JP2003141524A - Non-contact image reading method and system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a non-contact image reading system having a means for restoring the curved surface of a thick book or the like into a plane to read it without hindrance, without being affected by the thickness or size of the subject to be read. SOLUTION: The non-contact image reading system comprises a plurality of image input cameras 10 and 11 positioned so that their respective observation ranges overlap at least partially; feature point retrieving means 22 and 21 for retrieving the feature points of the images read; a corresponding point retrieving means 23 for retrieving the correspondence between the retrieved feature points between images; a distance value measuring means 24 for measuring the value of the distance to the subject according to misalignment between the corresponding points; a plane restoration means 25 for subjecting the read image to a modeling process according to the distance value to develop the image into a plane image; and an image connection means 26 for connecting the image developed into a plane.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image reading system for reading documents such as documents and slips in offices and homes, postal items, banknotes, etc., and correcting and processing the images thereof. In a natural environment, that is, without adding a light source exclusively for an image reading system, the surface of a three-dimensional object, a page of a thick book, a piece of paper such as a bent bill, or a large-sized reading target can be contacted with high image quality without contact. The present invention relates to a non-contact image reading system that reads with high accuracy and high speed and processes the image.

[0002]

2. Description of the Related Art In recent years, digital cameras and flatbed scanners have become popular as image input devices. However, there is a big gap between the need for anyone to easily use and capture anything while seeing it, and the function and performance of a real digital camera or flatbed scanner.

In a real digital camera, in order to accurately capture the shape of an object, the optical axis of the lens must be directly opposed to the center of the object and the framing must be determined strictly. If the optical axis does not directly face the object, perspective is generated as in the perspective projection method, and an image of the correct shape cannot be obtained unless it is corrected by so-called image processing software.

Further, in the flatbed scanner, it is necessary to turn the sheet original over and set it.

Therefore, a non-contact reading technique has been attracting attention. However, in the conventional non-contact image reading system, the thickness and size of the reading object are limited. For example, when reading a bent surface of a thick book or the like, there is a problem that an image is distorted.

As a conventional technique for solving the image distortion, Japanese Patent Laid-Open No. 9-322055 proposes a method of generating a panoramic image with a swing camera.

Japanese Unexamined Patent Publication No. 6-109450 proposes a method of inputting a three-dimensional image using a plurality of cameras without blind spots.

Japanese Unexamined Patent Publication No. 7-87282 discloses a plurality of Cs.
When combining multiple images taken using CD elements on a large screen, the images near the joints overlap in order to avoid the effects of distortion of the taking lens itself and to connect the joints smoothly. Taking lens and CCD
We propose a structure that arranges multiple units consisting of.

Japanese Unexamined Patent Publication No. 10-164566 discloses a rectangle for displaying a plurality of images as one image when a monitor television monitors an image signal obtained from a plurality of cameras installed adjacent to a ceiling. We have proposed a method to generate a composite video signal by overlapping areas.

[0010]

[Patent Document 1] Japanese Unexamined Patent Publication No. 9-32205
According to the technique disclosed in Japanese Patent Publication No. 5, a plurality of images can be pasted together to generate a large screen image. However, the observation image is generated only in consideration of the three-dimensional position of the camera, and the distance to the reading object is not obtained.

The technique disclosed in Japanese Patent Laid-Open No. 6-109450 has a function of measuring the distance to an object using a plurality of cameras. However, the main purpose is to observe the three-dimensional object as it is without any blind spots, and the surface observed obliquely was captured as an image as it was oblique.

In the technique disclosed in Japanese Patent Laid-Open No. 7-87282,
Although a method for avoiding the influence of the distortion aberration of the photographing lens itself is shown, there is no description about eliminating the influence of the thickness or bending of the reading object.

The technique disclosed in Japanese Unexamined Patent Publication No. 10-164566 also shows a method of compositing rectangular areas by overlapping them in order to display a plurality of images as one image. There is no consideration.

An object of the present invention is to provide a non-contact image reading method and a non-contact method which are provided with means for restoring a bent surface of a thick book or the like to a flat surface and reading without trouble, without being affected by the thickness or size of an object to be read. An image reading system is provided.

[0015]

In order to achieve the above object, the present invention searches for characteristic points of an image read by a plurality of image input cameras arranged so that at least a part of observation areas overlap. , The correspondence of the retrieved feature points is searched between the images, the distance value from the displacement of the corresponding points to the object is measured, the read image is modeled based on the distance value, developed into a plane image, and developed. We propose a non-contact image reading method that connects two-dimensional images.

