JPH09102030A - Picture reader - Google Patents

Abstract

PROBLEM TO BE SOLVED: To execute satisfactory correction around density which is usually used and to correct the distribution unevenness of whole light quantity. SOLUTION: At the time of an operation, a line sensor 14 is driven in a state where a reading light source 11 is turned off, and an offset value F(i) is obtained. Furthermore, a filter 15 is inserted and output distribution f1(i) is similarly obtained. The reading light source 11 is turned on and the output distribution L0(i) of maximum light is obtained. A statistical non-linear correction reference table is generated from the output distributions f1(i) and L0(i), and they are written into a reference table 20. A picture is taken in by using a real film 12 and the reference table executes non-linear correction so as to obtain a density value.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image reading device for optically reading a medical film image and digitizing it.

[0002]

2. Description of the Related Art In recent years, there has been active movement to digitize medical images such as X-ray fluoroscopic images and apply them to electronic filing and automatic diagnosis. As a device therefor, a film digitizer which optically reads an image-developed X-ray film and converts it into a digital value is known as an effective means.

Generally, medical X-ray films are large,
For example, the size is 14 inches × 17 inches, and in many cases, a fine image expression of about 2 to 10 lines / mm is a problem in diagnosis by a doctor. For example, even if the sampling pitch is set to 150 μm and a maximum of 3.3 lines / mm can be expressed, in the case of a film of 14 inches × 17 inches, the size of the entire image is considerably large of 2400 × 2900.

In the case of optically reading such a large size, a line sensor having a large number of pixels is generally used, and the film is physically moved to scan and read. Further, in a photoelectric conversion element of a line sensor typified by a CCD line sensor, a light amount and an output voltage generally have a linear relationship, and by utilizing this, it is possible to correct uneven distribution of the light amount, that is, shading.

FIG. 8 is a block diagram of a film digitizer of a conventional example, which faces a reading light source 1 such as a rod-shaped fluorescent lamp or a halogen lamp, a medium film 2, an imaging optical system 3,
A line sensor 4 such as a CCD line sensor is sequentially arranged. The film 2 is conveyed at a constant speed in the direction of arrow A by a mechanical means (not shown), and a visible image on the film 2 is read by transmitted light here.

Further, an A / D converter 5 for converting an analog value into a digital value is connected to the output of the line sensor 4,
The output of this A / D converter 5 is connected to the subtractor 7 together with the output of the line memory 6. The output of this subtractor 7 is
A logarithmic converter 8 for performing logarithmic conversion by a look-up table and a subtractor 9 are sequentially connected, and a line memory 10 is connected to an input of the subtractor 9.

During operation, first the reading light source 1 is turned on,
The output distribution L0 (i) is obtained by taking out one line of the highest light output a plurality of times without inserting the film 2 and averaging the same. However, if n is the number of pixels of the line sensor 4, then i = 1, ..., N.

Further, the reading light source 1 is turned off, and the output distribution L0
Similar to the case where (i) is obtained, the offset value F (i) of the line sensor 4 or the amplifier (not shown) is obtained by taking out and averaging one line a plurality of times, and stored in the line memory 6. Then, the transmittance T (i) of the film 2 is obtained from the output distribution L (i) obtained when the film 2 is transmitted through the normal film 2 by the following equation. T (i) ＝ {L (i) −F (i)} / {L0 (i) −F (i)}… (1)

This calculation also plays a role of correcting variations in individual characteristics of the photoelectric conversion elements on the line sensor 4, and if only the linearity of individual pixels is guaranteed, the offset value F (i) The variation of the slope is corrected by equation (1).

However, the value actually used is often not the transmittance T (i) shown by the equation (1) but the concentration value which is its logarithmic conversion. Since the calculation of division is complicated, the denominator and the numerator of equation (1) are logarithmically converted, and then the concentration value is obtained from the difference. If the density value is D (i),
Let log (10) be the base 10 logarithm, and D (i) = − log {T (i)} = log {L0 (i) −F (i)} −log {L (i) −F (i)} It can be shown as (2). However, if n is the number of pixels of the line sensor 4, then i = 1, ..., N.

