JPH0740719B2 - Document reader - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a document reading device that samples and reads a document image at predetermined intervals.

[Prior Art] Generally, a document image input unit used in a digital copying machine or a facsimile device, or a document reading device used to input an image drawn on a document into a data processing device, scans the document image in the main scanning direction. Each pixel is decomposed into pixels in the sub-scanning direction at a predetermined resolution, each pixel is photoelectrically converted into digital data, and image data is formed and output in time series in main scanning line units.

Normally, scanning the image in the main scanning direction and dividing it into pixels is a CCD.
(Charge-coupled device) A line image sensor such as a line image sensor is used, and scanning of an image in the sub-scanning direction and decomposition into pixels are performed by relatively moving a reading line by the line image sensor.

In such a document reading device, the document image is formed on the light receiving surface of the line image sensor by an optical system including a mirror and an optical lens, and therefore, the MTF (Moduqation Transfer Fun) of the image read by the line image sensor is formed.
However, there is a problem that the scanned image is blurred and the read image is blurred.

Therefore, in order to obtain a high-quality binary image, when pixels are photoelectrically converted, they are converted into multi-bit data, and the edges of the image are passed through, for example, an MTF correction filter (digital filter) as shown in FIG. 14 (a). Was emphasized and the image was corrected.

This MTF correction filter corresponds to two pixels adjacent in the main scanning direction of a correction pixel located in the center (hereinafter referred to as a pixel of interest) and two pixels adjacent to the pixel of interest in the sub-scanning direction. The calculation is performed by multiplying the above coefficients and adding them, and the sum of all the coefficients is set to 1 so as not to be involved in the low frequency component. Therefore, as shown in FIG. 14 (b), this MTF scan filter has a DC component gain of 1 and rapidly increases toward the Nyquist frequency fn (Nyquist limit) corresponding to the sampling interval of this image reading apparatus. It has a high-pass characteristic that makes it stand up.

Therefore, it is possible to increase the sharpness (definition) of the image deteriorated by the optical system and enhance the density difference between the adjacent pixel and the target pixel, and it is possible to correct the blurred image.

By the way, in recent years, the recording density of an image recording apparatus (for example, an optical writing printer) that binary-codes an image has been improved, and accordingly, a document reading apparatus capable of reading an image with high resolution is required. It has become to.

As described above, in a document reading device capable of reading a document image with high resolution, for example, if the resolution is twice as high as that of the conventional one, the area occupied by 1/4 and 1 pixel becomes very small, so that the MTF further deteriorates. In the output signal of the line image sensor, the MTF is reduced to about 15-20%.

In order to improve this, it is necessary to use an optical system having a high MTF (particularly a lens), but such an optical system is very expensive and the cost of the entire document reading apparatus increases, which is not preferable.

Therefore, it is conceivable to correct the image using the MTF correction filter as described above. However, when the MTF correction filter having the high-pass characteristic that maximizes the gain of the Nyquist frequency as in the conventional case is used, It causes inconvenience.

First, for example, when the MTF is 20%, as shown in FIG.
As shown in (b), there are black and white pixels of an image that changes to black and white at intervals corresponding to the reading resolution (for example, if the resolution is 16 dots / mm, there are 8 white pixels and 8 black pixels per mm). Even if there is a 100 level difference between the two levels, the level difference between the black and white pixels is only 20 at the signal level after being read by the line image sensor. Therefore, the output signal of the line image sensor whose frequency component is near the Nyquist limit is easily affected by noise in the signal system, and the reliability of the signal itself is low in that frequency region.

Therefore, as described above, when the MTF correction filter that maximizes the gain of the Nyquist frequency is used, the frequency component of the unreliable part is emphasized.
For example, background stains on the original are emphasized, which causes noise in the read image, or causes so-called notch noise.

Moreover, since the image data is read in high resolution, the data amount of the image data becomes very large, and high-speed analog processing is required to process the image in real time. Deteriorates and noise particularly in the vicinity of the Nyquist frequency becomes noticeable, and the above-mentioned inconvenience is more likely to occur.

[Purpose] The present invention has been made in order to solve the above-described inconvenience of the conventional technique, and an object of the present invention is to provide an original reading device capable of reading an original image with high quality and high resolution.

