JP3163703B2 - Document reading device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a reading apparatus for photographing a document from above, and more particularly to a document reading apparatus for detecting the degree of curving of a spatially curved document and correcting it.

[0002]

2. Description of the Related Art In an apparatus of this type, when a bound document is photographed and read from above, since the paper surface is curved at the seam of the book, the photographed data is reproduced as it is. The image is compressed in parts,
The image becomes distorted. To cope with this, conventionally, for example, as disclosed in Japanese Patent Application Laid-Open No. 62-143557, a line sensor for scanning and reading a document and a distance sensor for detecting a distance from the document surface are used for reading. In some cases, the reading pitch of the line sensor in the sub-scanning direction is changed in accordance with the detected distance, or as shown in Japanese Patent Application Laid-Open No. HEI 3-117965, according to the degree of bending of the original when reading the original. In some cases, a captured image is expanded.

[0003]

However, in the above-mentioned prior art apparatus, the reading pitch of the line sensor in the sub-scanning direction is changed when reading the original, so that the mechanical structure becomes complicated and the distance sensor is changed. The cost is high because of using. Further, in the latter device, since the density reproduction interpolation process is not performed when the photographed image is expanded, the image may not be reproduced beautifully near the seam of the document where the image expansion ratio is high. The present invention is to solve the above problem, without using a mechanical correction means of changing the reading pitch of the line sensor in the sub-scanning direction at the time of reading the original, detects the degree of bending of the original at the time of reading the original, It is an object of the present invention to provide a document reading apparatus that performs density reproduction interpolation processing when expanding an image in accordance with the result, and can reproduce an image as if a flat document was photographed.

[0004]

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention is directed to an image processing apparatus in which the surface of an original placed upward
Document consisting of imager with multiple pixels read from
Scanning means and a skew detection to detect the skew of the original
Output means, and based on the detection result of the bend detection means,
Calculating means for calculating an image expansion ratio of a pixel; and the calculating means
Number of pixels by density reproduction interpolation of each pixel output according to the output of
Original having expansion means for expanding an image by increasing
It is a reading device. In the density reproduction interpolation,
The pixel data to be interpolated is based on the pixel data on both sides.
It may be calculated based on this.

[0005]

According to the above arrangement, an original placed upwards
The original surface is read from above by
The degree of bending of the draft is detected, and each
Calculates the image expansion ratio of pixels, and calculates each image according to the calculated output.
Increasing the number of pixels by density reproduction interpolation of raw output
Is extended. Thereby, an accurate image can be reproduced. Interpolated pixel data in density reproduction interpolation is
Beautiful because it is calculated based on data
Images can be reproduced.

[0006]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings. FIG. 1 shows an external configuration of a document reading apparatus according to an embodiment of the present invention. An original 1 is placed upward on an original table 10 of an apparatus body having a desk shape, and a reading unit 2 is supported by a support member 3 at a position where the original 1 can be read from above. At a predetermined interval. This interval is a size that allows at least the readable area on one surface of the original document to be always visible, and a work space is formed between the one original surface and the reading unit 2 by this interval. The apparatus is provided with operation switches 4 and a bending correction circuit section 5 for correcting the output of the document reading means.

FIG. 2 and FIG. 3 show a detailed configuration of the reading unit 2. Image information of the document 1 is read by the image sensor 6 through the imaging lens 7 of the reading unit 2. The image sensor 6 is a semiconductor photoelectric conversion element such as a CCD, and may be an area sensor or a line sensor. FIG. 2 shows an area sensor,
FIG. 3 shows a configuration in the case of a line sensor. The bend detection board 9 measures the height of the original document 1 using a so-called triangulation method in which the reflected light emitted from the IR light emitting element 8a and reflected on the original document 1 is received by the light receiving element 8b to measure the distance. Then, the bending detection board 9 having the IR light emitting element 8a and the light receiving element 8b is scanned above the document 1 to measure a change in the height of the document 1 surface. The distance measuring method is not limited to the triangular distance measuring method, but may be a phase difference distance measuring method.

FIG. 4 is a block diagram showing a flow of electric signals of the document reading apparatus according to the present embodiment. FIG. 5 shows a series of copy operation processes using the document reading apparatus according to the present embodiment. Hereinafter, the configuration and operation of this embodiment will be described with reference to FIGS. When the document 1 is set on the document table 10 (YES in # 1) and the copy button is turned on (YES in # 2), the distance measuring circuit 22 is driven by a command from the CPU 21 in FIG. Then, the bend detection board 9 of FIG. 2 scans and measures the height of 10 points on one page of the document 1, that is, a total of 20 points on both left and right pages. The height of each point is measured by the light emitting element 8a and the light receiving element 8b of FIG. 2 by a triangulation method (# 3).