When the film 2 is read, first, in synchronization with a signal indicating the start of one line generated from a control device (not shown) for controlling the line sensor 4, the subtractor 7 outputs the output of the line sensor 4 pixel by pixel. The distribution L (i) and the offset value F (i) stored in the line memory 6 are subtracted to calculate the numerator of the equation (1).

Further, the logarithmic converter 8 converts the logarithm of {L (i) -F (i)} as shown in the expression (2) based on the look-up table. At this time, since the reference table handles integer values, if the number of bits of the reference table is k1, logarithmic conversion is performed on the input x using the function z (x) represented by the following equation. However, 0 ≦ x ≦ 2 ^k1 −1. z (x) = log (x + 1) ・ 2 ^k1 / (k1 ・ log 2)… (3)

Further, log {L0 (i) -F (i)} of the equation (2) calculated by the output distribution L0 (i) and the offset value F (i) taken in without inserting the film 2 is given as It is stored by the line memory 10, and the line memory 1 is stored for each corresponding pixel.
The log {L0 (i) -F (i)} output from 0 is subtracted from the log {L (i) -F (i)} output from the logarithmic converter 8 by the subtractor 9, and the expression (2) is obtained. The density value D (i) shown is obtained.

Here, the density value D (i) is obtained by assuming that the relationship between the light input and output voltage of each pixel of the line sensor 4 is completely linear, but the actual light input and output voltage of each pixel of the line sensor 4 is obtained. The relationship is not completely linear, and a typical characteristic is the saturation characteristic.

FIG. 9 is a light quantity output characteristic diagram of each pixel of the line sensor 4, where the horizontal axis represents the light input and the vertical axis represents the output voltage. It is almost a straight line up to the light amount a, but linearity is lost when it exceeds it, and although it is possible to suppress the incident light amount to a light amount a or less and correspond to the formula (1) or (2), S / N From this viewpoint, it is preferable that the range of the amount of incident light is wide.

Considering that the transmittance T (i) is converted into the density value D (i) to be used, it is not necessary to have the same sensitivity in the range of the total light amount, and the resolution is rough in a portion where the light amount is large. Therefore, it may be converted into a digital value while including the saturation characteristic, and the digital value is linearized by a reference table,
It is theoretically possible to apply equation (2).

In this case, after the characteristic function f (x) representing the light quantity output characteristic shown in FIG. 9 is obtained, the inverse characteristic function g (x) is calculated, and the output voltage of the line sensor 4 is calculated as the inverse characteristic function. Linearize by calculating with g (x). Then, the density value D (i) can be obtained by the following equation. D (i) ＝ log [g {L0 (i) −F (i)}] − log [g {L (i) −F (i)}]… (4)

FIG. 10 is a graph showing the standard deviation of the average density when the density value D (i) of the film 2 whose density is almost uniform is calculated by the equation (4). The horizontal axis shows the average density of the film 2. The vertical axis shows the standard deviation. Since the standard is the case where the average density without film 2 is zero,
When the average density is zero, the correction is completely performed and the standard deviation is zero, and the variation increases as the average density increases.

[0019]

However, in the above-mentioned conventional example, in the line sensor 4 having a plurality of photoelectric conversion elements, the characteristics of the individual elements are not always the same but vary. In addition, there is no method to measure the characteristics of individual photoelectric conversion elements with high accuracy, and even if accurate measurement is possible, it is necessary to
Since it is necessary to convert the equation (4), the number of circuit elements increases, which is economically difficult.

Further, as the amount of light increases, the influence on peripheral pixels increases, and the independence of individual pixels decreases. That is,
The output distribution L0 (i) when there is no medium does not show the characteristics of individual pixels or small unevenness of the light quantity, and even if it is used, it is possible to correct small fluctuations for each pixel when the light quantity of the pixel is small. Can not. In other words, when there is no medium or when it is close to it, a fairly good correction is performed because conversion to the transmittance is performed based on the characteristics close to the shape, but if it gets away from that, correction of fine fluctuations for each pixel Is not done.