According to the present invention, the characteristic points of the images read by the plurality of image input cameras arranged so that only the central portion of the observation region overlaps are searched, and the correspondence of the searched characteristic points is determined between the images. The distance value from the shift of the corresponding points to the object in the central portion is searched, the read image is modeled based on the distance value in the central portion, developed into a planar image, and non-overlapping portions are measured. We propose a non-contact image reading method that performs modeling processing based on the distance value predicted from the relationship with the distance value of the central part, develops it into a planar image, and connects the developed planar images.

In order to achieve the above-mentioned object, the present invention comprises a plurality of image input cameras arranged so that at least a part of the observation areas overlap each other, and a feature point search means for searching a feature point of the read image. , Corresponding point searching means for searching the correspondence of the searched feature points between images, distance value measuring means for measuring the distance value from the deviation of the corresponding points to the object, and modeling processing of the read image based on the distance value We propose a non-contact image reading system consisting of plane restoration means for developing a plane image and image connecting means for connecting the plane developed image.

The present invention also provides a plurality of image input cameras arranged so that only the central portion of the observation region overlaps with each other, feature point search means for searching the feature points of the read image, and the searched feature points. Corresponding point search means for searching the correspondence between images, distance value measuring means for measuring the distance value from the deviation of the corresponding points to the object in the central part, and modeling processing of the read image based on the distance value in the central part Connect the flattened image to the flattened image reconstruction unit that performs modeling processing based on the distance value predicted from the relationship with the measured distance value of the central portion and develops it into a flattened image We propose a non-contact image reading system consisting of an image connecting means.

It is possible to provide interval adjusting means for moving the plurality of image input cameras only in one-dimensional direction, and for moving the plurality of image input cameras apart from each other to set to the large-size input mode, and to bring them closer to each other to set the distance value measurement mode to the object.

It is also possible to have a structure in which the plurality of cameras are mounted on both ends and a boom which is extendable and retractable, and one column which supports the midpoint of the boom from the document table are supported.

The plurality of image input cameras are included in a non-contact image reading device that has a document table in each and can operate independently. One of the non-contact reading devices is a combination of a plurality of non-contact reading devices that can operate independently. It is also possible to provide a control means for operating the remaining non-contact readers in cooperation with the activation means of.

In any of the above non-contact image reading systems, a camera position calibration mark can be provided on the document table of each image input camera.

Further, a height calibration chart which can be freely placed may be provided on the original table of each image input camera.

The image connecting means includes means for generating an image having no specular reflection by laminating the corresponding images of the other cameras when an image captured by one image input camera has a gradation jump due to specular reflection. To do so.

[0025]

BEST MODE FOR CARRYING OUT THE INVENTION Next, referring to FIGS.
An embodiment of a non-contact image reading system according to the present invention will be described.

[0026]

First Embodiment FIG. 1 is a functional block diagram showing the configuration of a non-contact image reading system according to the present invention. The image reading system according to the first embodiment includes image input cameras 10 and 11 as image reading means, and document plates 12 and 13.
And the control means 20. In this image reading system, the image input cameras 10 and 11 are used for manuscript plates 12 and 13, respectively.
The document above is read, the control means 20 performs image processing, and the processing result is output device 27 such as a display or a printer.
Send to.

The image input cameras 10 and 11 include a document table 1
This is a device for reading a document placed with its face up to Nos. 2 and 13 in a non-contact manner.

Originally, the image input cameras 10 and 11 are devices which independently read the originals on the original plates 12 and 13, but here, at least the photographing range is set so that the boundary portions can be read in duplicate. Part of them are placed one above the other. The document placed on the boundary is read by the image input camera 10 and the image input camera 11. If it can be determined which portions of the images read by the two image input cameras 10 and 11 correspond to each other, the boundary portions can be accurately connected.

When the object to be read is a three-dimensional object, in order to connect the images more accurately, the height is recognized from the images of the two cameras by the stereo distance measuring process, and the height on the original platen is recognized based on the recognized height. Create a flat image projected on.

Therefore, the control means 20 has a feature point search means A21 and a feature point search means B24 for searching the feature points of the images read by the image input cameras 10 and 11, and how the searched feature points are related to each other. Corresponding point search means 22, a distance value measuring means 23 that measures a distance value from the shift of the corresponding points, and a plane restoration means that performs modeling processing on the read image based on the distance value and develops it into a plane image. 25 and an image connecting means 26 for connecting the respective images developed on the plane.