The measured height data is A / D-converted by an A / D converter 23 as shown in FIG. 4, and then sent to a spline interpolation circuit 24, where the data is subjected to cubic spline interpolation processing. The shape data of the document 1 is obtained (# 4). The shape data of the document 1 output from the spline interpolation circuit 24 is sent to the image expansion ratio calculation circuit 25, where the image expansion ratio is calculated (# 5), and the calculated expansion ratio value is prepared for execution of the image expansion process. Stored in the memory 26.

When the above processing is completed, the CPU 2 of FIG.
The CCD drive circuit 27 is driven by the instruction of the
The document 1 is imaged by the camera 8 (# 6). The captured data is A / D converted by the A / D converter 29 as shown in FIG. 4, and then sent to the image restoration circuit 30 to restore the image (# 7). Next, the image data output from the image restoration circuit 30 is sent to the image expansion circuit 31. On the other hand, the memory 26
Is also sent to the image decompression circuit 31, and an image decompression process is performed based on this data and the image data (# 8). The data output from the image decompression circuit 31 is sent to the output unit 32 and printed out (# 9), and a series of copying operations is completed.

Hereinafter, each of the above processes will be described in detail. First, the interpolation processing in the spline interpolation circuit 24 will be described. FIG. 6 is a diagram for explaining the processes of # 3 and # 4 in FIG. FIG. 6A shows a measured value of the height and its measurement position, and the envelope indicates the bending of the document. FIG.
The broken arrow in (b) is the height interpolation value calculated by the cubic spline interpolation process. By this interpolation processing, measurement data at a height of, for example, 450 points is obtained in accordance with the number of dots of the reproduction printer.

The reason why the height interpolation value is calculated by the spline interpolation process in addition to the height measurement value will be described below. FIG.
Taking the measurement positions x1 and x2 in (a) as an example, the length of the image compressed by the bending of the document and the actual length of the document are as follows.
In the triangle ABC of FIG. 7A, the length of the compressed image is the length of the straight line AC, and the actual length of the original is AB.
If the bending of the original between them can be approximated to a straight line, it can be approximated to the length of the straight line AB. Actually, when the image is decompressed, the length of the straight line AB can be easily obtained from the length of the straight line AC from AB = AC · (1 / cos θ).

However, the actual bending of the original is shown in FIG.
Since it has a curved shape like the curve AB in (b), it cannot be approximated to a straight line. However, if the interval between the height measurement points is made dense and a large amount of height measurement data is obtained, the bending of the document can be approximated to a straight line. Therefore, when measuring the height of the document, if the number of height measurement points is so small that the curvature of the document cannot be approximated by a straight line, a cubic spline interpolation process is performed based on the values of the measurement points. By calculating the height interpolation value at a plurality of points (hereinafter referred to as interpolation points) between the measurement points, a curve approximating the curvature of the document can be obtained, and the decompression process can be performed. Become. For the above reasons, the height interpolation value must be calculated by spline interpolation processing in addition to the height measurement value. However, if it is possible to obtain measurement data having the same number of heights as the number of pixels of one line imaged from the CCD in the distance measurement method of FIG. 2, the accuracy can be improved without performing cubic spline interpolation processing. High linear approximation can be obtained.

By cubic spline interpolation, x in FIG.
To find a cubic spline function from 1 to x10,
The height values at x0 and x11 in FIG. 8 are required. As shown in FIG. 8, x0 is just the joint of the book,
Process as height h ₀ = 0. Further, in order to obtain an interpolated value in the section of x0 to x1 near the seam of the original, FIG.
To indicate x _- height in ₁ requires, in this case, x _-
1 in the height h _{- 1} = -h by approximating the _1, x0
X1 is obtained. Also, since x11 is the edge of the document, it is assumed that the document and the platen are substantially horizontal,
It is treated as a height h _{1 1} = h _{1 0} in x11. As a result, a cubic spline function can be obtained in the entire range on the document surface. As an alternative, interpolation is further simplified by a quadratic spline function.