[Structure] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 shows a document reading apparatus according to an embodiment of the present invention.

In the figure, the contact glass 1 on which the read document is placed is illuminated by the light source 2, and the reflected light from the image surface of the read document passes through the mirrors 3, 4, 5, 6, 7 and the lens 8. An image is formed on the light receiving surface of the line image sensor 8.

Further, the light source 2 and the mirror 3 are mounted on the scanning body 10 which moves the lower surface of the contact glass 1 in parallel with the contact glass 1 in the sub-scanning direction, and the mirrors 4 and 5 are interlocked with the traveling body 10 to be 1/2.
It is mounted on the traveling body 11 that moves in the sub-scanning direction at the speed of, and the image surface of the read document is scanned in the sub-scanning direction by moving the optical system.

The image reading resolution in this embodiment is 16 dots / m.
It is set to m and can read A3 size originals. Therefore, as the line image sensor 9, a 500-pixel CCD line image sensor is used.

FIG. 2 shows an example of the control system of the apparatus shown in FIG.

In the figure, the line image sensor 9 is driven by a sensor driver 21, the output signal of the line image sensor 9 is amplified by an amplifier 22, and the shading correction / analog / digital conversion unit 23 causes unevenness in the illuminance of the light source and the line image sensor 9 The variation of the sensitivity of the element is corrected and converted into digital data of a predetermined number of bits (for example, 6 bits or 8 bits), and the MTF correction unit 24 performs the MTF correction process described later, and then the main scanning magnification changing unit. Entered in 25.

The main-scanning scaling unit 25 scales the image by performing logical operation processing or the like corresponding to the reduction ratio of the image in the main-scanning direction, and the output data thereof is set to a predetermined threshold value by the binarization processing unit 26. Are binarized by and output to the next stage device via the output circuit 27 as black and white binary data. In addition, this binarization processing unit
The threshold value of 26 may be dynamically changed according to the density of the image, and halftone processing for converting to pseudo halftone may be performed.

By the way, the line image sensor 9 used in this embodiment is provided with 5000 elements, and even-numbered elements and odd-numbered elements are separately provided from the main scanning start position so that the reading speed can be increased to some extent. It has been read in a different system.

Therefore, an image signal EV output from the line image sensor 9 as a signal of an even-numbered pixel and an image signal OD output as a signal of an odd-numbered pixel in a time series complementary to this image signal EV are The amplifiers 22a and 22b are separately amplified, and the output stages of these amplifiers 22a and 22b are connected and combined into one to perform shading correction / analog /
It is added to the digital conversion unit 23.

Thus, even-numbered image signal EV and odd-numbered image signal OD
And are respectively amplified by separate amplifiers 22a and 22b, the following inconvenience occurs due to the difference in characteristics of these amplifiers 22a and 22b.

In the shading correction / analog / digital conversion unit 23, the level when the reference white image (not shown) is read is set to 1, and the image signals EV and OD output from the amplifiers 22a and 22b are set relative to the level. Convert to a ratio. Therefore, the same white level as the reference white image is displayed on the image signal EV and the image signal EV.
OD and OD are converted into digital data having the same value (for example, all digits are data “1”), and there is no inconvenience caused by the difference in characteristics of the amplifiers 22a and 22b.

In addition, a black image of a level (high density, low brightness) that is surely judged to be black by the binarization processing unit 26 is converted into digital data of almost the same value by the shading correction / analog / digital conversion unit 23. Since the image is converted, there is no inconvenience caused by the difference in the characteristics of the amplifiers 22a and 22b even in the image of the high density portion.

However, in the intermediate level region from black to white, the value of the output data of the shading correction / analog / digital conversion unit 23 corresponding to the brightness of the pixel is, for example, about the image signal EV output via the amplifier 22a. Is as shown by the curve CV1 in FIG. 4, and the image signal OD output through the amplifier 22b is as shown by the curve CV2 in the figure, and they are different.

Therefore, in particular, the threshold value SH set in the binarization processing unit 26
When there is a large difference between the two in the periphery, even-numbered pixels in the main scanning direction in the image of that level are determined to be black, while odd-numbered pixels are determined to be white, which causes notch noise, etc. .

