JPH08202853A - Picture reader - Google Patents

Abstract

PURPOSE: To efficiently obtain the combination position of plural pictures which are read. CONSTITUTION: A mosaic picture 2110 obtained by converting the 8×8 picture element of a reference picture extracted from a picture 2100 into one picture element occupying the same range as the 8×8 picture element is generated. The generated mosaic picture 2110 and a picture 2000 are pattern-matched and the optimum combination position is obtained. When the optimum combination position is obtained, the mosaic pictures obtained by converting the 4×4 picture elements of the reference picture into one picture element occupying the same range of the 4×4 picture element are sequentially generated. Pattern matching with the picture 2000 is executed from the obtained optimum combination position, and the optimum combination position is updated. Above processings are repeated until one picture element in the mosaic picture and one picture element in the reference picture become the same size, and the final combination position is obtained.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image reading apparatus for reading an image, and more particularly to an image reading apparatus using a hand scanner for reading an image of an original while moving on the original for reading the image.

[0002]

2. Description of the Related Art Conventionally, there are roughly two types of scanners for sending an image of an optically read original to a connected personal computer (personal computer) or a word processor. Flatbed scanners and handy scanners) are known.

The flatbed scanner is a scanner of a type in which an original is placed on a reading surface prepared for the scanner and is read, and the scanner is necessarily larger than the original. Therefore, in order to handle a large document, it is necessary to enlarge the scanner body. Therefore, a large area is occupied around the personal computer or the like.

This is against the demand for space saving, and is a drawback of the flatbed scanner.

The other handy scanner is a type of scanner used by a user while moving on a document. The handy scanner sequentially reads the image of the document as it moves. Therefore, the handy scanner can be made smaller than the size of the document without depending on the size of the document to be read by the size of the scanner.
However, since the handy scanner reads an image while moving on a document, it is necessary to accurately detect the position where the image is read according to the movement of the scanner. That is, the handy scanner is usually provided with a rod-shaped roller that moves while rotating on the original, and the moving distance of the scanner is measured according to the rotation of this roller, and the distance for one line of the generated image is measured. Each time the scanner moves, a line sensor such as a one-dimensional CCD sensor captures an image of the original for one line, and the images of each captured line are arranged in the capture order to generate a screen of the entire original. . Or, one-dimensional CC at regular intervals
Using a line sensor such as a D sensor to capture the image of one line of the original, and to fit the image of each line captured at the position on the screen that corresponds to the position of the scanner specified by the moving distance of the scanner at that time. Creates the screen of the entire document.

Here, in the position detection by the rod-shaped roller, only movement information in the one-dimensional direction can be obtained. Therefore, if the speeds at both ends of the one-dimensional reading sensor are not constant (when there is movement including the rotational component of the scanner), the image for one line to be read is the line of the image read before that. However, the image of the entire original cannot be accurately generated.

Regarding this problem, taking as an example the case where an image of one line is taken in every movement of a fixed amount of movement,
This will be described in more detail.

FIG. 6 shows the structure of a conventional handy scanner.

In FIG. 6, reference numeral 1000 denotes a handy scanner, 2
000 is a personal computer to which a handy scanner 1000 is connected.

When the user manually runs the handy scanner 1000 on the document while pressing a reading start button (not shown), the handy scanner 100 is operated.
The illumination light source 1101 emits light from the originals arranged in a line of 0 to illuminate the surface of the original in a straight line.

The light emitted from the illumination light source is reflected on the surface of the original, and the reflected light is collected by the optical system 1102 and imaged on the one-dimensional sensor 1103. In the one-dimensional sensor 1103, the reflected light formed into an image is converted into an electric signal, and the control circuit 1 converts it into image data for one line.
To 104. Here, the one-dimensional sensor 1104
In the conversion of the reflected light into an electric signal by means of, that is, the reading of the image data for one line, the movement amount measured by one encoder which detects the rotation of the roller 1106 provided in the scanner 1000 is a constant amount ( It is assumed to be performed each time the distance (corresponding to the width of one line of the image) is reached.

The control circuit 1104 transmits the received image data for one line to the personal computer 2000 via the interface circuit 1105.

In the personal computer 2000, an interface circuit 2001 is used for a handy scanner 1000.
The image data for one line transmitted by the above is received and stored in the memory 2002.

