JP2724224B2 - Digital image signal scaling processing method and apparatus - Google Patents

Description

Description: Object of the Invention (Field of Industrial Application) The present invention relates to a method and an apparatus for processing a digital image signal for scaling.

(Prior Art) In the case of performing image scaling processing, that is, enlargement / reduction processing, in a digital image processing system, it is generally necessary to convert the discretization density of a digital image signal. That is, at the time of enlargement, the discretization density is increased, and at the time of contraction, the discretization density is reduced.

As a conversion method of the discretization density, a thinning method, an interpolation method, and the like are known. The thinning method converts a discretized density by thinning out a part of the original signal (increasing in case of enlargement). This method has a very simple circuit configuration, but has the disadvantage that the accuracy of the signal is reduced. Further, when the original image has a peak of a high frequency component such as halftone dot printing or the like, a peak of a low frequency component which is not present in the image due to aliasing, that is, a so-called moray array may be generated, and the image quality is significantly deteriorated.

As shown in FIG. 2, the interpolation method uses a linear function or a sinc function to interpolate a signal at a grid store corresponding to a new sampling density from an original signal defined on a grid point corresponding to the original sampling density. Go and ask for it. In the interpolation using a linear function (linear interpolation), since the multiplication and addition of coefficients are performed between adjacent pixels, the circuit scale is slightly increased as compared with the thinning method, but the accuracy of the signal is increased. In this method, moire noise is reduced, but cannot be completely removed.

Interpolation using the sinc function (sinc interpolation) is to perform interpolation using the sinc function of the Nyquist frequency corresponding to the smaller of the discretized density of the original signal and the discretized density of the converted signal. In this method, since there is no aliasing, no moarrays occur. However, since the sinc function has a long tail, it has the disadvantage that the number of terms in the interpolation operation increases and the number of operations increases greatly. Therefore, in the case of real-time processing such as pipeline processing, there is a problem that the circuit scale increases.

As for the projection method, for example, as proposed in Japanese Patent Application No. 62-297969, a grid-like boundary of a scaled image (hereinafter, referred to as a scaled image) is an image before scaling (hereinafter, referred to as an original image). The density of each pixel of the scaled image is interpolated by adding the densities of all pixels of the original image onto which each pixel of the scaled image is projected and weighted by the projected area. Moire noise does not occur in principle,
A disadvantage is that the circuit scale is slightly larger. Also, if the number of bits of the interpolation operation is reduced to reduce the circuit scale, the sum of the interpolation coefficients multiplied by the original pixel in order to obtain the value of the scaled pixel is not constant. (Referred to as "grazing noise"), and the image quality is significantly degraded.

On the other hand, the scaling factor is rounded to 2 ⁿ / k (k and n are integers),
When the interpolation calculation is performed with the precision of n bits after the decimal point, no grazing noise is generated, but the substantial magnification changes due to the influence of rounding. For this reason, in order to perform scaling processing with a high-precision magnification, the number of bits n after the decimal point must be set to a sufficiently large number, and as a result, there is a problem that the circuit scale increases.

(Problems to be Solved by the Invention) As described above, in the related art, when the thinning-out method or the linear interpolation method is used for the conversion of the discretization density for performing the image scaling processing, moire noise occurs, and the image quality is reduced. May deteriorate.

In addition, in the interpolation or projection method using the sinc function, the image quality does not deteriorate due to the more noise, but the number of operations increases because the number of terms of the interpolation operation is required, and the number of bits of the interpolation operation is reduced in the projection method. There is a problem in that the response of the DC component of the original image causes glare noise due to fluctuation.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a method and an apparatus for suppressing the generation of moir noise and grazing noise and performing a scaling process of a digital image signal with a high-precision scaling ratio by a simple process with a small number of operations. .