Further, when MTF correction is performed using the conventional MTF correction filter that emphasizes the Nyquist frequency shown in FIGS. 14A and 14B, the difference in signal level due to the difference in such characteristics is emphasized. Therefore, the above-mentioned inconvenience is increased, which is not preferable.

Therefore, the present inventor has the characteristics shown in FIG. 5, has a gain peak at a frequency slightly smaller than the Nyquist frequency fn (here, 3/8 of the sampling frequency), and does not have much in the low frequency component. We have found that the inconvenience can be solved by using a bandpass filter, which has high sensitivity to high frequency components, without being involved in the MTF correction filter.

That is, by using the MTF correction filter having such a bandpass characteristic, it is possible to prevent the signal component of the Nyquist frequency whose reliability is lowered due to the fact that the MTF becomes very small, from being emphasized. It is possible to eliminate the influence of periodic level fluctuations caused by using separate amplifiers 22a and 22b for even-numbered and odd-numbered amplifiers in the scanning direction.

Further, as a matter of course, the low frequency component passes through the MTF correction filter as it is, and the signal component near the frequency component of the gain peak is emphasized, so that the blur of the image is corrected.

In FIG. 5, the horizontal axis u represents the frequency represented by the frequency 1 corresponding to the sampling interval of the original image. Therefore, the peak of this MTF correction filter corresponds to the sampling frequency when the sampling interval is 12 dots / mm.

Now, the MTF correction filter having such characteristics can be formed as follows.

First, the frequency characteristics of the optical system of the optical system in the main scanning direction and the sub scanning direction are obtained.

There are several methods for obtaining this frequency characteristic, but here, the line spread function h (x) is measured, and the line spread function h (x) is Fourier-transformed to obtain. Here, the line spread function h (x) indicates the spread of read data in the line width direction when a line with a thickness of P is read, where P is the sampling pitch, and represents the impulse response of the electrical system. Equivalent to.

As shown in FIG. 6A, the line-spread function h (x) in the main scanning direction is a straight line having a width P (in this case, 1/16 (mm)) in the direction orthogonal to the main scanning direction. When drawing this straight line and reading this line, it is obtained by measuring the level of the pixel reading that line and the level of the pixels that are continuous from that pixel in the left-right direction. For example, as shown in FIG. It became something like. Therefore, this line spread function h (x)
The function H (u) obtained by Fourier transform of is as shown in FIG.

In FIG. 7 (a), the pixel reading the straight line is set as the reference position (“0”) at the distance x, and the level at that position is set to “1”, and the levels at other positions are standardized. Further, the distance x is shown as a coordinate value when normalized to P = 1. Further, in FIG. 3B, the frequency corresponding to the sampling pitch P is set to 1 and the frequency is standardized.

The line-spread function h (x) in the sub-scanning direction is the level when a straight line having a width P is drawn on a document in the direction orthogonal to the sub-scanning direction and the straight line is read, as shown in FIG. 6 (b). When,
It was obtained by measuring the reading levels of a plurality of lines before and after the reading line, for example, as shown in FIG. 8 (a). Therefore, this line spread function h (x)
The function H (u) obtained by Fourier transform of is as shown in FIG.

In FIG. 8 (a), the sub-scanning position reading the straight line is set as the reference position ("0") of the distance x, the level at that position is set to "1", and the levels at other positions are standardized. ing. Further, the distance x is shown as a coordinate value when normalized to P = 1. Further, in FIG. 3B, the frequency corresponding to the sampling pitch P is set to 1 and the frequency is standardized.

Therefore, the MTF correction filter uses the reciprocal 1 / H of each function H (u) in the main scanning direction and the sub-scanning direction.
It may be configured so as to have the frequency characteristic of (u).

Now, letting i (x) be the function that corrects the MTF, the discrete value i (m) of the function i (x) that represents the elements of the MTF correction filter.
Δx) can be obtained by the following equation (I) using the function H (u) described above.

Here, M is the number of samples, M ′ = M / 2, Δx is the sampling pitch (= P), Δu ₀ = 1 / M, K is the normalization coefficient, and k, m (= −M ′, −M ′ + 1). , ..., 0,1, ..., M ',-
1),

In addition, the grayscale data at each sampling point is S (mΔ
x), the data 0 (mΔx) of the MTF correction number by i (mΔx) is expressed by the following equation (II).