FIG. 2 is a block diagram showing a hardware configuration when the function of the control means 20 in the non-contact image reading system of FIG. 1 is realized by a computer. The processor 16 monitors the operation of a reading start button (not shown) in the image input camera 10 via the input adapter 15. When the reading start button is operated, the processor 16 issues a reading instruction not only to the image input camera 10 but also to the image input camera 11. The images read by the image input cameras 10 and 11 are stored in the memory 19 via the input adapter 15. The processor 16
When the reading is completed, the image is transferred to the computer 18 via the output adapter 17. Computer 18
Processes the image and accurately connects the images read by the image input cameras 10 and 11.

FIG. 3 is a flow chart showing the processing procedure in the computer 18 of FIG. When the reading by the image input cameras 10 and 11 is completed, the computer 18
When the image is transferred to the internal memory, the process starts in step 30.

In the feature point search process in step 31, the image captured by the image input camera 10 is binarized, and a block in which white pixels and black pixels are uniformly included in a block of 16 pixels × 16 pixels is extracted. In the corresponding point search process of step 32, 16 pixels × 1 found in the feature point detection process 31
The pattern near the 6-pixel block is searched for in which position in the image read by the other image input camera 11.

In the distance value measuring process of step 33, triangulation is carried out from the positional relationship between corresponding points and the known photographing position of the camera to measure a three-dimensional position. Step 34
Then, the above processing is repeated until the distance measurement of the feature points is completed.

After the distance values of all the feature points are measured, in the plane restoration processing of step 35, a model of the read object is created from the measured three-dimensional distances, divided into triangular patches, and these triangular patches are formed into a plane. Unfold. After being developed into a two-dimensional image, in the image connection processing of step 36, the developed images are connected to one sheet and combined.

FIG. 4 is a diagram showing the relationship between the camera and a thick book when reading a thick book. FIG. 4A shows a document table 1
The thick book 40 placed on Nos. 2 and 13 is attached to the image input camera 1
The state of reading at 0 and 11 is shown. FIG. 4B shows an image read by the image input camera 10, and FIG.
Indicates an image read by the image input camera 11. In the case of FIG. 4B, the thick book 40 is photographed mainly on the left, and the right part is not photographed. In the case of FIG. 4C, the image is taken centering on the right side of the thick book 40, and part of the left side is not taken.

A point 41 on the original table 12 and a point 42 on the thick book 40 shown in FIG. 4 (a) are in the vertical direction in the image read by the image input camera 10 as shown by 41L and 42L in FIG. 4 (b). Are observed side by side, but in the image read by the image input camera 11, the images are observed laterally displaced as 41R and 42R in FIG. This is because the visible position differs depending on the height. Stereo distance measurement uses this shift.

FIG. 5 is a diagram for explaining the principle of distance measurement. FIG. 5A schematically shows this shift. If the positions of the observation points 41 and 42 seen from the image input cameras 10 and 11, that is, the angles viewed from the image input camera 11 can be measured, the position of the observation points in the height direction can be accurately calculated.
As shown in FIG. 5B, the heights of all points can be calculated within the shooting range of the image input cameras 10 and 11.

FIG. 6 is a diagram showing a method of restoring an original planar image from a bent object to be read. Figure 6
(a) has shown the planar image image | photographed. Figure 6 (b)
Shows an example in which a two-dimensional image is made into a three-dimensional model by referring to distance measurement data, FIG. 6C shows an example in which it is divided into triangular polygons, and FIG. An example of the plane development is shown. In this way, if the three-dimensional position of the feature point is known, it is possible to develop it on the plane estimated as the original shape.

[0040]

Second Embodiment FIG. 7 is a diagram for explaining an image restoration method when it is impossible to obtain all three-dimensional positions.
FIG. 7A shows an example in which all three-dimensional positions cannot be obtained. If the observation target can be covered with multiple image input cameras, all three-dimensional positions can be obtained, but if the target is large, it is expected that there will be an area where only one image input camera can capture the image. It When the thick book 40 is photographed by the image input cameras 10 and 11, since the central portions overlap, the distance can be measured by stereo processing. However, the distance information cannot be obtained because there is only one camera image at both ends. in this case,
Based on the distance information measured in the overlapping area 50, the plane restoration is executed only from the shape of the image. As a premise, it is assumed that the paper has elasticity and the distance value changes gently. For example, a rectangular page image is obtained by stretching the image of the area having only the image shape in the vertical direction so as to match the vertical length of the page in the portion where the distance value in FIG. 7B can be used.