By repeating the above operation until the reading start button is released, image data for one line is sequentially stored in the memory 2002 of the personal computer 2000 to form one screen. . That is,
Personal computer 2000 received i-th 1
The image data for a line constitutes an i-line on the screen. Now, in this case, when the scanner 1000 moves so as to draw a curve, the scanner 1 moves at a different inclination with respect to the original.
The image data for the lines are sequentially read.
In this case, two lines that are inclined with respect to each other on the original form the screen as two parallel lines in the memory of the personal computer 2000, so that it is not possible to generate an accurate image of the original. .

For this reason, in the technique disclosed in Japanese Patent Laid-Open No. 3-256448, two position sensors are provided at both ends of the reading portion of the scanner, and the position sensors respectively detect the amounts of movement in the mutually orthogonal directions. 10
When 00 moves curvilinearly, an accurate screen is generated by correcting the mutual positional relationship between the read lines in accordance with the moving amounts detected by the two position sensors.

Further, by reading the position information engraved on the sheet installed on the upper or lower part of the original together with the reading of the image data, the position on the sheet of the scanner is detected, and accordingly, the position is detected. A technique is known in which an accurate screen is generated by correcting the positional relationship between the read lines.

[0017]

DISCLOSURE OF THE INVENTION Problems to be Solved by the Invention
According to the technology described in Japanese Patent No. 56448, in reality, 2
The two position sensors cannot accurately detect the movement amount, resulting in an error. Therefore, in the generated screen, characters, straight lines, etc. are not smoothly connected, and a shift or the like is unavoidable.

On the other hand, according to the method of using the sheet in which the position information is engraved, the space saving cannot be realized because a sheet or the like in which the position information of the size of the original is recorded is required, and thus the handy scanner cannot be realized. The result is that the greatest advantage is destroyed.
Moreover, since it is necessary to install the seats one by one, the annoyance at the time of use increases.

Therefore, it is an object of the present invention to provide an image reading apparatus capable of more accurately and efficiently generating an image of a document without requiring a special sheet or the like.

[0020]

[Means for Solving the Problems] To achieve the above object,
According to the present invention, an image pickup means that moves a document on a document by a user and sequentially captures a two-dimensional image of each part of the document with the movement, and a partial range overlaps with the two-dimensional image sequentially captured last time with the movement. Control means for causing the image pickup means to pick up a two-dimensional image of a document portion so as to pick up the two-dimensional image, and the i + 1th two-dimensional image picked up by the image pickup means sequentially from the first to i-th And an image combination unit that combines the two-dimensional images captured up to 2 to a two-dimensional image, the image combination unit overlapping with the other two-dimensional image included in one of the two two-dimensional images to be combined. A reference image that is an image in the range to be extracted is extracted, and the extracted reference image is stored for each of a plurality of pixels in the extracted reference image, the gradation value having an average value of the gradation values of the plurality of pixels. of Converted to a mosaic image converted into one pixel having the same spatial size as the element, comparing the mosaic image and the other two-dimensional image, the mosaic image, the other two-dimensional image most approximate A combination position determining means for determining an inner range, and a reference image in the one image, so that the one of the images and the reference image in the other two-dimensional image is located in the range determined by the combination position determining means. There is provided an image reading device having a combination means for combining the other image.

[0021]

According to the image reading apparatus of the present invention, the combination position determination means of the image combination unit includes an image in a range included in one of the two two-dimensional images to be combined and overlapping with the other two-dimensional image. Is extracted, and the extracted reference image is stored in the same space as the plurality of pixels having the average value of the gradation values of the plurality of pixels as a gradation value for each of the plurality of pixels in the extracted reference image. Converted to a mosaic image converted into one pixel having a physical size, comparing the mosaic image and the other two-dimensional image, the mosaic image, the range in the other two-dimensional image most approximate Ask.

Then, the combining means sets the one image and the other image so that the reference image in the one image is located in the range obtained by the combination position determining means in the other two-dimensional image. Combine with the image.

That is, in the present invention, since the combination position is obtained by the so-called pattern matching method, the sequentially read two-dimensional images can be combined in the correct positional relationship even when the two-dimensional image is moved in a curved manner.
Further, in the present invention, one of the two-dimensional images to be combined is converted into an image mainly composed of a component having a low spatial frequency, and then the other two-dimensional image is compared to obtain the position where the two should be combined. In this way, if an image mainly composed of low spatial frequency components is used as an image for determining the position to be combined, high-speed processing can be expected due to its low information amount. Further, it is possible to prevent erroneous determination due to high spatial frequency components due to noise or the like.

[0024]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the image reading apparatus according to the present invention will be described below.