Therefore, in order to perform the calculation based on the equation (II), it is necessary to process pixels adjacent to each other by the sampling number in each of the main scanning direction and the sub scanning direction. Here, there are 3 pixels in the main scanning direction and 4 pixels in the sub scanning direction.

However, since the circuit configuration becomes complicated to fully express such an operation, it cannot be adapted to real-time processing of an image signal. Further, the line spread function h (x) when reading a document image is not necessarily constant over the entire range of the image, and is not necessarily the same between devices.
Therefore, in such a document reading apparatus, it is considered that it is not necessary to increase the accuracy of the MTF correction calculation so much. Therefore, in the present embodiment, the MTF correction calculation is performed approximately for the adjacent 1 to 2 pixels to prevent the circuit configuration from becoming complicated and to realize a high-speed calculation so that it can be adapted to the real-time processing. ing.

Therefore, first, consider the MTF correction using one peripheral pixel. That is, the function i (mΔx) is set as in the following equation (III).

i (mΔx) = [b, a, b] (III) Therefore, the frequency characteristic I of this function i (mΔx)
(U) is expressed by the following equation (IV).

I (u) = | a + 2b (cos2πu) | (IV) That is, a and b satisfying the following expression (V) may be obtained.

I (u) = 1 / H (u) (V) However, if the data of 3 pixels in the main scanning direction and 4 pixels in the sub scanning direction are not used as described above, this formula (V) is completely satisfied. Cannot be done, so u = 0 and u =
Equation (V) is approximated by two points of α. Here, as α, 3/8 is set in order to obtain the characteristics shown in FIG. 5, but a value less than that can be set.

By the way, in the formula (IV), when a is changed to −2, −1,0,1,2 while fixing b = 1, for example, the frequency characteristic I
Since (u) changes as shown in FIG. 9, the bandpass characteristic shown in FIG. 5 cannot be realized.

In order to realize such a band pass characteristic, it suffices to combine the high pass characteristic and the low pass characteristic. That is, I _H (u) having the hyper characteristic is also multiplied by I _L (u) having the low-pass characteristic to form a band-pass characteristic of I _B (u).

Here, I _H (u), I _L (u), and I _B (u) are represented by the following equations (VI), (VII), and (VIII), respectively.

_{I H (u) = | a} + 2b (cos2πu) | ... (VI) I L (u) = | c + 2b (cos2πu) | ... (VII) I B (u) = I H (u) · I L (u) = ac + 2 (ad + bc) cos2πu + 4bdcos 2 2πu ... (VIII) however, the conditions of the high-pass filter and a low pass filter, a> 0, b <0 , c> 0, d> 0 _{Furthermore, I H (0) = I} L (0) = 1, a + 2b = 1, c + 2d = 1 (IX) Furthermore, in order to have high-pass characteristics in the region where I _B (u) is u = 3/8 or less, ad + bd in the second term of equation (VIII) ≦ 0 ... as (X) the peak value matches the _{u = 3/8, dI B} (u) du = -4π (ad + bc) sin2πu -16πbd (sin2πu · cos2πu) = -4πusin2πu {(ad + bc) + 4bd ( cos2πu)} ...
From (XI), (1 / 2π) cos ⁻¹ (− (ad + bd) / 4bd) = 3/8 (XII) Furthermore, the reciprocal of the function H (u) in FIGS. 7 and 8 1 / H ( u) and I _B (u), u = 0 and u = α (α ≦ 3/8)
The condition is given to the formula (VIII) so as to match with.

As a result, a, b, and
c, d can be obtained.

When a, b, c, d are determined in this way, two one-dimensional filters, that is, filters with high-pass characteristics [b, a, b]
And a filter [d, c, d] having a low-pass characteristic are multiplied to obtain a band-pass filter represented by the following equation (XIII).

[B, a, b] * [d, c, d] = [bd, ad + bc, ac + 2bd, ad + bc, bd] (XIII) That is, the expression (XIII) is the fifth
Function i _B (mΔx) (m =-
2, -1,0,1,2), which enables MTF correction.

