JP2003187244A - Image correction method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image correction method by which an image with high precision and joints of the same images are inconspicuous when the same images are arranged in line, without requiring the skill and manual work of a worker, with high processing efficiency and with excellent reproducibility. <P>SOLUTION: When an area required to be corrected is specified near the joints of images (indicated by bold lines), a corrected place is set as a blank pixel. When a pixel value is reset to the blank pixel P, non-blank pixels (indicated by a diagonal mesh) around the blank pixel P are defined as reference masks, comparison masks in the same shape as that of the reference masks are arranged on the images (Fig. (a)) and similarity of the areas masked by both masks is calculated. The similarity is calculated by transferring the comparison masks on the image and a pixel value of a comparison pixel Q (Fig. (c)) corresponding to a position of the comparison mask the similarity of which is maximized is provided to the blank pixel P. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to the creation of images used for printed building materials for decorating interior surfaces of buildings such as walls, floors and ceilings, furniture, fittings, surfaces of featured parts, etc. The present invention relates to a technique for creating a seamless endless image so as not to be conspicuous.

[0002]

2. Description of the Related Art Building material printed matter is much longer (longer) than general printed matter such as publication printed matter and commercial printed matter. For example, considering a printed material to decorate the wall surface, the length from the floor to the ceiling of the building is usually about 2.5 m,
In addition, the horizontal length from the end to the end of the room is about 3.5 m in the case of 6 tatami mats. Therefore, the images for printed materials for building materials are also required to be endless images, that is, a continuous continuous image of at least that length, that is, an endless image.

The long image as described above is printed on the plate cylinder 1 for printing.
It is impossible to fit within the circumference. Therefore, the side of the start end and the side of the end of the image for one circumference in the printing direction are connected (connected) on the plate cylinder to perform plate making. Then, each time the plate cylinder makes one rotation during printing, an image for one circumference is repeatedly printed without interruption, thereby obtaining a long printed matter of an arbitrary length.

However, if the side of the start end and the side of the end in the printing direction of the image for one circumference are simply butted and connected, the image becomes discontinuous at the connecting portion (joint) and the endless image is obtained. Can't get In order to make the plate cylinder image endless, it is necessary to make the discontinuity of both images continuous at the seam of the image and make the seam unrecognizable, that is, seamless. Becomes

Therefore, as a method of making the plate cylinder image endless by seamlessizing the image joints, which is particularly effective in the case of a wood grain pattern, as described in Japanese Patent Publication No. 43-2065, etc. As a wavy line or zigzag broken line, the joint itself is made inconspicuous and the area near the joint is manually added, decolorized, dyed, deleted, etc. by a skilled worker to correct the discontinuity of the image and seamless Has been widely used (prior art).

Further, in recent years, due to the progress of digital image processing technology applying an electronic computer, an original image is read by a scanner to be converted into digital data, and a correction portion is designated by a mouse, a tablet, a cursor or the like on a display device such as a CRT. It has been attempted to further improve workability by a technique such as copying an appropriate image from another portion of the screen or the like and transplanting the image (prior art). Furthermore, the work of making the seams seamless is performed by averaging calculation processing of both side images.
A method of improving efficiency by automatically calculating with a predetermined formula has also been proposed (prior art).

[0007]

However, in the above-mentioned prior art, since the seamless operation is performed manually by a skilled worker, the work efficiency is low and it has become difficult to secure a skilled worker in recent years. Moreover, it is also difficult to reproduce a completely seamless plate when reprinting.

In the above-mentioned prior art, by adopting the digitized and mechanized processing for the work of the seamless operation, the work efficiency can be improved in the steps such as the joint of the images as compared with the above-mentioned prior art. . However, the seamlessness work itself has not been improved much in that it depends on the intuition and manual work of a skilled worker.