Next, the processing in the image expansion rate calculation circuit 25 will be described. , X450 high measurement data h ₁ ,..., H ₄ obtained by the spline interpolation circuit 24.
_{5 0} data to calculate the image elongation in the original. The image expansion rate is a value of AB / AC in FIG. From FIG. 7A, the image expansion rate α

(Equation 1)

## EQU1 ## By sequentially calculating this from k = 1, α (1),..., Α (449) can be obtained. This image expansion rate α (k) (k = 1,..., 44
9) is performed by the image expansion ratio calculation circuit 25.

Next, the processing in the image restoration circuit 30 will be described. The image restoration circuit 30 performs two processes, a blur restoration process and a brightness unevenness correction process. First, the blur restoration process will be described. Near the seam of the document, a large difference in height between a high portion and a low portion of the document causes image blurring. As a method for repairing this blur, there is a Laplacian process. By performing the Laplacian process on the image data, the edge of the light and shade portion of the image in the gradation gradient is emphasized, and the blur of the image is improved. further,
When the A / D conversion processing is performed on the path in which the electric signal flows as in the present embodiment, noise of the electric signal is likely to be generated. Therefore, by removing the noise by the gamma (γ) conversion processing, the image can be further reduced. Sharpening can be realized.

Next, the brightness unevenness correction processing will be described. For example, in the read image data, the image data of the i-th row and the j-th column before the luminance unevenness correction processing is set to a (i, j),
The image data after the brightness unevenness correction processing is represented by b (i, j). In the brightness unevenness correction processing, first, one line of image data a (k, j) in a white background portion having no character or picture at the upper end or lower end of the document is detected. This a
FIG. 12 shows the density distribution at (k, j). The density distribution in FIG. 12 is the density distribution in the x-axis direction in FIG. FIG.
, The point A at which the density is reduced corresponds to the seam portion of the document placed upward, and in order to correct the image data in this portion, the value of max / a (k, j) is calculated,
This value is multiplied by each row of image data a (k, j) to obtain b
The density values of (k, j) are all max, and luminance unevenness is eliminated. This is shown in FIG. This max / a (k,
When the value of j) is multiplied by all the image data corresponding to the same column, luminance unevenness is corrected for all the image data.

Assuming that the image data on the k-th row before the luminance non-uniformity correction processing is a white background, the image data a (i, j) in the i-th row and the j-th column before the luminance non-uniformity correction processing and the image data after the luminance non-uniformity correction processing The relationship with the data b (i, j) is

(Equation 2) Becomes Image restoration can be realized by the above-described blur correction processing and luminance unevenness correction processing.

Next, a method of expanding image data in the image expansion circuit 31 will be described. FIG. 9 shows [x _k ,
[x _{k +1} ]], a case where 10 pieces of image data in the section of [x _{k + 1} ] are expanded at an image expansion rate of 1.4. In this example,
_{[X k, x k + 1} ] Ten number of pixels in the interval of 1, 2 from x1 sequentially, ..., and 10. Where i
When the pixels in the row j column are expanded at the image expansion rate α, the relational expression of m = α · i-in is established, where m is the number of expanded pixels and n is the cumulative number of expanded pixels. Here, m is a value obtained by rounding the right side.

Accordingly, the relational expression of the decompression process is as follows: m = f (αi-i-n), when xy ≧ 0.5, f (x) = y + 1, when xy <0.5 f (x) = y x: Value of (αi-in) y: Value of integer part of (αi-in) When i = 1, m = f (0.
4) Since n = 0, n = 0. Hereinafter, similarly, i =
, 10 are calculated. K = 1 in FIG.
Pixel data i before expansion processing corresponding to 2,.
, 10 are shown in FIG. These processes are called [x
1, x450], it is possible to expand the image by applying it to the pixels in all the sections.

Further, as a better method, when the data in the i-th row and the j-th column is expanded to m pieces, the value of the m pixels after the expansion processing is set to the value of the pixel data in the i-th row and the j-th column. Instead, the processing shown in FIG. 11 is performed based on the pixel data of the i-th row and the j-th column and the i-th row (j + 1) column. That is, the image data is expanded by performing density interpolation on the data of the pixel to be expanded using the pixel data on the i-th row and the j-th column and the i-th row (j + 1) column while giving an average density gradation. Row i (j +
l) Image data a of the column (l = 0, 1,..., m-1)
(I, j + 1) can be calculated by the following equation.