[0041]

[Third Embodiment] FIG. 8 is a diagram showing a method of restoring an original planar image when specular reflection occurs on an object to be read. A specular reflection portion 51T exists in the left portion 40L of FIG. In the specular reflection section 51T, since the illumination is totally reflected and taken into the image input camera 10, the input value exceeds the maximum value and is saturated, and a correct image is not observed.
Similarly in the right portion 40R of FIG.
There is T. Such specular reflection frequently occurs when the object to be read has a rounded three-dimensional shape and causes image quality deterioration. Therefore, in the present invention, FIG.
As shown in, when connecting the images after the plane restoration, if the specular reflection portion is detected, the portion corresponding to the specular reflection is restored with the image on the side without the specular reflection.

FIG. 9 is a diagram showing the configuration of a non-contact image reading system in which a plurality of image input cameras are combined and a double-sized object can be read.

FIG. 9 (a) shows the apparatus as viewed from above. The image input cameras 10 and 11 are respectively A
Designed to input 4 size images,
You can read A3 size by combining two units and shooting.

Generally, when a plurality of devices are combined, inconveniences often occur such that the support shaft of the camera interferes with the operation. In this non-contact image reading system, each device is designed to have a symmetrical shape so that there is no discomfort even if a plurality of devices that originally operate independently are combined. Also in this case, since the central portion can be read redundantly, three-dimensional distance measurement can be performed. Platen 12, 13
Reference marks M10-M21 are arranged in the peripheral portion of. The image input camera 10, using the reference marks M10-M21 as reference points when connecting two A4 size images and reading A3 size or when measuring a three-dimensional distance,
By calibrating the height and posture of 11, a high-definition composite image can be obtained.

FIG. 9B shows two image input cameras 10,
In this example, the intervals of 11 are closely arranged. In this example, almost only A4 size can be read, but since overlapping images are obtained over the entire surface, all three-dimensional distances within A4 size can be measured.

[0046]

[Fourth Embodiment] FIG. 10 shows the distance between two image input cameras.
FIG. 6 is a diagram showing a structure of a document table for smoothly adjusting (one-dimensional distance).

FIG. 10 (a) shows a structure capable of simplifying the one-dimensional movement of the two devices such that the distance between the two devices is reduced or increased. In order to facilitate movement while maintaining parallelism when two document plates are combined and the distance between the devices is adjusted, a fitting portion is formed on each of the document plates. It is possible to further increase the number of divisions to form the comb-like portions for maintaining the parallelism.

FIG. 10 (b) shows an example in which the photographing positions of the image input cameras 10 and 11 are changed by, for example, expanding and contracting the columns while the original table remains unchanged. Also in this case, if the movement of the supporting column is restricted so that the reading position moves only to the left and right, the processing time for image correction in the subsequent stage can be shortened.

[0049]

[Fifth Embodiment] FIG. 11 is a view showing a mechanism for supporting two cameras with one support and adjusting the heights and intervals of the cameras in order to ensure workability. Embodiment 5
Shows an example in which one support is used so that the support mechanism of the image input cameras 10 and 11 does not hinder work as much as possible. Arms are extended from one column toward the center of the original plate, and image input cameras 10 and 11 are attached to both ends of a boom that can extend and contract to the left and right. Therefore, the support mechanism of the image input cameras 10 and 11 does not hinder work such as page turning of the object to be imaged, and the efficiency of image reading is improved.

[0050]

Sixth Embodiment FIG. 12 is a diagram showing a calibration procedure using a reference mark. FIG. 12A shows a processing procedure of calibration using the reference mark, and FIG. 12B is a perspective view showing an example of the height calibration chart.

In the processing procedure of FIG. 12A, in the mark position detection in step 61, the image input cameras 10 and 11 are detected.
Each time, the mark on the platen for the own camera is read, and then the mark on the platen for another camera is read. Whether or not the mark is for the own device can be easily identified because the marks are oriented in the same direction. It is also possible to change the mark for each position. Then, the position can be detected easily and without error even if the document table is expanded or contracted.