In the above-mentioned prior art, the efficiency has been improved by automating the work of making it seamless by digital processing. However, since the seamlessization simply adds (averages) the images on both sides at the joint, the high frequency component of the spatial frequency is lost near the joint, and the fine structure or sharpness of the image is reduced. Therefore, the problem remains that the joint is still recognized when viewed from a distance.

In view of the above points, according to the present invention, it is possible to create an image with high accuracy and inconspicuous seams, and at the same time, the skill of the operator and the manual work are unnecessary, and the processing efficiency is high, Another object is to provide an image correction method having excellent reproducibility.

[0011]

In order to solve the above problems, according to the present invention, when a plurality of the same images are arranged, the pixels near the edges of the images are corrected so that the joints at the boundaries are not noticeable. At this time, a position changing step in which the position of the pixel is changed so that the end portion of the correction target image is located at a position other than the end portion, and a position changing step is performed, in which each pixel in the area designated for the position changing image is A blank pixel setting step of setting as a blank pixel, a pixel value determining step of determining the value of each blank pixel are executed, and as the pixel value determining step, for each blank pixel, the blank pixel is defined as a reference pixel, A mask shape determining step of determining the existence position of non-blank pixels around the reference pixel as a reference mask, arranging a comparison mask of the same shape as the reference mask in the position change image, and changing the position of the comparison mask. While moving on the image, calculate the similarity between the area masked by the comparison mask and the area masked by the reference mask at each position, and select the pixel corresponding to the comparison mask when the similarity is maximum. The value of each blank pixel is determined by executing a pixel value giving step of giving the pixel value of the selected pixel as the pixel value of the reference pixel. According to the present invention, in performing the seamless processing of the created image, a blank pixel is set near the boundary of the correction target image, and the blank pixel is similar to the value of the pixels around the blank pixel. Even if the image has both a relatively rough structure and a relatively fine texture, the pixels that have a value are specified in the vicinity and the value of that pixel is given as the value of the blank pixel. It becomes possible to make corrections without feeling uncomfortable.

[0012]

DETAILED DESCRIPTION OF THE INVENTION An embodiment of the present invention will be described in detail below with reference to the drawings. This corresponds to the case where the method of the present invention is applied to the application of making a plate cylinder for a rotary printing press based on a wood grain original. FIG. 1 is a flowchart showing an outline of an image correction method according to the present invention. First, a document image to be subjected to seamless processing is input by using a scanner, a digital camera or the like, AD-converted into digital information (step S1). Here, in FIG.
Original image G having a grain pattern as shown in (a)
Enter _ORI (indicated by the dashed-dotted rectangle in the figure). Then, a desired area is cut out from the input original image G _ORI , and this is used as a correction target image G _RET (shown by a solid line rectangle in the drawing). Of course, the original image G _ORI may be directly used as the correction target image G _RET . Here, the Y-axis direction of the coordinate system is the printing direction, that is, the circumferential direction of the plate cylinder,
The correction target image G _RET is a quadrangle abcd whose four sides are parallel to the X-axis direction and the Y-axis direction.

Next, of the input correction target image G _RET ,
A process of exchanging the upper and lower sides is performed (step S2). Here, for convenience, the + Y axis direction is referred to as an upward direction and the −Y axis direction is referred to as a downward direction. Specifically, as shown in FIG. 2B, a cutting line ef that is parallel to the upper side ab and the lower side dc of the correction target image G _RET (parallel to the X axis) and is located between the upper side and the lower side is defined. Then, the upper half image GA and the lower half image GB of the correction target image G _RET are divided, and then the upper half image GA and the lower half image GB are moved in parallel to completely exchange the upper and lower positions. Since the purpose here is to sufficiently include the vicinity of the seam of the image in the range of the image, the position (Y coordinate) of the cutting line ef is not necessarily in the middle between the upper side ab and the lower side dc. good. Also,
In FIG. 2, the upper and lower sides of the correction target image G _RET are interchanged, but in the present invention, the upper and lower sides of the images may not necessarily be interchanged, but the left and right sides may be interchanged. In either case, it is sufficient that the direction of image replacement corresponds to the portion where the images are joined.