The configuration of the image reading apparatus according to this embodiment is shown in FIG.

In the figure, 100 is a handy scanner, and 200
Is a personal computer. Handy scanner 1
Reference numeral 00 denotes a control circuit 102, an area CCD sensor 105 for capturing a two-dimensional image, a light source 104 including a light emitting diode array for illuminating a document, and an area CCD sensor 10.
5, an optical system 106 for forming an image of reflected light of an original, an image combining means 103, a memory 107, and a position detecting means 10.
9. An interface circuit 111 is provided. Area C
The CD sensor 105 photographs only a part of the document at one time.

Further, the handy scanner 100 is placed at a target position with respect to the center of the area CCD sensor 105.
It is provided with one ball-shaped sphere 101. Ball-shaped sphere 1
In each of 01, although not shown in order to avoid complication of the drawing, a predetermined 2
Two rollers that rotate in the same direction are provided so as to rotate with the rotation of the sphere, and an encoder that detects the rotation of the roller is connected to each roller.

The ball-shaped sphere is a handy scanner 100.
Rotates with the movement of. This rotation is decomposed into mutually orthogonal components by a roller provided adjacent to the rotation, and the amount of movement of each component is detected by the encoder. The detected movement amount is in the form of pulse information (the movement amount is represented by the number of pulses), and the position detecting means 10 is used.
9 is transmitted.

On the other hand, the personal computer 200 is
The interface circuit 201 and the memory 202 are provided.

The operation of the image reading apparatus according to this embodiment will be described below.

When image reading is started, the control circuit 102 of the handy scanner 100 causes the area C to be read at regular intervals.
The image captured by the CD 105 is temporarily stored in the memory 107 as a partial image.

During the certain period, even when the expected maximum speed handy scanner 100 is moved, the area CC
D105 determines that the range of the document including a part of the range of the document photographed last time can be photographed each time.

The image combining means 103 combines the partial images in the memory 107 every time a partial image is captured or after the partial images of the entire document are completely captured to represent the image of the entire document. Create an image. This whole image is created by sequentially combining the photographed partial images. For example, each time a partial image is captured, if you want to combine the partial images captured until then, combine the partial image captured this time with the entire image created by combining the partial images captured up to the previous time, This is used as a new whole image.

The details of creating the whole image will be described below.
An example will be described in which a partial image is combined every time a partial image is captured to create an entire image representing an image of the entire original.

First, the partial image stored in the memory 107 for the first time is set as the entire image. Then, when the second partial image is stored in the memory 107, the image combination means 103 receives the movement amount of the handy scanner 100 from the position detection means 109. The movement amount of the handy scanner 10 is divided into a straight-moving component and a rotation component of the movement amount, and is obtained by the position detecting means 109 as follows.

That is, the positions of the two ball-shaped spheres with respect to the reference origin, which are obtained according to the current movement amounts of the two ball-shaped spheres 101, are expressed by (X1,
Y1) (X2, Y2), and the position when the previous image was captured is (XO1, YO1) (XO2, YO2).

Then, the straight-ahead component amount at the central position of the area CCD sensor 105 is set to ({(XO1 + XO2)-(X1 + X2)} / 2, {(YO1 + YO2)-(Y1 + Y2)} / 2).

Further, the rotation component amount is (X1-X2) ^ 2 + (Y1-Y2) ^ 2 = (XO1-XO2) ^ 2 + (YO1-YO2) ^ 2 = R ^ 2, and cos θ = 1- [{(X2-X1)-(XO2-XO1)} ^
2 + {(Y2-Y1)-(YO2-YO1)} ^ 2] / 2R
Let's say 2

Now, the image combining means 103 uses the movement amount received from the position detecting means 109 to determine the second
The relative position and angle of the partial image of the first time with respect to the partial image of the first time are obtained.

Then, the relative position thus obtained,
Depending on the angle, the overlapping range between the second partial image and the entire image when combined is obtained. In addition, an image of an appropriate size range is extracted as a reference image from the range of the second partial image that overlaps the obtained overall image. However, the entire range of the second partial image that overlaps the obtained overall image may be used as the reference image. Then, the process A is performed.

(Processing A) As shown in FIG. 2b, the 8 × 8 pixel of the reference image extracted from the second partial image (FIG. 2a) is converted into one pixel which occupies the same range as the 8 × 8 pixel. To generate a mosaic image 2110. And
The generated mosaic image is overlaid on the whole image 2000 so as to match the relative position and angle received from the position detection means 109 as shown in FIG. 2c. That is, if the movement amount processing obtained by the position detection unit 109 is correct, the range on the document corresponding to the mosaic image is set so that the range of the entire image where the mosaic images overlap is equal to the range on the corresponding document. Overlap. Then, the process B is performed.