In the measurement of step 62, the coordinate position, rotation angle and height of the own camera are measured. Next, the coordinate position, rotation angle, and height of the camera are detected from these mark positions.

In the height calibration process of step 63, the process shown in FIG.
Using the height calibration chart 69 shown in 2 (b), the distance in the height direction is calibrated.

[0054]

Seventh Embodiment FIG. 13 is a diagram showing the configuration of a non-contact image reading system in which the input adapter is composed of two adapters. The only difference from the hardware of FIG. 2 is that the input adapters 15A and 15B are separated. If the processor 16 has a sufficient bus, it is only necessary to use a plurality of designed adapters 15, and it is not necessary to design the new input adapter 15 of FIG.

[0055]

Eighth Embodiment FIG. 14 is a diagram showing the configuration of a parallel non-contact image reading system including a computer. Although the total device cost is high, the amount of design change for multiplexing the image input system is significantly reduced. The communication method may be any of a master / slave method, a collision detection method, a token method, an interrupt method, and the like. Since the number of connected devices is limited, for example, when the master / slave system is adopted, when the reading button of the camera 10 is operated, the computers communicate with each other, and the computer on the side where the button is operated becomes the master, and image input,
Issue data transfer instructions. The slave computer transfers the read image to the master according to the instruction. With such a configuration, for example, a mode for inputting a plurality of images in parallel with an input counter for business use, a mode for inputting a large-sized original, and a mode for inputting an image of a three-dimensional object are used for the original of the apparatus. It can be reconfigured freely just by adjusting the base and camera position.

The combination of the non-contact reading devices is not limited to two, and may be three, four or more.

In the stereo processing, the binarized images are used for matching, but if the processing capability is allowed, the accuracy can be further improved by using multivalued data.

The movement of the document table and the movement of the camera position are not limited to the one-dimensional direction. If the processing time does not matter, the camera position may be moved in a circle and the distance between the devices may be changed.

A camera equipped with a two-dimensional image pickup device or a one-dimensional line sensor can be used for capturing an image.

[0060]

According to the present invention, a plurality of image input cameras arranged so that at least a part of the observation areas are overlapped with each other, feature point search means for searching the feature points of the read image, and the searched image are searched. Corresponding point searching means for searching the correspondence of feature points between images, distance value measuring means for measuring the distance value from the deviation of the corresponding points to the object, and the read image is subjected to modeling processing based on the distance value to form a planar image. Since a non-contact image reading system comprising a plane repairing means for developing and an image connecting means for connecting images developed on a plane is provided,
Without being affected by the thickness or size of the object to be read,
The bent surface of a thick book or the like can be restored to a flat surface and read without any trouble.

[Brief description of drawings]

FIG. 1 is a functional block diagram showing a configuration of a non-contact image reading system according to the present invention.

2 is a block diagram showing a hardware configuration when a function of a control unit in the non-contact image reading system of FIG. 1 is realized by a computer.

FIG. 3 is a flowchart showing a processing procedure in the computer of FIG.

FIG. 4 is a diagram showing a relationship between a camera and a thick book when reading a thick book.

FIG. 5 is a diagram illustrating the principle of distance measurement.

FIG. 6 is a diagram showing a method of restoring an original planar image from a bent object to be read.

FIG. 7 is a diagram illustrating an image restoration method when it is impossible to obtain all three-dimensional positions.

FIG. 8 is a diagram showing a method of restoring an original planar image when specular reflection occurs on an object to be read.

FIG. 9 is a diagram showing a configuration of a non-contact image reading system in which a plurality of image input cameras are combined to read a double-sized object.

FIG. 10: Distance between two image input cameras (one-dimensional distance)
FIG. 7 is a diagram showing a structure of a document table for smoothly adjusting the document.

FIG. 11 is a diagram showing a mechanism for supporting two cameras with one support and adjusting the height and interval of the cameras to ensure workability.

FIG. 12 is a diagram showing a calibration procedure using a reference mark.

FIG. 13 is a diagram showing a configuration of a non-contact image reading system in which an input adapter is composed of two adapters.

FIG. 14 is a diagram showing a configuration of a non-contact image reading system including a computer in parallel.