As shown in FIG. 2B, in the vertically inverted image G _REV in which the upper half image GA and the lower half image GB of the correction target image G _RET are exchanged, the upper side ef and the lower side ef are both original. Since it is the same part of the original image _ORI , the upper side ef and the lower side ef of the vertically inverted image G _REV do not form a joint even if they are joined in a cylindrical shape in the circumferential direction (Y-axis direction). Instead, the joined portion of the upper side ab of the upper half image GA and the lower side dc of the lower half image GB generated inside the reverse image G _REV is joined to another portion of the original image, and thus is a discontinuous image. (The annual rings N are misaligned in FIG. 2 (b)), and the joints are conspicuous. Here, the attachment lines ab (or d) connected by the replacement
c: G _AB is designated as a strip-shaped area to be corrected by the operator in the vicinity thereof including (indicated by a dotted line in FIG. 2B) (step S3). Specifically, the replaced image is displayed on the screen, and the operator uses an input device such as a mouse while looking at the image and displays the image at an appropriate position above the attachment line ab (or dc) as shown in FIG. The upper boundary line E _UP (X) is set as shown in a), and the lower boundary line E _LOW (X) is set at an appropriate position below the joining line ab (or dc). At this time, the area to be modified G _AB can be set in any shape and size. The upper boundary line E _UP (X) and the lower boundary line E _LOW (X) may be straight lines, broken lines, or curved lines. Practically, it is preferable to specify the correction required area G _AB by leaving the areas having the preferable patterns included in the original designs / patterns around the joining line ab (or dc).

Pixels existing in the band-shaped area set as the correction-required area G _AB are once reset to be blank pixels (uncorrected pixels) and then given new pixel values to generate continuous images. (Step S4). Here, blank means that the value of the part becomes meaningless. The specific processing is performed by setting the value of the pixel in the blank portion to 0, setting the fourth channel different from the three colors of RGB, and setting a flag indicating that to the blank pixel. . At this time, the state of the reverse image G _REV in which the blank area is set is shown in FIG.

Now, the process of generating continuous images in step S4 will be described in detail with reference to the flowchart shown in FIG. First, a blank pixel is searched (step S41). This process is performed sequentially while scanning from the upper row to the lower row of the image, and from the left pixel to the right pixel within the same row. That is, the search is started from the upper left end of the inverted image G _REV shown in FIG. FIG. 5 shows an enlarged view near the joining line ab. In FIG. 5, a thick line crossing the center is a joining line of the lower half image GB and the upper half image GA. Each square represents a pixel, of which a shaded pixel represents a pixel having a value, and a non-shaded pixel represents a pixel set as a blank. It can be seen that the blank pixels are sandwiching the joining line,
This is based on the designation of the worker in step S3. If the search is performed from the upper row in the example shown in FIG. 5 and the search is performed from the left pixel in the same row, the pixel indicated by "P" is detected as a blank pixel. As a specific blank pixel search process, for example, when a flag indicating a blank is set on the fourth channel as described above, it is sufficient to detect the flag set.

Subsequently, the blank pixel is used as a reference pixel P.
(X _S, Y _S) is defined as the reference pixel P (X _S, Y _S) to determine the mask shape necessary to determine the value of (step S42). Specifically, as shown in FIG. 6, a range of 5 × 5 pixels centering on the reference pixel P (X _S , Y _S ) is set, and a mask is set for the non-blank pixel portion. This is called a reference mask. For the reference pixel P (X _S , Y _S ) shown in FIG. 6, a reference mask extending over 15 pixels is set as shown by hatching.

