binary image for correction processing is normalized based on the image processing method according to claim 1. An image processing method comprising: performing rotation and tilt correction in a direction, storing the correction data in a memory, and correcting the direction and tilt of the input image based on the correction data. 5. A first for image direction detection from a multi-valued image
Image of the first binary image based on the image processing method according to any one of claims 1 to 3, wherein the second binary image for detecting the inclination of the image and the second binary image for detecting the inclination of the image are formed. Rotation in the normal direction and inclination correction for the second binary image are performed, each correction data is stored in the memory, and rotation and inclination correction of the multi-valued image in the normal direction are performed based on the correction data. And an image processing method including each step. 6. A program for executing processing of each step in the image processing method according to claim 1. Description: 7. A computer-readable recording medium in which the program according to claim 6 is recorded. 8. An image input means, a means for storing the program according to claim 6, and an image processing control means for executing rotation and inclination correction processing of an input image in a normal direction based on the program. An image processing device characterized by: