JPH0127467B2 - - Google Patents

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention relates to a digital image distortion correction method.
Specifically, the present invention relates to a method of correcting shading distortion of an input image by using image data taken with a sensor with little uneven shading or by storing coefficients of paralinear approximation in advance.

[Background of the invention]

In order to analyze and process an image using a digital computer, it is first necessary to convert an analog signal indicating the intensity of each point of the image into a digital signal and input it to the computer.

FIG. 1 is a schematic diagram of a conventional image input device.

In Figure 1, 1 is the subject, 2 is the sensor, and 3
4 is an analog-to-digital converter, and 4 is a storage device.

The radiation intensity of the object 1 is formed by the sensor 2 as a two-dimensional intensity distribution as shown in the image 8, and an analog signal indicating the intensity at each point is sampled and quantized by the analog-to-digital converter 3 into a digital signal. and is read into the storage device 4.

In the sensor 2 shown in FIG. 1, as shown in FIG.
A light ray 11 from the subject passes through an optical system 12 such as a lens.
As a result, an image is formed as an electric charge on a photosensitive surface 13 of a vidicon or the like in the focal plane. A reading device 15 reads out this image using an electron beam 14, converts it into voltage, and sends it out via a signal line 16.

Incidentally, the image captured by the input device shown in FIG. 1 includes unevenness in density due to unevenness in sensitivity at each point on the photosensitive surface 13 of the sensor 2. In other words, even if a light source with a uniform brightness distribution is used for the subject 1, the gradation intensity distribution of the sensor output image 8 or the gradation intensity distribution of the image 9 read into the storage device 4 will not be uniform in area. This results in unevenness.

Therefore, in order to use image data for precise analysis and processing, it is necessary to remove this unevenness in density.

Conventionally, in order to remove uneven shading from an image, as shown in FIG. 1, a correction device consisting of a conversion device 6, a storage device 7, and an input device 5 that provides a conversion table is connected downstream of the input device.

The conversion device 6 converts each image data stored in the storage device 4, that is, the intensity of each sampled image unit, to
Convert to intensity with unevenness removed according to the conversion table. The conversion table is for making the gray scale intensity distribution of the image uniform for the subject 1 of a light source with uniform brightness, and the pixel data converted according to this conversion table is stored in the storage device 7. can be stored in

This conversion table is created in the order shown in FIG. 3 a, b, and c.

In Figure 3a, the maximum value of the input radiation intensity that can be captured by sensor 2 is x _nax (unit: cd, for example).
Assuming that a known uniform light source is the subject and the output voltage appearing on the sensor output signal line 16 for the input radiation intensity x is y (unit: v, for example), then at the position on the sensor surface corresponding to each pixel of the development, (x ,
The relationship y) is obtained. That is, each point 17 in FIG. 3a is measurement data corresponding to each pixel of one light source when the radiation intensity x is varied from 0 to x _nax and measured using several light sources.

FIG. 3b shows a model of the characteristics of an analog-to-digital converter. If the number of levels of digital data after conversion is L when the sensor output voltage range is changed from 0 to y _nax , the output Z of the analog-to-digital converter for the sensor output voltage y is given by the following equation. It can be done.

Z=[L×y/y _nax ]...(1) Here, [a] is the maximum integer that does not exceed a non-negative real number a.

In the correction device, a variable table given from the input device 5 to remove uneven shading is obtained as follows.

For each point 17 in FIG. 3a, the correction intensity u obtained by removing uneven shading is expressed by the following equation.

u=[L×x/x _nax ]...(2) FIG. Curve 20 connecting 19
This shows that.

The function 20 obtained by linearly interpolating each point 19 in FIG. 3c is given by the following equation.

u=f(Z)...(3) Also, the conversion table shows the level 0, 1...L of variable Z.
−1, it is expressed by the following equation.

u=[f(Z)]...(4) In a normal vidicon sensor, the value u of the above equation (3)
can be approximated by a linear function with sufficient accuracy. In this way, for each pixel of the image,
By performing the above operations, a desired conversion table can be obtained.

The conventional method of correcting the uneven shading of an input image using a conversion table is such that the uneven shading characteristics of the sensor 2 can be completely measured, and the function f in u=[f(Z)] in (4) above can be completely measured. This method is based on the premise that it can be modeled.

However, in the image data that is actually obtained, there are cases where the function f that expresses the uneven density characteristics cannot be measured.
Something unknown, or even once the function f is measured,
In many cases, due to deterioration of the sensor 2, the characteristics of the unevenness in density of an image actually obtained do not match the characteristics of the unevenness in density measured in advance.