When the reference mask is set, the reference pixel P
To determine the value of (X _S , Y _S ), a comparison mask having the same shape as the reference mask is arranged on the inverted image G _REV , and the similarity sim of the regions masked by both masks is obtained (step S43). . In the process of step S43, in order to determine the value of one reference pixel P (X _S , Y _S ), the comparison mask is arranged at positions corresponding to all the pixels of the inverted image G _REV , and the line is set for each position. Be seen. However, no comparison is made at positions corresponding to blank pixels and positions where the comparison mask does not fit within the reverse image G _REV . When the similarity sim is calculated for all positions to be compared on the inverted image G _REV (step S44), the pixel of the pixel Q (X _R , Y _R ) corresponding to the comparison mask having the highest similarity sim. The value is set as the pixel value of the reference pixel P (X _S , Y _S ) (step S45).

Here, steps S43 to S45
The process of FIG. 7 will be described in detail with reference to the drawings.
As shown in (a), a comparison mask is arranged at the upper left corner of the inverted image G _REV , and the similarity sim between the area masked by the comparison mask and the area masked by the reference mask is obtained. Next, as shown in FIG. 7B, the comparison mask is shifted to the right by one pixel, that is, the Y coordinate is left as it is, and the X coordinate is shifted to a position where the X coordinate is increased by +1 and masked by the masked region and the reference mask. Find the similarity sim with the region. Similarly, the process of obtaining the similarity sim is performed for the comparison mask centered on each pixel Q (X _R , Y _R ) and when the comparison mask reaches the right end (maximum X coordinate value) on the image, the comparison mask To the left edge one line down (Y coordinate value decremented by 1, X coordinate value minimum)
To calculate the similarity sim between the masked area and the area masked by the reference mask. In this way, the comparison mask is the inverted image G except for the blank part on the image.
Comparison masks are arranged at positions corresponding to all pixels Q (X _R , Y _R ) within a range that falls within _REV , and the similarity sim between the masked region and the region masked by the reference mask is obtained. For example, when the comparison mask is arranged at the position shown in FIG. 7C, and it is determined that the similarity sim with the masked area is the highest, the positional relationship with the comparison mask is based on the reference value. The value of the comparison pixel Q having the same positional relationship as that of the mask and the reference pixel P is set as the value of the reference pixel P. Here, setting the value of the comparison pixel Q as the value of the reference pixel P means, for example, R
In the case of a GB color image, R · that the comparison pixel Q has
This is performed by giving the respective values of G and B as the values of R, G and B of the reference pixel P, and setting the flag indicating the blank portion to 0 for the fourth channel different from the three colors of RGB.

Here, in the present embodiment, a function for calculating the similarity sim between the area masked by the reference mask and the area masked by the comparison mask will be described.
As this function, any function may be used as long as it can determine the similarity between the pixel existing at the mask position of the reference mask and the pixel existing at the mask position of the comparison mask. The formula shown in [Formula 1] is used.

[0021] [Equation 1] _{Sim = 1 / {Σ k (} V Pk -V Qk) 2} However, k = 1 to K (the number of reference pixels in the mask)

In the above [Formula 1], Sim is the similarity.
Yes, V_Pk , V_Qk Is a reference pixel with the reference pixel P.
Positional relationship within the mask, comparison mask with comparison pixel Q
The pixel value of a pixel having the same positional relationship in is shown. Well
Σ_kCalculates the sum when k = 1 to K
Indicates that That is, Σ_k(V_Pk-V_Qk)²= (V _P1
-V_Q1)²+ (V_P2-V_Q2)²+ ... + (V_PK-V_QK)²When
Become. Here, K indicates the number of pixels in the mask.
For example, using a mask with the shape shown in FIG. 6 and FIG.
If so, K = 15. Pixel value V_PkAnd pixel value V_QkDifference
For a monochrome image, compare the pixel values as they are.
Good, select one color as the representative color for color images
May be calculated based on the difference, or two colors or all three colors may be calculated.
You may calculate the sum total of the square of all the differences.

The pixel having the same positional relationship in the mask as described above means that the coordinate value of the reference pixel P is (Px, P
y), where the coordinate value of the comparison pixel Q is (Qx, Qy), each reference pixel (Qkx, Qky) in the comparison mask and each reference in the standard mask defined by the following [Formula 2]. The relationship of pixels (Pkx, Pky) is established.