As described above, the function i _B (mΔx) is calculated for each of the main scanning direction and the sub scanning direction, and if the MTF correction filters having these elements as coefficients are Fx and Fy, the filter F
x and Fy are expressed as follows.

Fx ＝ [C, B, A, B, C], Fy ＝ [F, E, D, E, F] Then, if these two filters Fx, Fy are combined,
It is possible to obtain the MTF correction filter FC as shown in FIG.

This MTF correction filter FC has bandpass characteristics as shown in FIG. 5 in the main scanning direction and the sub-scanning direction, and as a result, the MTF is originally very small, resulting in reduced reliability. The signal component of the Nyquist frequency is prevented from being emphasized, and at the same time, the influence of the periodic level fluctuation caused by using the separate amplifiers 22a and 22b for even-numbered and odd-numbered in the main scanning direction is removed. be able to. In addition, this MT without changing the low frequency component
While passing through the F correction filter, the signal component near the frequency component of the gain peak is emphasized, so that the blur of the image is corrected.

Further, another example of the MTF correction filter using each coefficient of the filters Fx and Fy is shown in FIG.

This MTF correction filter FK consists of a 5 × 5 matrix, and each matrix element D _ij (i, j = 1,2,3,4,5) is indicated by a broken line in the drawing in the main scanning direction and the sub-scanning direction. As shown, when each element is added, the addition result is matched with each coefficient of the filters Fx and Fy.

By the way, in the above-described embodiment, particularly in the main scanning direction, the influence of the periodic level fluctuation caused by the use of separate amplifiers 22a and 22b for even-numbered and odd-numbered is strong, It is necessary to give the band-pass characteristic to the MTF correction filter in the azimuth direction, but in the sub-scanning direction, since the effect of such level fluctuations is not significant, give the MTF correction filter in the sub-scanning direction the high-pass characteristic. In some cases it may be enough.

In such a case, each coefficient of the filter may be set so that the MTF correction filter characteristic in the sub-scanning direction becomes a high-pass characteristic. That is, for example, an MTF having a size of 5 in the main scanning direction and 3 in the sub scanning direction as shown in FIG.
The correction filter FD can be used.

An example of each coefficient of the MTF correction filter FD is shown in FIG.

Now, in the above-described embodiment, the scaling in the main scanning direction is performed by executing a digital operation in the main scanning scaling unit 25, but the scaling in the sub scanning direction changes the scanning speed of the optical system. I am doing it. For example, when the magnification is changed by 50%, the scanning speed of the optical system is doubled, and when the magnification is changed by 200%, the scanning speed of the optical system is halved. Even when the scanning speed in the sub-scanning direction is changed in this way, the scanning cycle in the main scanning direction is constant.

In this way, the scanning speed of the optical system is changed to 1 in the sub-scanning direction.
When the distance on the read document corresponding to the step is changed, the line spread function h (x) in the sub-scanning direction may change.

For example, the line spread function h when scaling at a magnification of 50%
(X) is as shown in Fig. 12 (a), and the magnification
The line spread function h (x) when the magnification is changed at 200% is as shown in FIG. 13 (a).

Further, the function H (u) obtained by Fourier transforming each line spread function h (x) becomes as shown in FIG. 12 (b) and FIG. 13 (b), respectively.

Therefore, if each coefficient of the MTF correction filter is set according to the image reading magnification, more effective MTF correction can be performed.

That is, the MTF correction filter FE when the magnification is 50% and the MTF correction filter FF when the magnification is 200% are as shown in FIGS. 10 (d) and 10 (e), respectively.

By the way, in the above-mentioned embodiment, in order to set each coefficient of the MTF correction filter, while setting each line spread function of the optical system in the main scanning direction and the sub-scanning direction,
The one-dimensional filters in the main scanning direction and the sub-scanning direction are formed and combined, but if the point spread function of the optical system is measured, the two-dimensional filter can be directly formed based on the measurement result.

In the embodiment described above, each coefficient of the MTF correction filter is set by using the data of 1 to 2 pixels because it is determined that the accuracy of the MTF correction calculation is not required so much. Then, it is necessary to set each coefficient of the MTF correction filter by using the data of pixels for each sampling number.