For such images, conventional shading unevenness correction methods have the disadvantage that it is difficult to remove shading unevenness.

Further, in the conventional method for correcting uneven shading, it is necessary to prepare a conversion table for each pixel on the screen, and there is also the drawback that the memory capacity becomes large to store the conversion table.

[Purpose of the invention]

In order to eliminate these drawbacks, the present invention aims to correct the shading distortion of image data with uneven shading for which the correction function f is not clear, and to reduce the capacity of the conversion table to be stored. An object of the present invention is to provide a digital image distortion correction method capable of correcting uneven shading at high speed and with high accuracy.

[Summary of the invention]

In order to achieve the above object, the image distortion correction method of the present invention includes means for inputting an image including uneven shading, and after or simultaneously converting the pixel signals of the input image from analog to digital.
Divide the screen into an N×N grid, compare small areas of the image to be compared around the segmented grid points, and
Using the least squares method from the distribution of two histograms,
The feature is that it is equipped with a means for correcting the unevenness of shading by finding a transformation model on a grid point using the model of y = αx + β and interpolating the surrounding four transformation models (here, N is Any positive integer, y is the pixel intensity of an image without uneven shading, x is the pixel intensity of an image with uneven shading, α and β are functions of the location of the image (coefficients of the paralinear function). In addition, when calculating a transformation model on divided grid points as a means for correcting the unevenness of shading, instead of calculating from the histogram distribution by the least squares method, data of an image with less unevenness of shading is used to correct the unevenness of shading. With respect to a reference point P set for correction on an image, point Q is found on an image with no uneven shading, and by matching the histograms of the images around point P and point Q, the shading of the image can be adjusted. Another feature is that it removes unevenness. Furthermore, when obtaining coefficients for paralinear approximation at each segmented lattice point as a means of correcting the uneven shading, the coefficients are calculated in advance and stored in the memory within the analog-to-digital converter. There are also characteristics.

[Embodiments of the invention]

First, we will explain the device (first device) that uses image data with less unevenness of shading, and then next, when such image data is not available, by inputting the coefficients of the paralinear approximation, we can effectively change the shading. A device (second device) for correcting distortion will be explained.

In the first device, a reference point P (l, p) is set on an image with uneven shading to correct uneven shading, Q is calculated for P on an image without uneven shading, and P and Q are A method is used to remove unevenness in image density by matching the histograms of surrounding images.

Now, let x be the pixel intensity of an image with uneven shading, and y be the pixel intensity of an image without uneven shading. Generally, the uneven shading correction function f can be approximated with sufficient accuracy by a linear function, so the following equation holds. do.

y=ax+b (5) Here, a and b can be expressed as functions a(l, p) and b(l, p) of the location (l, p) of the image. Therefore, if we take a histogram around P, and set the mean to μ _x and variance to σ _x ² , and then take a histogram around Q, and let the average to μ _y and the variance to σ _y ² , then on point P, a(l, p) and b(l, p) are given by the following equations.

b(l, p)=μ _y −a(l, p) μ _x (7) The details until the above equations (6) and (7) are derived will be described. When dispersion is calculated around the average intensities μ _x and μ _y for a pixel x in an image with uneven shading and a pixel y in an image without uneven shading, the frequency curves shown in FIGS. 4a and 4b are obtained. When calculating the average value E{y} of the frequency in Figure 4b, from the above equation (5), y=ax+b...(5) E{y}=E{ax+b} E{(y-μ _y ) ² =E {{(ax+b)−(aμ _x +b)
} ² } =E{a ² (x−μ _x ) ² } =a ² E{(x−μ _x ) ² } …(8) Here, σ _y ² = (y−μ _y ) ² σ _x ^{2 = (} _x ⁻ _{μ _} ^_ be guided.

Furthermore, from FIG. 4 a and b, the following relationship is established between the average intensities μ _y and μ _x based on the above equation (5).

μ _y =aμ _x +b (11) Therefore, the above formula (7) is derived from the above formula (11).

The following processing given by equations (6) and (7) above is performed on all reference points P(l _i , p _j ) set in the image with grayscale distortion, and a(l _i , p _j ), b(l _i , p _j ) and use these values to calculate the functions a(l, p), b(l, p) at any point (l, p) on the screen.
is calculated by spatial interpolation.