[Formula 2] Qkx = Qx + Xk Qky = Qy + Yk However, Xk = Pkx-Px Yk = Pky-Py

When the value of the reference pixel P as shown in FIG. 6 is determined, as shown in FIG. 8, in order to obtain the value of the next blank pixel P, a range of 5 × 5 pixels centering on the reference pixel P is obtained. Set a new reference mask inside. When the reference mask is set, a comparison mask having the same shape as this reference mask is set, and the similarity Sim between the reference mask and the comparison mask is calculated by [Equation 1] for all pixels in the image as shown in FIG. Seeking,
The value of the comparison pixel Q having the maximum similarity is set as the value of the reference pixel P. At this time, since there are 13 pixels in the mask as shown in FIG. 8, [Formula 1] is applied with K = 13.

Similarly, pixel values are successively given to blank pixels. When the non-blank pixels of the original upper half image GA are included in the range of 5 × 5 pixels centered on the blank pixel P, the value of the blank pixel P is influenced by the upper half image GA. . For example, when determining the value of the blank pixel P as shown in FIG. 9, the upper half image GA is included in the reference mask.
7 non-blank pixels are included. Even when the reference mask as shown in FIG. 9 is set, the pixel value of the blank pixel P is determined after calculating the similarity with the comparison mask in the same manner as shown in FIG. It In this case, since the non-blank pixels of the upper half image GA are included in the mask, the pixel value of the blank pixel P is affected. Therefore, the joint with the upper half image GA becomes inconspicuous. That is, seamlessness is achieved. When the pixel values are given to all the blank pixels on the reverse image G _REV , the pixel value determination process ends (step S46).

In this way, when the continuous image generation process in step S4 is completed, the corrected image is output by a predetermined output means (step S).
5).

FIG. 3 shows how the image is modified according to the present invention. In step S41 of FIG. 1, FIG.
The inverted image G _REV in which blank pixels are set as shown in
In step S45 of FIG. 4, pixel values are sequentially given to the blank pixels. That is, the generated image G _GEN is generated in the correction required area G _AB . The state of the image at this time is shown in FIG. Furthermore, by giving pixel values to the blank pixels, a seamless image including the generated image G _GEN as shown in FIG. 3C is finally obtained.

(System Configuration) Next, a system configuration for executing the image correction method will be described. Figure 1
0 is a block diagram showing an embodiment of a system for executing the image correction method. In FIG. 10, the image input means 1 has a function of inputting a document image G _ORI . As the image inputting means 1, when the original image G _ORI is already prepared as digital data, it can be realized by a reading device capable of reading an electronic recording medium on which the digital data is recorded, and the original image G _ORI is a paper or the like. When prepared as a physical medium, it can be realized by an optical reading device such as a scanner.

The display means 2 displays the upper half image GA of the original image G _ORI , the correction target image G _RET , and the correction target image G _{RE T.}
It is for displaying a reverse image G _{REV in} which the lower half image GB is exchanged, and various display devices such as a CRT and a liquid crystal can be used. The instruction input means 3 is used by the operator to designate the required correction area G _AB while looking at the reverse image G _REV displayed on the display means 2, and can be realized by an input device such as a mouse or a keyboard.