This means that a small fluctuation of the transmittance of 1/10 or less, that is, a density of 1 or more, which is a normally used light amount range, always remains, and the small fluctuation is a fixed pattern and appears as vertical stripes on the image. Then, good correction is performed only in a high transmittance range that is not normally used.

The object of the present invention is to solve the above problems,
An object of the present invention is to provide an image reading apparatus that corrects a good distribution of light amount near a normally used density and corrects the uneven distribution of the entire light amount.

[0023]

An image reading apparatus according to the present invention for achieving the above object illuminates an image recorded on a medium with illumination light to obtain a light amount distribution of transmitted light or reflected light. In an image reading device that reads and forms an image with a plurality of photoelectric conversion elements, after obtaining a light amount distribution obtained by weakening illumination light or its transmitted light or reflected light as compared with that during image reading, a value obtained by multiplying the light amount distribution by a constant Is provided as a reference distribution, and means for calculating the transmittance or reflectance of the image is provided.

[0024]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail with reference to the embodiments shown in FIGS. FIG. 1 is a block diagram of the first embodiment, which faces a reading light source 11 such as a rod-shaped fluorescent lamp or a halogen lamp, a medium film 12, an imaging optical system 13, a line sensor 14, and a line sensor. Filters 15 that can be inserted and removed and have a concentration of about 1 are sequentially arranged on the front surface of 14.

Further, an A / D converter 16 for converting an analog value into a digital value is connected to the output of the line sensor 14, and the output of this A / D converter 16 is sent to a subtractor 18 together with the output of the line memory 17. It is connected. This subtracter 1
The output of 8 is sequentially connected to a logarithmic converter 19 for performing logarithmic conversion by a reference table, a reference table 20 having a function of a statistical non-linear correction reference table, and a subtractor 21.

A line memory 22 for shading correction is connected to the input of the subtractor 21, and the line memory 17, the logarithmic converter 19, the reference table 20, and the line memory 22 are provided by a microprocessor, software, or the like. A control unit 23 configured to control the present system is connected.

Further, a memory 24 for temporarily storing the output data and an interface 25 for outputting the data to an external device are connected to the control unit 23, and the output of the subtracter 21 is also stored in the memory 24. Through interface 25
It is connected to the.

At the time of operation, first, the line memories 17 and 22.
Data are initialized to zero, and the reference tables of the logarithmic converter 19 and the reference table 20 are linearly set so that the input and output values are the same. Subsequently, the line sensor 14 is driven in a state where the reading light source 11 is turned off, the output is A / D converted by the A / D converter 16, and the subtractor 18, the logarithmic converter 19, the reference table 20, and the subtraction are performed. Data is stored in the memory 24 for n lines via the device 21.

At this time, since the data in the line memories 17 and 22 is zero and the reference tables of the logarithmic converter 19 and the reference table 20 are linear, the memory 24 has A
Data obtained by sign-reversing the output itself of the / D converter 16 by the subtractor 21 is stored.

Here, assuming that r (i, j) is the i-th pixel data of the j-th data on the memory 24, the offset value F (i) of the data when the reading light source 11 is turned off is as follows. Is calculated by the average calculation of. F (i) = − Σr (i, j) / n (1 ≦ i ≦ p) (5)

Where Σ is j = 1 to j = n for j
Shows the sum up to. This offset value F (i) is stored in the line memory 17 and used for the next calculation.