(Process B) First, for each pixel of the mosaic image, a pixel group overlapping the pixel is extracted from the entire image. This extraction is performed by extracting a pixel in which the area overlapping the pixel of the mosaic image is half or more of the pixel as shown in FIG. 3a. Then, for each pixel of the mosaic image, the average value of the gradation of each pixel of the pixel group overlapping the pixel is taken.

Then, the sum IT, which is the difference between the average value of the gradations of the pixels of each mosaic image and the pixel group of the entire image which overlaps with it, is calculated by the equation 1.

IT = Σ ((GO) − (GN)) 2 (Equation 1) GN: gradation of pixel N of mosaic image GO: gradation of each pixel of the pixel group of the entire image overlapping N The average value Σ indicates that the sum is calculated for all pixels of the mosaic image.

Then, a) processing the mosaic image from the current position on the whole image in parallel in the vertical and horizontal directions as shown in FIGS. Perform B to obtain IT at each position. Then, the position where IT becomes the minimum is set as the reference position. Then, b) processing B is performed at the position where the mosaic image is translated in parallel in the vertical and horizontal directions on the entire image one pixel at a time from the reference position, and IT of each position is obtained. Then, if the obtained IT is smaller than the IT at the position before the parallel movement, the process of b) is performed again using the position where the IT obtained in b) is the minimum as a new reference (pattern matching is performed). Equivalent to that). On the other hand, if the calculated IT is larger than the IT at the position before translation,
The position before the parallel movement (reference position) is stored as the next final reference position, and the process C is performed.

(Processing C) A mosaic image in which the pixel size of the mosaic image is reduced to 1/4 or 1/16 as shown in FIG. 3d is generated. For example, if the pixels of the mosaic image up to now correspond to the 8 × 8 pixels of the reference image, this time, the 4 × 4 pixels of the reference image are converted to the 4 × 4 pixels of the reference image.
A mosaic image converted into one pixel occupying the same range as four pixels is generated.

Then, the generated mosaic image is superimposed on the stored final reference position on the entire image, and the process B is performed.

However, if the pixel of the mosaic image so far corresponds to 1 × 1 pixel of the reference image, the stored final reference position is set as the optimum position,
The second partial image is combined with the whole image so that the reference image portion in the partial image is located.

Then, at the boundary between the images of the entire image and the partial image, it is detected whether or not there is a deviation in which characters, straight lines, etc. do not continue smoothly, and if so, the deviation is eliminated. Rotate the partial image at a small angle, a few times with slightly different angles. If the displacement is not eliminated by this rotation, the entire character is horizontally moved pixel by pixel until the displacement is eliminated.

When the displacement is eliminated, the position conversion means 109 carries out a process of converting the partial image into a partial image having a mesh suitable for the whole image mesh (step 211). Here, the mesh defines the arrangement of pixels as shown in FIG. 4, and each profit in the mesh corresponds to a pixel.

That is, the arrangement of pixels in the rotated or moved partial image (reference numeral 502 in FIG. 4a) is not aligned with the arrangement of pixels in the entire image (reference numeral 501 in FIG. 4a). Therefore, based on the applied rotation or movement, the rotated or moved partial image is converted into a partial image (reference numeral 503 in FIG. 4b) made up of pixels arranged in the same arrangement as the pixels of the entire image.

In this conversion, the mesh of the whole image is extended and overlapped on the rotated and moved partial image, and the pixel (reference numeral 601) of the converted partial image in each mesh (reference numeral 604). Value (gradation) of the partial image before conversion (reference numeral 603 in FIG. 5) that overlaps that position (reference numeral 602 in FIG. 5)
The value is calculated by the following formula.

Gradation of pixel after conversion = Σ (area overlapping with pixel before conversion × gradation of pixel before conversion) / area of pixel Then, the obtained gradation is the floor of the pixel on the reference mesh. To the key (In the case of generating a color image, it is possible to convert the color image by obtaining this gradation for each of R, G, and B.) Finally, the converted second partial image is converted into the whole image. The connected image is created in the memory 107, and this image is used as the whole image for the third partial image. At this time, a range in which the whole image and the second partial image overlap each other occurs. Therefore, in this range, either one of the partial image and the whole image is valid, and the other image is invalid.
In this embodiment, only the range of the partial image that does not overlap with the entire image is connected to the entire image.