In the above-described embodiment, the magnification is 100% (1: 1), 50%.
% And 200%, MTF correction filters are formed, but MTFs that support other magnifications
A correction filter can also be formed. Again each
To apply the MTF correction filter, for example, if the magnification is 70% or less, use an MTF filter that supports 50% magnification and 70% or more 1
When it is 50% or less, an MTF correction filter corresponding to a 100% magnification can be used, and when it is more than that, an MTF correction filter corresponding to a 200% magnification can be used.

Further, in the above-described embodiment, the read original is fixed and the optical system is moved to scan the image in the sub-scanning direction. However, the original of the mechanism that fixes the optical system and conveys the read original in the sub-scanning direction The present invention can also be applied to a reading device.

Further, the line spread function and the MTF in the above-described embodiment
The coefficient of the correction filter is an example, and the coefficient is not limited thereto.

Further, in the above-described embodiment, the peak of the gain of the MTF correction filter is set at the position of 3/4 of the Nyquist frequency, but the setting condition of this peak is not limited to this.

The configuration of the control system in the present invention is not limited to the configuration of the above-mentioned embodiment.

[Effect] As described above, according to the present invention, a filter having a bandpass characteristic having a peak at a frequency smaller than the Nyquist frequency corresponding to a predetermined interval is provided, and image data is corrected by this filter. Since the signal component of the Nyquist frequency is not used as much as that of No. 1, the original image can be read with high quality and high resolution.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic configuration diagram showing an example of an optical system of a document reading apparatus according to an embodiment of the present invention, and FIG. 2 is a document reading apparatus according to an embodiment of the present invention. FIG. 5 is a block diagram showing an example of a control system, FIG. 3 is a block diagram showing a configuration example of an amplifier, FIG. 4 is a graph diagram showing a difference between even-numbered and odd-numbered characteristics in the main scanning direction, and FIG. Is an MT according to an embodiment of the present invention.
6A and 6B are graphs showing an example of the characteristics of the F correction filter, and FIGS. 6A and 6B are explanatory diagrams showing the outline of the method for measuring the line spread function in the main scanning direction and the sub scanning direction. Is a graph showing an example of the line spread function in the main scanning direction, FIG. 8B is a graph showing the Fourier function thereof, and FIG. 8A is an example of the line spread function in the sub scanning direction. FIG. 9B is a graph showing its Fourier function, FIG. 9 is a graph showing an example of frequency characteristics, and FIG. 10A is a schematic view showing an example of an MTF correction filter. FIG. 6B is a schematic diagram showing another example of the MTF correction filter, FIG. 6C is a schematic diagram showing a specific example of the MTF correction filter, and FIG. Fig. 11 (e) is a schematic diagram showing another specific example of the MTF correction filter, and Fig. 11 is an MTF correction filter. Furthermore schematic view showing another example of a filter, Figure 12 (a) is a graph chart showing another example of the line spread function in the sub scanning direction, FIG. (B)
Is a graph showing the Fourier function, FIG. 13 (a) is a graph showing another example of the line-spread function in the sub-scanning direction, and FIG. 13 (b) is a graph showing the Fourier function. FIG. 14 (a) is a schematic view showing a conventional example of an MTF correction filter, FIG. 14 (b) is a graph showing its characteristics, and FIG.
The figure is an explanatory view for explaining MTF deterioration. 9 ... Line image sensor, 22,22a, 22b ... Amplifier,
23 …… Shading correction / analog / digital conversion unit, 24 …… MTF correction unit, 26 …… Binarization processing unit, 27 …… Output circuit, FC, FD, FE, FF, FK …… MTF correction filter.

Claims (3)

[Claims] 1. An image reading apparatus for sampling an original image at a predetermined interval and converting it into image data, comprising filter means having a bandpass characteristic having a peak frequency lower than a Nyquist frequency corresponding to the predetermined interval. A document reading device, wherein the image data is corrected by a filter means. 2. The document reading device according to claim 1, wherein the filter means has different characteristics in the main scanning direction and the sub scanning direction. 3. The document reading apparatus according to claim 1, wherein the filter means is MTF correction filter means.