The arithmetic and control unit 4 realizes the image correction method.
It plays a central role in the system for
Image inversion means 4a, blank pixel setting means 4b, mask shape
It has a determining means 4c and a pixel value giving means 4d. image
The inverting means 4a executes the process of step S2 in FIG.
It has a function, and in the present embodiment, the correction target image G
_RETSwap the upper half image GA and the lower half image GB
ing. The blank pixel setting means 4b uses the step S of FIG.
41 has a function of executing the processing of 41 and has an instruction input means.
Set the pixels in the range specified by 3 as blank pixels
To do. The mask shape determining means 4c uses step S4 of FIG.
It has a function to execute the process of 2 and is used as a reference pixel.
Set a 5 × 5 range centered on the blank pixels
The position of the non-blank pixel in the box is set as the mask shape of the reference mask.
It The pixel value giving means 4d starts from step S43 of FIG.
It has a function to execute the process of step S45, and
Areas masked with the quasi-mask and masked with the comparison mask
Adjacent to the comparison mask when the similarity of the
The value of the comparison pixel Q to be applied is given as the value of the reference pixel P.
The arithmetic and control unit 4 is realized by the CPU and memory of the computer.
Image inversion means 4a, which is a specific function, and blank pixels
Setting means 4b, mask shape determining means 4c, pixel value assigning hand
Step 4d is dedicated software installed in the computer.
It is realized by a program.

The image output means 5 is means for outputting an image which is made seamless by being modified,
An external storage device such as a hard disk, a recording device for a portable electronic recording medium, and a communication device for transmitting to a remote place via a network are applicable.

[0033]

As described above, according to the present invention,
When arranging the same image more than once, when correcting pixels near the edge of the image so that the joints at the boundary are not noticeable, the pixel position so that the edge of the image to be corrected is located outside the edge To obtain the position-changed image, set each pixel in the area specified for this position-changed image as a blank pixel, determine the value of each blank pixel, and determine this pixel for each blank For a pixel, the blank pixel is defined as a reference pixel, the positions where non-blank pixels around the reference pixel are present are determined as a reference mask, and a comparison mask having the same shape as the reference mask is placed in the position-changed image for comparison. While moving the mask on the position-changed image, the similarity between the area masked by the comparison mask and the area masked by the reference mask at each position is calculated, which corresponds to the comparison mask when the similarity becomes maximum. Since the value of each blank pixel is determined by selecting a pixel and giving the pixel value of the selected pixel as the pixel value of the reference pixel, both a relatively rough structure and a relatively fine texture can be obtained. Even if the image is included, it is possible to correct the image without feeling uncomfortable.

[Brief description of drawings]

FIG. 1 is a flowchart of an image correction method according to the present invention.

FIG. 2 is a diagram showing a relationship among a document image G _ORI , a correction target image G _RET , and a vertically inverted image G _REV .

FIG. 3 is a diagram showing how an image is corrected by generating a generated image G _GEN in a correction required area G _AB .

FIG. 4 is a flowchart showing details of processing in step S4 of FIG.

FIG. 5 is an enlarged view of the vicinity of a joining line when a blank pixel is set.

FIG. 6 is a diagram showing how a mask shape is determined.

FIG. 7 is a diagram showing how a comparison mask is moved to calculate a degree of similarity.

FIG. 8 is a diagram showing how mask shapes of adjacent blank pixels are determined.

FIG. 9 is a diagram showing how a mask shape is determined when non-blank pixels of the upper half image GA of the correction target image G _RET are included.

FIG. 10 is a diagram showing a configuration of a system that realizes an image correction method according to the present invention.

[Explanation of symbols]

1. Image input means 2 ... Display means 3 ... Instruction input means 4 ... Arithmetic control device 4a ... Image reversing means 4b: blank pixel setting means 4c ... Mask shape determining means 4b ... Pixel value providing means 5: Image output means

Claims (3)