By repeating the above processing for each partial image, the entire image of the entire original is created in the memory 107.

Now, when the entire image of the entire original document is created on the memory 107, the control circuit 102 sends it through the interface circuit 111 to the personal computer 20.
Send to 0. In the personal computer 200, the interface circuit 201 receives this and stores it in the memory 202.

The embodiment of the present invention has been described above.

As described above, according to the present embodiment, when the handy scanner is moved, even when the hand-held scanner is moved in a curvilinear manner to which a rotation component is added, the partial images are sequentially reproduced and the image of the original is accurately reproduced. Can be spliced together.

Further, in the above description, the control circuit 102 of the handy scanner 100 has the area CCD 1 at regular intervals.
The image photographed by 05 is captured as a partial image. This is because the control circuit 102 of the handy scanner 100 causes the area CCD 105 to photograph the range of the original including a part of the range of the original taken every time. The image captured by the area CCD 105 may be captured as a partial image according to the amount of movement sent from the encoder.

In the present embodiment, the partial images are joined on the handy scanner 100 side and the completed whole image is sent to the personal computer 200. However, the partial images are directly output from the handy scanner 100 as they are. Par
Alternatively, the partial images may be sent to the personal computer 200, and the partial images may be joined on the personal computer side to obtain the final entire image.

Further, in the present embodiment, the mosaic image is generated from the partial image side to be combined with the whole image and the optimum position where the two are combined is obtained. On the contrary, the mosaic image is generated from the whole image side, You may make it obtain | require the optimal position which should combine both.

Further, in the present embodiment, as described above, when obtaining the optimum position of the combination of the partial images to the entire image, both the partial image and the entire image, or either one of them is previously calculated. You may make it convert into the enlarged image by the interpolation by intermediate value generation etc. By doing so, the position of the image can be more finely adjusted, and even if the pixel density of the read image is rough, it is possible to prevent the deviation from becoming conspicuous.

Further, when obtaining the optimum position of the combination of the partial images to the entire image, the value of each pixel of the partial image and the entire image, or either one of them is calculated in advance as the moving average value. You may make it convert into the image converted into. That is, the pixel value may be converted into an image in which the average value of the pixel values within a predetermined range around the pixel is converted. By blurring the image in this way, the high frequency components present in the image are removed,
The accuracy of pattern matching can be improved.

By the way, according to the present embodiment as described above, the read image is converted into the mosaic image to obtain the optimum position of the combination. However, this is because information is truncated at first sight, It may seem like a disadvantage in finding the optimal position for the combination.

However, in actuality, for example, when a person compares images with each other, rather than comparing the details one by one, at first, it is determined whether or not the entire target image is similar. Are compared, and then the differences in the details are compared. In other words, first compare the low spatial frequencies (rough overview), then
This is equivalent to making a comparison for a portion having a high spatial frequency. And, such comparison is a relatively efficient approach because it is hierarchical.

Therefore, in the present embodiment, learning from this, as described above, by generating a mosaic image for the read image, an image mainly composed of low spatial frequency components is obtained from the read image. Is to realize an efficient process and speed up the process by roughly determining the optimum position based on this image. This is also because it is useful for lowering the spatial frequency of the image that the image is converted into an image in which the value of each pixel is converted to the moving average value thereof in advance.

Further, as described above, by using the mosaic image, it is difficult to determine whether or not the images, which have been subtly different from each other due to noise or the like, have a minute portion (high frequency component), which has been a problem. It is less likely that In a reading element such as a CCD, the reading position of pixels is not always constant,
Even if you read the same image, if the reading angle or position shifts even a little, you can get an image with different minute parts.
Even in such a case, the present embodiment is effective.

[0067]

As described above, according to the present invention, it is possible to provide an image reading apparatus capable of efficiently producing an image of a document more accurately and without requiring a special sheet or the like. You can

[Brief description of drawings]

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

FIG. 2 is a diagram showing how images are combined by an image reading apparatus according to an embodiment of the present invention.

FIG. 3 is a diagram showing how images are combined by an image reading apparatus according to an embodiment of the present invention.

FIG. 4 is a diagram showing how images are combined by an image reading apparatus according to an embodiment of the present invention. It is a figure which shows the state of the image conversion for mesh adaptation.

FIG. 5 is a diagram showing how images are combined by an image reading apparatus according to an embodiment of the present invention. It is a figure which shows the state of the image conversion for mesh adaptation.