Next, the filter 15 is inserted in front of the line sensor 14 in a state where the reading light source 11 is turned on and the film 12 is not inserted. As in the case where the reading light source 11 is turned off and the offset value F (i) is calculated, the output distribution f1 (i) of the data transmitted through the filter 15 is taken in as data r (i, j) for n lines. The following average calculation is performed as in Eq. (5). f1 (i) = -Σr (i, j) / n (1≤i≤p) (6)

Where Σ is j = 1 to j = n for j
Shows the sum up to. And this output distribution f1 (i)
From the offset value F (i) shown in the equation (5) is subtracted by the subtractor 18 and stored in the memory 24. Assuming that the density of the filter 15 is 1.0 (transmittance 1/10), the output distribution f1 (i) obtained by passing through the filter 15 having a density of 1.0 has a small amount of light, and thus the individual pixel independence. Since each pixel is used in a region where the photoelectric conversion characteristic is substantially linear, it is considered that the unevenness of the light amount distribution of the reading light source 11 and the characteristic variation of each pixel are reproduced quite well.

Conversion to the actual density value D (i) is 10 · f1
It can be calculated by the following formula using (i) as a reference. D (i) = log {10 · f1 (i) −F (i)} −log {L (i) −F (i)} (7) where L (i) is the read value of the film 12. is there.

By using 10f1 (i) as a reference, it is possible to intensively correct the unevenness of the light amount distribution in the vicinity of the density 1 and the variation in the characteristics of individual pixels.

When it is actually executed, S (i) = log {f1 (i) ^.multidot.10 + 1} ^{.multidot.2k1} / (k1.multidot.log2) (8) The logarithmic converter 19 has Z (x) = log (x + 1) * ^2k1 / (k1 * log2) ... (9) is memorize | stored and a formula (7) is implement | achieved. Here, k1 is the number of bits in the calculation.

However, in the above operation, the reference table 20
Therefore, if there is a saturation characteristic in a portion where the light amount is large, there is a possibility that the unevenness of the light amount distribution may not be corrected correctly. A portion having a large amount of light, that is, a low density region usually has a low probability of being present in a medical image film, and it is not necessary to correct small variations, but it is desirable to correct at least a large amount of light unevenness.

Therefore, in a portion where the light amount is large (large value), it is necessary to store characteristic data for statistically correcting the saturation characteristic of the entire line sensor in the reference table 20 without considering the variation of individual sensors. , It is possible to correct the unevenness of the light quantity in a large light quantity area almost correctly.

As a means for obtaining the input / output characteristics of the sensor, normally, a reference medium should be read and calculated from the input / output relationship, but it is usually difficult to obtain an accurate reference medium. A method for statistically obtaining the input / output characteristics without using such a reference medium will be described as an example.

First, the line memory 22 is cleared to 0,
Remove the filter 15 from the front of the line sensor 14,
As in the previous case, the data r (i, g) is taken in for n lines.
Then, the output distribution L0 (i) corresponding to the maximum light is calculated by the following formula. Lo (i) = − Σr (i, j) / n (1 ≦ i ≦ p) (10)

Here, the output distribution f1 shown in the equations (6) and (10)
(i) and L0 (i), a method of creating a statistical nonlinear correction reference table will be described in detail. First, the maximum light output distribution L0 (i) is on the X-axis, and the output distribution f1 of the data transmitted through the filter 15 is f1. A curve that smoothly passes through the point with (i) as the Y axis by the least squares method is created.

At this time, for example, let m be a value in the range where the characteristic of the line sensor 14 is linear, and a constraint condition is given such that it is a straight line when X <m and Y = 0 when X = 0. When X> m, for example, assuming that the characteristic of the line sensor 14 is a fifth-order polynomial, 1 at X = m.
A curve is calculated such that the second and second differential values are equal.
As a method of obtaining it, for example, the simultaneous polynomial equation (11) below is solved.