(Equation 3) When m = 0 a (i, j + 1) = a (i, j)

Next, the reason why image restoration is performed before image expansion will be described. Since the density interpolation processing is performed when the image is expanded, the density gradient is reduced in the image data after the expansion processing, and the edges are gentle. Therefore,
Even if Laplacian processing in image restoration is performed after image expansion, edges are hardly emphasized, and the effect of image restoration is small. On the other hand, when the image is expanded after the Laplacian process, the edges are emphasized and become thick, so that it is extremely effective for restoring a thin character or a small character.

The image sensor serving as a means for reading a document may be a line sensor or an area sensor.
The range of application will be wider. Further, when performing distance measurement, as shown in FIG. 14, the distance measurement data near the seam of the document is increased,
The distance measurement data can be reduced as the distance from the vicinity of the seam increases, and thus, more accurate spline interpolation can be performed. Spline interpolation can be performed with an arbitrary number of distance measurement data. This makes it possible to perform an appropriate interpolation process according to the specifications of the device (hardware).

FIGS. 15 and 16 show a distance measuring means according to another embodiment different from the above-described embodiment. In this example, a board 90 on which a photo sensor 91 is arranged perpendicularly to the document table 10 is placed on the document table 10, and the document 1 is brought into close contact with the board 90, and the photo sensor 9
The original 1 is arranged so that the light 1 is shielded. With this configuration, the height of the document 1 can be measured in accordance with the output of the photo sensor 91.

[0026]

As described above, according to the present invention, when a document is read, the degree of bending of the document that is spatially curved is detected,
Since the image expansion processing and the density reproduction interpolation processing are performed in accordance with the result of the detection of the bend, an image can be reproduced as if a flat original were photographed. In particular, the density reproduction interpolation processing for performing the decompression processing can be performed by obtaining the average value of the density data before and after the pixel. By performing density interpolation and expanding in this way, a beautiful image can be restored.

[Brief description of the drawings]

FIG. 1 is an external configuration diagram of a document reading apparatus according to an embodiment of the present invention.

FIG. 2 is a perspective view illustrating a configuration of a reading unit.

FIG. 3 is a perspective view illustrating a configuration of a reading unit.

FIG. 4 is a block diagram showing a flow of an electric signal.

FIG. 5 is a flowchart of an image copying operation.

FIG. 6 is an explanatory diagram of spline interpolation.

FIG. 7 is an explanatory diagram of linear approximation.

FIG. 8 is an explanatory diagram of approximate data in spline interpolation.

FIG. 9 is an explanatory diagram of an image decompression process.

FIG. 10 is an explanatory diagram of an image decompression process.

FIG. 11 is an explanatory diagram of density reproduction interpolation.

FIG. 12 is a diagram showing a density distribution before correction of uneven brightness.

FIG. 13 is a diagram showing a density distribution after correcting uneven brightness.

FIG. 14 is an explanatory diagram of spline interpolation in a case where the number of measurement points is changed depending on a location.

FIG. 15 is an external configuration diagram of a document reading apparatus according to an embodiment using a photo sensor as a distance measuring unit.

FIG. 16 is a side view of a distance measuring unit using a photo sensor.

[Explanation of symbols]

 2 Reading unit 5 Bend correction circuit unit 6 Image sensor 8a IR light emitting element 8b Light receiving element 21 CPU 22 Distance measuring circuit 24 Spline interpolation circuit 25 Image expansion ratio calculation circuit 28 CCD 30 Image restoration circuit 31 Image expansion circuit

──────────────────────────────────────────────────続 き Continuation of the front page (72) Inventor Satoyuki Nakamura 2-3-13 Azuchicho, Chuo-ku, Osaka City Osaka International Building Minolta Camera Co., Ltd. (72) Inventor Shinya Matsuda Ajicho, Chuo-ku, Osaka City Chome 3-13 Osaka International Building Minolta Camera Co., Ltd. (56) References JP-A-4-199478 (JP, A) JP-A-4-331562 (JP, A) JP-A-60-223376 (JP, A) A) JP-A-58-130360 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) H04N 1/04-1/207

Claims (2)

(57) [Claims] An original document placed upward is placed on the original surface upward.
An original comprising imaging means having a plurality of pixels read from
Manuscript reading means, a bend detecting means for detecting the degree of the bend of the document, and each of the pixels based on a detection result of the bend detecting means.
Calculating means for calculating the image expansion ratio; and density reproduction interpolation of each pixel output according to the output of the calculating means.
Expansion means for expanding the image by increasing the number of pixels
Document reading apparatus characterized by having a. 2. An image interpolated in the density reproduction interpolation.
Raw data is calculated based on the pixel data on both sides.
The image reading device according to claim 1, wherein