FIG. 6 is a block diagram showing how the conventional image reading apparatus operates.

[Brief description of reference numerals]

 100 Handy Scanner 101 Sphere 102 Control Circuit 103 Image Combination Means 104 Light Source 105 Area CCD Sensor 106 Optical System 107 Memory 109 Position Detecting Means 111 Interface Circuit 200 Personal Computer

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI technical display location G06F 15/66 355 C

Claims (5)

[Claims] 1. An image pickup unit that is moved on a document by a user and sequentially captures a two-dimensional image of each part of the document as the document moves, and a two-dimensional image that is sequentially captured last time and a partial range. Control means for causing the image pickup means to pick up a two-dimensional image of a document portion so as to pick up overlapping two-dimensional images; And an image combination unit that combines the two captured two-dimensional images into a two-dimensional image, the image combination unit including the other two-dimensional image included in one of the two two-dimensional images to be combined. A reference image that is an image in an overlapping range is extracted, and the extracted reference image has a gradation value that is an average value of gradation values of the plurality of pixels for each of a plurality of pixels in the extracted reference image. Compound Converted to a mosaic image converted into one pixel with the same spatial size as several pixels,
Comparing the mosaic image and the other two-dimensional image,
In the mosaic image, a combination position determination unit that finds the range in the other two-dimensional image that is most approximate, and a reference image in the one image is obtained by the combination position determination unit in the other two-dimensional image. An image reading apparatus comprising: a combining unit that combines the one image and the other image so that the image is located in a different range. 2. An image pickup means which is moved by a user on a document and sequentially picks up a two-dimensional image of each part of the document according to the movement, and a movement amount detection means which detects a movement amount moved by the user. Based on the movement amount detected by the movement amount detecting means, a two-dimensional image of the document portion is sequentially displayed on the image pickup means so as to pick up a two-dimensional image whose partial range overlaps with the two-dimensional image taken last time. A control means for capturing an image; and an image combining means for sequentially combining the two-dimensional images captured by the image capturing means i + 1th with the two-dimensional images captured from the first to i-th images. The image combination unit includes combination position determination means,
Combining means, wherein the combination position determining means: a) extracting a reference image which is an image in a range overlapping with the other two-dimensional image included in one of the two two-dimensional images to be combined, A) Based on the movement amount detected by the movement amount detecting means,
A range in the other two-dimensional image estimated to be a range of an image obtained by photographing the same original part as the extracted reference image is set as a candidate range, and c) the extracted reference image is n in the extracted reference image. For each pixel, the average value of the gradation values of the plurality of pixels is converted into a mosaic image converted into one pixel having the same spatial size as the plurality of pixels having a gradation value, and the mosaic image , The range of the other two-dimensional image that most closely approximates
Obtaining by searching the candidate range from the other two-dimensional image as the center of the search, the process of making the range in the other two-dimensional image a new candidate range, while sequentially decreasing the value of n. Repeated a predetermined number of times, d) The finally obtained candidate range is set as the final candidate range, and the combining means sets the reference image in the one image to be located in the final candidate range in the other two-dimensional image. An image reading apparatus characterized by combining the one image and the other image. 3. The image reading apparatus according to claim 1 or 2, wherein the combination position determining means determines the gradation value of the pixel of the mosaic image and the pixel of the mosaic image, which are obtained for each pixel of the mosaic image. The sum of the differences from the average value of the gradation of a plurality of pixels in the other corresponding two-dimensional image is the smallest, the range in the other two-dimensional image, the other of the other closest to the mosaic image. An image reading device characterized by being obtained as a range in a two-dimensional image. 4. The image reading apparatus according to claim 1 or 2, wherein the image-image combining unit is configured to combine the two two-dimensional images with each other prior to combining the two two-dimensional images by the combination position determining unit. Is provided with an interpolating unit that enlarges at least the overlapping range of the two two-dimensional images by interpolation, and the combination position determining unit performs processing on the two two-dimensional images enlarged by the interpolation. Image reading device. 5. The image reading apparatus according to claim 1 or 2, wherein the image-image combination unit is configured to combine the two two-dimensional images with each other before combining the two two-dimensional images by the combination position determination unit. For converting at least the overlapping range of the two two-dimensional images into a two-dimensional image in which the gradation value of each pixel of the two-dimensional image is a value obtained by a moving average. Is an image reading device, which performs processing on two converted two-dimensional images.