[Constitution of the Invention] (Means for Solving the Problems) According to the present invention, when scaling a digital image signal, the pixel position A _j (corresponding to the re-discrete position) of the scaled image signal is highly accurate (n-bit accuracy) And the pixel position A _j is m
First coarse pixel position signal rounded to bit precision (m <n)
A _j to 'seek further first magnification rounded designated magnification coarse pixel position signal A _j or specified magnification second coarse pixel position signal and a coarse reciprocal with rounded reciprocal of B _j,' Ask. Then, the first and second coarse pixel position signals A _j ′, B _j ′
From the reference pixel range i of the original digital image signal
(K ≦ i <l) and original image signals X _k , X at both ends of the reference range
_The scaled image signal _Yj is obtained by determining the weighting coefficients a and b for _l .

According to one aspect, a scaling processing device according to the present invention includes a cumulative addition unit that cumulatively adds a constant determined by a designated scaling factor to a predetermined initial value and outputs an n-bit-accurate pixel position signal; Means for obtaining a first coarse pixel position signal by rounding the pixel position signal output from the MPU to m-bit (m <n) accuracy, and a coarse scaling factor obtained by rounding the first coarse pixel position signal and a designated scaling factor. Means for obtaining a second pixel position signal; first multiplication means for multiplying an original digital image signal by an interpolation coefficient determined by the first and second coarse pixel position signals to obtain a plurality of multiplication results; It comprises an adding means for adding the result, and a second multiplying means for multiplying the addition result of the adding means by a normalization coefficient determined by a coarse scaling factor to output a digital image signal subjected to scaling processing.

According to another aspect, the original digital image signal is multiplied by an interpolation coefficient determined by the first and first coarse pixel position signals and the coarse scaling factor, instead of the first and second multiplying means, to perform a plurality of multiplications. Multiplication means for obtaining the result is used. In that case,
By adding the plurality of multiplication results by the adding means, a digital image signal subjected to scaling processing is output.

(Operation) In the scaling process according to the present invention, since smoothing is performed with a width substantially equal to the re-discretization interval of the digital image signal, moire noise hardly occurs as described later.

Further, the interpolation operation is performed with a small number of bits using the rounded coarse pixel position signal and the coarse scaling ratio, but the pixel position A _j of the scaled image signal that determines the accuracy of the scaling ratio is calculated with high precision. Therefore, the magnification can be set with high accuracy without increasing the number of operations.

Further, in the conventional projection method, the second pixel position signal is A _{j + 1}
Because of the rounding relationship between the first pixel position signal A _j and the second pixel position signal A _{j + 1} , the interpolation function for obtaining the scaled image becomes inconsistent and the DC gain changes. In contrast to the occurrence of the grazing noise, the present invention calculates the second coarse pixel position signal B _j ′ by calculating the rounded first coarse pixel position signal A _j ′ and the rounded scaling factor. Since the reference region of the original pixel per one pixel of the double image is constant, the value of the interpolation coefficient is constant, and no grazing noise occurs.

That is, according to the present invention, by simple processing with a small number of operations, it is possible to perform a scaling process with a high-precision scaling ratio while suppressing the occurrence of moiré noise and faint noise.

(Example) Hereinafter, an example of the present invention will be described.

FIG. 3 shows a configuration of an image input apparatus using the present invention. This apparatus includes an image reading unit 11 and a signal processing unit 12, and the present invention is applied to the signal processing unit 12.

FIG. 4 shows the configuration of the image reading unit 11. The linear region of the image on the document 22 proved by the light source 21 is imaged on the light receiving surface of the line sensor 24 by the contact optical system 23 using the gradient index cylindrical lens array, and is converted into an electric signal. .
The line sensor 24 is constituted by, for example, a CCD, and sequentially outputs one line of analog image signals by main scanning. This image signal is amplified by an amplifier 25, and then converted to an A / D conversion circuit 26.
Is converted into a digital image signal 27. With this configuration, the light source 21, the optical system 23, and the line sensor 24 are integrated and moved vertically (sub-scanning) with respect to the carriage main scanning direction, and the reading of each line is repeated while the original 23 is read.
Can be obtained.