[0043]

(Equation 1)

[0044] ^{Here, t1 (x) = x 5} -10m 3 x 2 -6m 5 t2 (x) = x 4 -6m 2 x 2 -3m -4 t3 (x) = x 3 -3mx 2 -m 3 Where Σ is the sum of i from 1 to p, where p is the number of pixels in one line. In the simultaneous equations (11), a,
b, c and e are obtained, and d and f are calculated by the following equation. d = −10 am ³ −6 bm ² −3 cm f = −6 am ⁵ −3 bm ⁴ −cm ³

From these d and f, a function q (x) which is the following fifth-order polynomial is defined. q (x) = ax ⁵ + bx ⁴ + cx ³ + dx ² + ex + f (12)

Using this function q (x), the function v (x) shown by the following equation
Is calculated, a characteristic curve approximating the characteristic of the line sensor 14 can be obtained. v (x) = xm / q (m) 0 ≦ x ≦ m v (x) = mq (x) / q (m) x> m (13)

According to the above calculation method, dg (x) / dx = q (n) / m, D ² q (x) /
This can be easily derived by introducing the constraint condition of dx ² = 0. Then, the inverse function of this function v (x) is obtained and written in the reference table 20.

FIG. 3 shows an example in which the calculation of the equation (13) is performed on the light amount distribution shown in FIG. 2 to calculate a reference table for statistically non-linear correction. Input data is shown on the horizontal axis.
Output data is shown on the vertical axis. It can be seen that the output data is linear up to a certain input data value, and is nonlinearly corrected in the range exceeding the input data.

Next, the shading data S (i) written in the line memory 22 is obtained from the output distribution f1 (i) by the following equation. S (i) ＝ log {f1 (i) ・ m / q (m) +1} ・ 2 ^k1 / (k1 ・ log 2)… (14)

Here, k1 is the number of input bits, and a reference table for logarithmic conversion shown by the logarithmic converter 19 is created for the input x according to the following equation. However, 0 ≦ x ≦ 2 ^k1
It is -1. z (x) = log (x + 1) ・ 2 ^k1 / (k1 ・ log 2)… (15)

The advantage of this method is that the density is statistically estimated by m / q (m) even if the density of the filter 15 is unknown.

When the above preparation is completed, an image is captured using the actual film 12. The above-mentioned preparation operation is shown in FIG.
Is shown in the flow chart of FIG. If a reference medium having a plurality of correct transmittances can be obtained, the characteristics may be corrected.

FIG. 5 is an example in which the density value is calculated from the output distribution shown in FIG. 2. Since the film 12 which is almost completely uniform is used as the medium, the result must be uniform.

In FIG. 6, the abscissa of each film 12 shows the average value of the densities of the film 12, and the ordinate shows the standard deviation thereof. In this method, filter 1 with a density of about 1.0
Since the standard is 5, the film 12 with a density of about 1.0
Has been completely corrected, the standard deviation is zero,
The overall variation is quite small.

According to this method, good correction is performed in the vicinity of the density of about 1.0, which is often used in an image having light and shade, and vertical stripes are almost eliminated. On the other hand, vertical stripes appear due to pixel variations in areas with extremely low transmittance, but there are few such low-density images in normal medical X-ray film images. Since this is rarely done, there is no practical problem. Most films have a base density of about 0.1 to 0.3, and it is rare to measure a density below that.

In this embodiment, when the statistical non-linear correction reference table is created, the filter 15 is inserted in front of the line sensor 14 to weaken the light quantity.
If the light quantity of 1 is controllable and the light quantity distribution does not change with respect to the change of the light quantity, the light quantity of the reading light source 11 may be reduced without using the filter 15 to create the non-linear correction reference table. It is possible.

Further, in the present embodiment, the image reading apparatus for forming an image from the light distribution transmitted through the medium has been described. However, if the medium is a reflection type image reading apparatus, the image is read by the same procedure for the reflected light distribution. Can read.

Further, the image reading device for reading the image by irradiating the image with the reference light has been described, but in the device for imaging a device which emits a luminous flux having an image distribution by itself such as fluorescence without the reference light by a photoelectric conversion element. Also, the present invention can be applied to the present embodiment by using the distribution obtained with a relatively weak light amount as a reference. Instead of using the line sensor, other surface sensors, for example, can be easily applied.