Using these a(l, p) and b(l, p), the pixel intensity x^(l, p) with uneven shading is calculated from the pixel intensity x(l, p) of the image with uneven shading. It is determined by the following calculation.

x^(l,p)=a(l,p)・x(l,p)+b(l
,
p)...(12) FIG. 5 is a diagram explaining the principle of the present invention, and FIG. 6 is a block diagram of a digital image distortion correction apparatus showing a first embodiment of the present invention.

The image 23 with uneven shading and the image 21 without uneven shading in FIG. 6 are a and b in FIG. 5, respectively.
corresponds to FIG. 5a shows an image with uneven shading divided into n×m blocks, and the shading distortion is corrected by using the surrounding histogram for each block and making it correspond to the image in FIG. 5b.

An image 21 without uneven shading and an image 23 with uneven shading are input from digital image input devices 22 and 24 in response to a control signal from a processing device 25 .

At this time, as shown in FIG.
The peripheral histogram of (l _i , p _j ) and the peripheral histogram of Q on the image with no unevenness of shading corresponding to P are expressed as
Calculation is performed using the image input devices 22 and 24 when inputting the image.

The image with uneven shading is transferred to the image storage device 26 at the same time as the histogram is calculated.

After inputting the image data, the processing device 25 retrieves the histograms from the image input devices 22 and 24, and calculates the average μ _x of the histograms at the reference point P(l _i , p _j ).
(i, j) and the histogram variance σ _x ² (i, j)
and the corresponding mean μ _y (i, j) and variance σ _y of Q
² (i, j), and further a(i, j), b(i,
Calculate j).

Next, regarding the histogram around the reference point,
This will be explained with reference to FIGS. 7 and 8.

As shown in Fig. 7, when we take a reference point P in an image with uneven pixel intensity x, it becomes a function of slow changes in some places, and a function of rapid changes in other places. Therefore, in order to approximate, the region of P is set wide for slow changes and narrow for rapid changes. A function y (y ₁ to y ₄ ) at the grid points of the area is calculated, and other points are subjected to spatial interpolation.

y ₁ = a ₁ x + b ₁ , y ₂ = a ₂ x + b ₂ y ₃ = a ₃ x + b ₃ , y ₄ = a ₄ x + b ₄ ...(13) a (i, j), b (i, j) are lattice Coefficients on points
Calculate as a paralinear function using a ₁ , a ₂ , a ₃ , a ₄ , b ₁ , b ₂ , b ₃ , and b ₄ .

If each of a and b is described three-dimensionally using the gradation level of the image, the shapes shown in FIG. 8 a and b will be obtained.

a(l, p)=α ₀ +α ₁ L+α ₂ P+α ₃ L・P b(l, p)=β ₀ +β ₁ L+β ₂ P+β ₃ L・P …(14) Here, (L, P) is This is the relative position within the block.

In the above equation (14), by calculating as a paralinear function of l and p and a polynomial function of l and p, the coefficient a
(l, p), b(l, p), transfer these to the unevenness correction device 27, and image storage device 26.
A corrected image 28 can be obtained by transferring the image data of 2 to the unevenness correction device 27.

FIG. 9 is a block diagram of the shading unevenness correction device when the functions a(l,p) and b(l,p) are paralinear functions.

The signal 40 sent from the processing device 25 is a
The determining circuit 29 determines which signal is the coefficient of (l, p), b(l, p) or the start signal of uneven density correction.

If the data are coefficients of a(l,p) and b(l,p), the data is sequentially transferred to the bilinear model table 33, and the contents of the table 33 are set so that unevenness in shading can be corrected.

If it is a start signal for density unevenness correction, counter 3
A signal is sent to 0. When the counter 30 receives the signal, it clears the contents of the counter to 0 and sequentially counts up the contents of the counter, but while the contents of the counter are incremented by one count, the unevenness correction device performs the following operations. .

First, the content K of the counter is sent to the address calculation circuit 32 of the bilinear model table 33, so the address calculation circuit 32 calculates the pixel position (l, p) on the image and the corresponding block ( i, j).

K=(l-1)×P _L +p (15) Here, P _L is the number of pixels in one line.

Furthermore, from (l, p) calculated by equation (15), it is calculated which block (i, j) the point is included in.

Subsequently, the address calculation circuit 32 calculates the address A _D of the corresponding table and the relative line L within the block using the following equation.

A _D =(i-1)×m+j (16) where m is the number of blocks in the horizontal direction.

L=l-(i-1)×L _B -1 (17) Here, L _B is the number of lines in the block.