[Claims] 1. An image input step of inputting a document image G _ORI and designating a correction target image G _RET having a rectangular shape from the input document image, and X in parallel with four sides of the correction target image G _RET. With an XY coordinate system having axes and a Y-axis, the + Y direction is upward, the -Y direction is downward, and the vertices of the quadrangle forming the correction target image G _RET are sequentially turned clockwise from the upper left in the order of a, b, c ,. When the correction target image G _RET is defined as d, the upper side a that forms the boundary line of the upper end portion
It is located at an intermediate portion between b and the lower side cd which forms a boundary line between the lower end portion and the parallel with the cutting line ef and is divided into an upper half image GA and a lower half image GB by an upper half image GA and a lower half image GA. An image inversion step in which the upper and lower positions of the half image GB are exchanged by parallel movement and the lower end dc of the lower half image GB and the upper end ab of the upper half image GA are joined so as to obtain an inverted image G _REV. In the reverse image G _REV , the upper boundary line E _UP (X) is located above the joining line where the upper end ab and the lower end dc are joined together, and below the joining line below the joining line. A boundary line E _LOW (X) is set, and then a strip-shaped correction area G _{AB is provided} between the upper boundary line E _UP (X) and the lower boundary line E _LOW (X).
A main correction area specifying step to the while sequentially scanning the pixels of the correction required in the region G _AB, sequentially performs determination processing of the pixel value, the upper half image from the lower half image GB in the correction required area G _AB has a continuous image generation step of image toward GA to produce a product image G _GEN as continuously connected, the continuous image generation step, (a) unmodified pixel P (X of the principal modified region G _{AB S,} Y _S) to the fixed one 1, (b) on the lower border E _LOW (X) above the lower region than in and a lower half image GB upper boundary line E _UP (X) of half images GA All pixels Q (X _R , Y _R ) in the area are scanned, and pixels P (X _S ,
Y _s ) and the pixel Q (X _R , Y _R ) similarity sim {(X _S ,
Y _S ), (X _R , Y _R )} is calculated for all pixels Q (X _R , Y _R ), and (c) similarity sim {(X _S , Y _S ), (X _R ,
Y _R )} is the maximum, and the pixel value of the pixel Q (X _R , Y _R ) is replaced with the pixel value of the pixel P (X _S , Y _S ). (D) If all uncorrected pixels P (X _S , Y _S ), if the above processes (a) to (c) are completed, the process is terminated, and
Uncorrected pixel P (X
_S , Y _S ) remains, another uncorrected pixel P
An image correction method, which is executed by the processes of (a) to (d), in which (X _S , Y _S ) is selected and the process returns to the process of (a). 2. The calculation of the degree of similarity in the step of generating continuous images is performed by setting a square area around the reference pixel P and using a non-blank pixel existing position included in the square area as a reference mask. This is performed by setting a square area around Q, setting the existence position of the non-blank pixels included in the square area as a comparison mask, and obtaining the similarity between the reference mask and the area masked by the comparison mask. The image correction method according to claim 1, wherein: 3. An original image G _ORI is input to a computer, and from the input original images,
An image input step of designating a correction target image G _RET having a rectangular shape, an XY coordinate system having an X axis and a Y axis in parallel with four sides of the correction target image G _RET , + Y direction is an upward direction, and −Y direction is a downward direction. When the directions and the vertices of the quadrangle forming the correction target image G _RET are sequentially set to a, b, c, and d in the clockwise direction from the upper left, the correction target image G _RET has an upper side ab that forms a boundary line of its upper end portion. And a lower side cd forming a boundary line between the lower end and the lower end, and a cutting line ef parallel to the lower side cd divides the image into an upper half image GA and a lower half image GB, and an upper half image G.
A and the lower half image GB are moved in parallel to exchange the upper and lower positions, and the lower end cd of the lower half image GB and the upper half image GA are
The reverse image G is joined so that the upper end ab of the
_{In the} image inversion step of obtaining _REV , in the inverted image G _REV , the upper boundary line E is located above the joining line where the upper end portion ab and the lower end portion cd are attached.
_UP (X) and lower boundary line E below the joining line
_LOW (X) is set, and then a correction-required region specifying step in which a strip-like correction-required region G _AB is defined between the upper boundary line E _UP (X) and the lower boundary line E _LOW (X), Generating such that the pixel value determination process is sequentially performed while sequentially scanning the pixels in G _AB , and the images from the lower half image GB to the upper half image GA are continuously connected in the correction required region G _AB . A program for performing successive image generation steps for generating an image G _GEN .