[Explanation of symbols]

10 image input camera 10a reading means 10b Read start button 11 Image input camera 11a reading means 11b Read start button 12 Platen 13 Platen 15 Input adapter 16 processors 17 Output adapter 18 Computer 19 memory 20 Control means 21 Feature point search means 22 Corresponding point search means 23 Distance measuring means 24 Feature point search means 25 Plane repairing means 26 image composition means 40 thick books 41 observation points 42 observation points 45 observation points 46 observation points 50 overlapping areas 51 Specular reflection part 52 Specular reflection part 69 Height calibration chart M10-M21 standard mark

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI theme code (reference) G06T 7/60 180 G06T 7/60 180B H04N 1/387 H04N 1/387 (72) Inventor Shinichi Shinoda Ibaraki Hitachi City Omika-cho 7-1-1, Hitachi Ltd. Hitachi Research Laboratory (72) Inventor Yoshiharu Konishi 1 Ikegami, Haruoka-cho, Owariasahi-shi, Aichi F-term (reference) 5B047 AA01 BA02 BB06 BC15 CA12 CA14 CB09 CB23 5B057 AA11 BA02 BA19 CA12 CA16 CB12 CB16 CC03 CD11 CD20 CE09 CE10 DA07 DB03 DC05 DC32 DC36 5C076 AA40 BA06 5L096 AA09 BA20 CA05 DA01 FA09 FA66 FA69 GA08

Claims (10)

[Claims] 1. A feature point of an image read by a plurality of image input cameras arranged so that at least a part of an observation region is overlapped, a correspondence of the retrieved feature point is searched between images, and correspondence is performed. A non-contact image reading method in which the distance value from a point shift to the object is measured, the read image is modeled based on the distance value, developed into a planar image, and the developed planar images are connected. 2. A feature point of an image read by a plurality of image input cameras arranged so that only a central portion of an observation region overlaps is searched, and correspondence of the retrieved feature point is searched between images, and correspondence is performed. The distance value from the deviation of the point to the object in the central part is measured, the read image is modeled based on the distance value in the central part and developed into a planar image, and the non-overlapping part is the measured distance in the central part. A non-contact image reading method that performs modeling processing based on the distance value predicted from the relationship with the value, develops it into a planar image, and connects the developed planar images. 3. A plurality of image input cameras arranged so that at least a part of an observation area overlaps, a feature point search means for searching a feature point of the read image, and an image showing correspondence between the searched feature points. Corresponding point searching means for searching between the two, distance value measuring means for measuring the distance value from the deviation of the corresponding points to the object, and plane restoration means for modeling the read image based on the distance value and developing it into a planar image. , A non-contact image reading system comprising image connecting means for connecting images developed on a plane. 4. A plurality of image input cameras arranged so that only the central portion of the observation region overlaps, a feature point search means for searching feature points of the read image, and an image showing correspondence between the searched feature points. Between the corresponding point searching means for searching between the corresponding points, the distance value measuring means for measuring the distance value from the shift of the corresponding points to the object in the central part, the read image is subjected to modeling processing based on the distance value in the central part An image that connects the images developed on a plane with a plane restoration means that develops a plane image by modeling based on the distance value predicted from the relationship with the measured distance value of the center part that does not overlap with the image developed A non-contact image reading system consisting of a connection means. 5. The non-contact image reading system according to claim 3, wherein the plurality of image input cameras are movable only in a one-dimensional direction, and when the plurality of image input cameras are separated, a large format input mode is set. A non-contact image reading system provided with a space adjusting means for setting a distance value measuring mode to an object. 6. The non-contact image reading system according to claim 3 or 4, wherein the plurality of cameras are mounted on both ends of the boom, and the boom is extendable.
A non-contact image reading system, comprising: one column supporting the midpoint of the boom from a document table. 7. The non-contact image reading system according to claim 3 or 4, wherein the plurality of image input cameras are included in a non-contact image reading device that has an original table for each and can operate independently. A non-contact image reading system comprising a plurality of independently operable non-contact reading devices, and a control means for causing the activation means of one of the non-contact reading devices to coordinately operate the remaining non-contact reading devices. 8. The non-contact image reading system according to claim 3, wherein a camera position calibration mark is provided on a document table of each image input camera. . 9. The non-contact image reading system according to claim 3, further comprising a height calibration chart that can be placed on a document table of each image input camera. Non-contact image reading system. 10. The non-contact image reading system according to any one of claims 3 to 9, wherein the image connecting unit causes gradation skipping due to specular reflection in an image captured by one image input camera. A non-contact image reading system including means for laminating corresponding images of other cameras to generate an image without specular reflection when the above-mentioned case.