As a result of calculation, table address A _D is address register 3 of bilinear model table 33.
1, and the relative line L is calculated by the calculators 34 and 3.
Sent to 5.

Coefficients α ₁ , α 2 , α ₃ , _{α 4 ,} β ₁ , β ₂ , β ₃ of the paralinear model according to the contents of the address register 31
,
β ₄ is read from the bilinear model table 33 and transferred to the computing units 34 and 35.

FIG. 10 shows the computing units 34 and 35 in FIG.
FIG.

The calculation circuit 44 calculates α ₁ +α ₂ L or β ₁ +β ₂ L, and the calculation circuit 45 calculates α ₃ +α ₄ L or β ₃ +β ₄ L. The values calculated by the arithmetic circuits 44 and 45 are α ₁ + α ₂ L, α ₃ +
α ₄ L, β ₁ +β ₂ L and β ₃ +β ₄ L regarding the coefficient b, and each is sent to the arithmetic units 36 and 37.

The detector 43 detects the address A _D of the parallel line model table 33 which is the output of the address calculation circuit 32.
Then, it is detected whether or not the relative line number L is the same as the previous one. If either one is different from the previous one, a signal "1" is sent to the calculators 36 and 37, and if there is no difference, a signal "1" is sent to the calculators 36 and 37. Send 0”. The purpose of the detector 43 is to detect the change of blocks within one line.

Arithmetic unit 36 calculates a(l, p), and arithmetic unit 3
7 calculates a(l,p).

FIG. 11 shows the computing units 36 and 37 in FIG.
FIG.

If the signal 46 from the detector 43 is "1",
The selection circuit 49 selects α ₁ +α ₂ L (or β ₁ +β ₂ L) which is the output of the arithmetic unit 34 (or 35),
The selection circuit 50 selects 0.

At this time, the adder 52 outputs a coefficient corresponding to the case where the value of the relative pixel P in the block is P=0. Further, the selection circuit 51 inputs α ₃ +α ₄ L (or β ₃ +
β ₄ L) is stored.

When the signal 46 from the detector 43 is "0", the selection circuits 49 and 50 select the contents of the register 54 storing the previous output contents of the adder 52 and the contents of the register 53 representing the gain. and input these to the adder 52. Output of adder 52
OUT is the coefficient a (or b) to be sought.

Since the relative pixel P is an integer, by repeating the addition by the adder 52 that number of times, the value a(l,p)=(α ₀ +α ₁ L)+P(α ₂ +α ₃ L) is obtained.

In FIG. 9, the outputs of arithmetic units 36 and 37 are transferred to a multiplier 38 and an adder 39, and the image storage device 2
Using the image intensity x of the uneven image read from 6, the multiplier 38 calculates ax, and the adder 39
By performing the calculation of ax+b, the following equation is obtained.

x^=ax+b...(12)' As mentioned above, x^ is the corrected pixel intensity. In this way, in the first device, by simply superimposing the intensity of an image with uneven shading on the intensity of a reference image without uneven shading,
You can correct uneven shading.

Next, the second device will be explained.

Unlike the first device, when an image without uneven shading cannot be obtained, the second device stores the coefficients based on the parallel line model in advance, so that it can be used without increasing the memory capacity. This corrects uneven shading with high precision.

FIG. 12 is a block diagram of a digital image distortion correction device showing a second embodiment of the present invention.

Light from the subject 1 is converted into voltage by the sensor 2 and stored in the storage device 4 via the analog-to-digital converter 63 with correction. Analog with correction
The digital converter 63 has a conversion characteristic 57 as shown in FIG. 13, inputs the coefficients of bilinear approximation from the input device 64, and performs gradation unevenness correction simultaneously with analog-to-digital conversion. The conversion characteristic 57 in FIG. 13 is obtained using equations (1), (2), (3), and (4) as follows: u=[f(Z) =[f([L×y/y _nax ]) ] d=[g(y)] ...(18) Given as follows.

Here, as described above, in a normal vidicon sensor, based on the fact that f(Z) can be linearly approximated, g(y) can also be linearly approximated.

Now, suppose that the linear approximation of g(y) in the above equation (18) is expressed by the following equation.

g(y)=ay+b...(19) After converting the shading intensity according to the above equation (18) for each pixel of the input image 55 having uneven shading, the shading of the image read into the storage device 4 is 5.
It becomes uniform as shown in 6.

Next, a method of sampling the sensor output signal will be explained.