FIG. 7 shows a block diagram of the second embodiment.
Although the configuration is almost the same as that of the first embodiment, the shield 26 can be inserted into the front surface of the line sensor 14. In the first embodiment, the natural shading used in the system is used at the stage of creating the statistical reference table, but when the variation range is narrow, data covering the characteristics of the line sensor 14 cannot be obtained. There are cases. Therefore, only when a statistical reference table is created, shading is actively created and data is loaded.

In operation, instead of inserting the filter 15 to obtain the output distribution f1 (i) in the first embodiment, the filter 15 is inserted and the shield 26 is inserted in front of the line sensor 14 at the same time. Since the focus of the imaging optical system 13 is not fixed on the shield 26, the light flux is diffused and a pseudo shading characteristic having a wide light amount range is created. Here, the output distribution f1 ′ (i) when the light amount is small is calculated by the following equation. f1 '(i) =-Σr (i, j) / n (16)

From this state, only the filter 15 is removed, and the output distribution L0 '(i) when the amount of light is large is obtained by the following equation. L0 '(i) =-Σr (i, j) / n (17)

However, Σ is j = 1 to j = n for j.
Shows the sum up to. The output distributions f1 ′ (i) and L0 ′ (i) thus obtained include many light amount ranges.

Here, the output distributions f1 (i) and L0 (i) of the equation (12) are
Are replaced with output distributions f1 '(i) and L0' (i), respectively, to solve the simultaneous equations and create a reference table for statistical nonlinear correction. Also, the offset value F (i), shading data
S (i) may be obtained in the same manner as in the first embodiment in which the shield 26 is not used.

[0064]

As described above, in the image reading apparatus according to the present invention, the density near the density which is usually used, which causes a problem of vertical stripes, is corrected by, for example, obtaining and correcting the light quantity distribution through a uniform filter. The fixed pattern can be removed almost completely by a low-cost device. Furthermore,
If the statistical non-linear characteristic of the photoelectric characteristic is obtained by using the unevenness of the light amount distribution, the shading correction can be performed over the entire density range.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a first embodiment.

FIG. 2 is a graph showing the output of the line sensor.

FIG. 3 is a graph of an example of a statistical non-linear correction reference table.

FIG. 4 is a flowchart of a preparation operation.

FIG. 5 is a graph of density values calculated using a non-linear correction reference table.

FIG. 6 is a graph showing standard deviation with respect to a density value.

FIG. 7 is a configuration diagram of a second embodiment.

FIG. 8 is a configuration diagram of a conventional example.

FIG. 9 is a light amount output characteristic diagram.

FIG. 10 is a graph showing standard deviation with respect to density value.

[Explanation of symbols]

 11 light source for reading 12 film 14 line sensor 15 filter 16 A / D converter 17, 22 line memory 18, 21 subtractor 19 logarithmic converter 20 reference table 23 control unit 25 interface 26 shield

Claims (6)

[Claims] 1. An image reading apparatus for irradiating an image recorded on a medium with illumination light to read a light amount distribution of transmitted light or reflected light by a plurality of photoelectric conversion elements to form an image, the illumination light or the transmitted light thereof. After determining the light amount distribution obtained by weakening the light or reflected light more than when reading the image, a means for calculating the transmittance or reflectance of the image is used with a value obtained by multiplying the light amount distribution by a constant as a reference distribution. Image reading device. 2. The image reading apparatus according to claim 1, further comprising means for creating a reference table for correcting non-linearity of the photoelectric conversion element. 3. The image reading apparatus according to claim 1, wherein a filter is inserted in front of the photoelectric conversion element as a means for weakening the reference light. 4. The image reading apparatus according to claim 1, wherein a light source capable of photoelectric control is used as a means for weakening the reference light. 5. The means for creating the reference table uses an inverse function of what is statistically derived from the correlation between the outputs of the individual photoelectric conversion elements when the reference light is weakened and strengthened. The image reading device according to claim 2. 6. The image reading apparatus according to claim 2, wherein the means for creating the reference table derives it from output characteristics of the photoelectric conversion element of a standard medium having a plurality of transmittances or reflectances in advance.