As shown in FIG. 14, the image data appearing as charges on the sensor photosensitive surface is read out sequentially one line at a time from the upper left of the screen by an electron beam, for a total of N lines. Read sensor output signal 59
is a voltage that fluctuates with time, as shown in FIG. This is sampled at a period of △t ₁ , and 1
Obtain M pixel data per line. Here, let Δt ₂ be the time during which there is no image data between lines.

FIG. 16 is a block diagram of the analog-to-digital converter with correction in FIG. 12.

The conversion characteristic generator 60 generates the coefficients a and b of equation (19) for each pixel on the screen, and the sampler 65 generates the timing signal from the timing generation circuit 67 (a timing signal for every Δt ₁ ,
However, there is no signal at △t ₂ between the lines), so
The output signal of sensor 2 is sampled every Δt ₁ . The address generation circuit 68 is a timing generation circuit 67
An address (line number l and pixel number P on the line) corresponding to the pixel being sampled is sent to the conversion characteristic generating device 60 by a timing signal from. The conversion characteristic generating device 60 calculates the coefficients a and b of equation (19) corresponding to the above address and outputs them to the conversion circuit 66.
send to

FIG. 17 is an internal configuration diagram of the conversion circuit 66 in FIG. 16.

Among the coefficients sent from the conversion characteristic generator 60, b in equation (19) is converted into an address signal by the digital-to-analog converter 661 and sent to the operational amplifier 663.
In addition, a in equation (19) is converted into an analog signal by the digital-to-analog converter 662 and sent to the operational amplifier 6.
64, respectively. After that, sampler 6
The image data sampled at 5 is transferred to an operational amplifier 66.
After being multiplied by a by 4, the operational amplifier 663 adds by b. Thereafter, the signal passes through an analog-to-digital converter 665 having the same characteristics as in FIG. In this way, by driving the timing signal, grayscale distortion correction, sampling, and quantization are sequentially performed for each pixel at high speed.

18 and 19 are diagrams showing the conversion characteristic creation principle of the conversion characteristic generating device 60 in FIG. 16.

As shown in Fig. 18, there are n screens each horizontally and vertically,
Divide into m blocks. It is assumed that the coefficients a and b of equation (19) have been determined at each grid point of divided block B by the method described above.

At this time, a and b of the pixel 70 (coordinates are (l, p)) within one block B are determined as follows.

As shown in Figure 19, the size of one block is
Let L _B lines x P _B pixels. At this time, pixel 70
The corresponding block number (i, j) is determined by the following equation.

Also, the relative coordinates (u, v) within the block are determined by the following equation.

u=l-(i-1)×L _B v=p-(j-1)×P _B (21) At this time, a and b in (u, v) are expressed by the following equation.

a (u, v) = α ₀ + α ₁ u + α ₂ v + α ₃ uv … (22) b (u, v) = β ₀ + β ₁ u + β ₂ v + β ₃ uv … (23) This bilinear approximation , n is set to a sufficiently large number (10 or more), it can be applied to a normal vidicon camera.

Here, the coefficients α _i and β _i (i=0,
1, 2, and 3) are obtained from a and b measured at the grid points of block division as follows.

That is, a _i , b _i (i=
0, 1, 2, 3) by the following formula.

FIG. 20 is an internal configuration diagram of the conversion characteristic generating device 60 in FIG. 16.

From the input device 64, block B (1, 1), . . .
For each of B(n, m), the above (24) and (25)
Coefficients α _i , β _i determined in advance according to the formula
(i=0, 1, 2, 3) is input to the memory 602 and stored. a and b for each pixel position are
It can be found as follows.

From the pixel coordinates (l, p) sent from the address generation circuit 68, the address conversion circuit 601
According to the above formula (20), the corresponding block number (i, j)
In addition, the relative coordinates (u, v) within the block are calculated using the above equation (21). Block number (i, j)
The coefficient α _i (i=0, 1, 2,
3) and the coefficient β _i (i=0, 1, 2, 3) of the above equation (23)
are read from memory 602 and sent to arithmetic circuits 603 and 604, respectively.

Arithmetic circuits 603 and 604 use the relative coordinates (u, v) output from address conversion circuit 601 and the coefficients α _i and β _i to calculate coefficient a and coefficient b according to equations (22) and (23), respectively. Calculate.

This calculated result is sent to the conversion circuit 66.

In this way, according to the second device, by storing the coefficients α _i and β _i of paralinear approximation determined in advance, the grayscale distortion can be corrected for each pixel of an image taken with a vidicon camera, etc. and digitalization,
It is done at high speed at the same time.

〔Effect of the invention〕

As explained above, according to the present invention, shading distortion of image data including shading unevenness can be reduced by using an image with little shading unevenness or by calculating and storing coefficients for parallel line approximation in advance. It can be corrected with high accuracy, and since conversion tables etc. are not stored, there is no increase in memory capacity, and high-speed image analysis and processing is possible by converting to digital signals and inputting them into a digital computer at high speed. becomes.

[Brief explanation of drawings]

Fig. 1 is a schematic diagram of a conventional image input device, Fig. 2 is a principle diagram of the sensor shown in Fig. 1, Fig. 3 is an explanatory diagram of the creation order of the conversion table in Fig. 1, and Fig. 4 is used in the present invention. A dispersion curve diagram of pixel intensity of an image, FIG. 5 is a diagram explaining the principle of the present invention, FIG. 6 is a block diagram of a digital image distortion correction device showing the first embodiment of the present invention, and FIG. 7 is a reference point. An explanatory diagram of the surrounding histogram, FIG. 8 is an explanatory diagram of the bilinear model of coefficients a and b, FIG. 9 is a block diagram of the unevenness correction device in FIG. 6, and FIG. 10 is the arithmetic unit 3 in FIG.
4 and 35, FIG. 11 is a block diagram of arithmetic units 36 and 37 in FIG. 9, FIG. 12 is a block diagram of a digital image distortion correction device showing a second embodiment of the present invention, and FIG. Figure 12 shows the conversion characteristics, Figures 14 and 15 are explanatory diagrams of the sampling method of the sensor output signal, and Figure 16 shows the conversion characteristics of the
17 is an internal configuration diagram of the conversion circuit in FIG. 16, and FIGS. 18 and 19 illustrate the conversion characteristic creation principle of the conversion characteristic generator in FIG. 16. The figure shown in FIG. 20 is an internal configuration diagram of the conversion characteristic generating device in FIG. 16. 21: Image without uneven shading, 23: Image with uneven shading, 22, 24: Image input device, 25: Processing device, 26: Image storage device, 27: Non-uniform shading correction device, 28: Corrected image, 29: Judgment circuit, 3
0: Counter, 31: Address register, 3
2: Address calculation circuit, 33: Paralinear model table, 34 to 37: Arithmetic unit, 38: Multiplier, 3
9: Adder, 44, 45: Arithmetic circuit, 49, 5
0, 51: selection circuit, 52: adder, 53, 5
4: Register, 63: Analog-digital converter with correction, 64: Input device, 57: Conversion characteristics, 5
8: Scanning, 59: Sensor output signal, 60: Conversion characteristic generator, 65: Sampler, 66: Conversion circuit,
67: Timing generation circuit, 68: Address generation circuit, 661, 662: Digital-analog converter, 663, 664: Operational amplifier, 665: Analog-digital converter, 70: Pixel, 71-
74: Lattice point, 601: Address conversion circuit, 60
2: Memory, 603, 604: Arithmetic circuit.

Claims (1)

[Claims] 1. After inputting an image containing uneven shading and converting the pixel signals of the input image into analog/digital, or at the same time as converting, the screen is divided into an N×N grid, and the divided Compare the image around the grid point with the corresponding small area of the reference image, and calculate y using the least squares method from the distribution of the two histograms.
A digital image distortion correction method characterized in that the above-mentioned unevenness in density is corrected by finding a transformation model on a lattice point using a model of = αx + β and interpolating the surrounding four transformation models (where N is Any positive integer, y is the pixel intensity of the image without uneven shading,
x is the pixel intensity of an image with uneven shading, α and β are functions of the location of the image (coefficients of the paralinear function). 2. As a method for correcting the uneven shading, when calculating a transformation model on the divided grid points, instead of calculating from the distribution of the histogram by the least squares method, data of an image with little uneven shading is used to correct the uneven shading. With respect to a reference point P set on the image for correction, point Q is found on the image without uneven shading, and by matching the histograms of the images around point P and point Q, uneven shading in the image can be corrected. 2. A digital image distortion correction method according to claim 1, characterized in that: . 3. As a method of correcting the unevenness of shading, when obtaining coefficients for parallel line approximation at each segmented grid point, calculate the coefficients in advance and use analog
Claim 1 or 2, characterized in that the data is stored in a memory within a digital conversion device.
Digital image distortion correction method described in Section 2.