The digital image signal 27 obtained by the image reading unit 11 is subjected to digital processing such as correction and conversion in the signal processing unit 12. In the signal processing unit 12, first, the normalization circuit 13 normalizes the image signal.
This normalization circuit 13 is also called a shading correction circuit, and is used in the main scanning direction of the image reading unit 11 due to unevenness in sensitivity and dark current of each light receiving element of the line sensor 24, uneven illuminance in the main scanning direction of the illumination light source, and the like. The specific configuration is described, for example, in Japanese Patent Application No. 59-199263. The normalization circuit 13 normalizes the image signal to be "1" when the original is white and "0" when the original is black. Also, line sensor
In an apparatus that reads a color image using a color sensor 24, the white balance is also automatically corrected by the standardization circuit 13.

Next, the image signal standardized by the standardization circuit 13 is enlarged / reduced by the enlargement / reduction circuit 14 in the main scanning direction.
A reduction process, that is, a scaling process is performed, and an image signal having a predetermined discretized density is output. As a method of changing the magnification, a method of changing the imaging magnification of the reading optical system to perform enlargement / reduction has conventionally been used, but a moving mechanism for enlarging / reducing the imaging optical system takes up space. Disadvantages. In this apparatus, enlargement / reduction in the sub-main inspection direction is performed by changing the main inspection speed of the carriage, and enlargement / reduction in the main scanning direction is performed by the enlargement / reduction circuit 14.

Hereinafter, the scaling process performed by the enlargement / reduction circuit 14 will be described. In the scaling process, the scaled image signal Y _j (j is an integer) of the j-th pixel is converted from the original image signal X _i (i is an integer of k ≦ i <l) of the i-th pixel by the following equation (1). Get)

Here, r is a designated scaling factor given by a user or the like, and the reciprocal of this designated scaling factor r is q, and this value is m
Q 'is the value rounded to bit precision, n-bit precision (m <n)
Let A _j = j · q + c be the pixel position of the scaled image signal Y _j calculated in
(C is a constant), and a first coarse pixel position signal obtained by rounding A _j to m-bit precision is a second coarse pixel calculated from A _j ′, A _j ′ and q ′ rounded to m-bit precision. Let the position signal be B _j ′ = A _j ′
+ Q '. Here, k and a represent the integer part and the decimal part of A _j ′, respectively, and l and b represent the integer part and the decimal part of B _j , respectively. R 'is a value obtained by rounding the reciprocal of q' to m-bit precision, that is, a scaling factor rounded to m-bit precision.

Therefore, equation (1) is equivalent to the first and second coarse pixel position signals.
A _j ', B _j' reference original image signal X _i by the scope i (k ≦ i
And <l), the original image signal X _k of the reference range ends, weighting of X _l coefficients a, seeking b, represents calculating a scaled image signal Y _j.

The meaning of the expression (1) will be described in more detail with reference to FIG. FIG. 1 is a diagram for explaining an operation method in a scaling processing method according to an embodiment of the present invention.

When the specified scaling ratio r is 11/15, the bit precision m is 3
Then, since the reciprocal of r is q = 15/11 = 1.36, when this is rounded, q '= 10/2 ³ = 10/8, and the rounded scaling factor is r' = 8/10 = 0.8. Since the constant c corresponds to the deviation of the origin between the original image and the scaled image, the constant c may be appropriately selected depending on the position of the end of each image.

Hereinafter, a calculation example of the scaled image signal _Yj of each pixel will be described.

When j = 0, k = a = 0 from A ₀ = A ₀ ′ = 0 (c = 0), and l = 1, b = 0.25 from B ₀ ′ (= A ₀ ′ + q) = 1.25. Become. Accordingly, scaled image signal Y ₀ of j = 0 becomes _{Y 0 = 0.8 (X 0 +0.25} 1).

In the case of j = 1, since A ₁ = 1.36, A ₁ ′ = 1.25, B ₁ ′ (= A ₁ ′ + q) = 2.5, k = 1, a = 0.25, l = 2, and b = 0.25. magnification image signal Y ₁ is, Y ₁ = 0.8 - a _{{(0.25X 1) + X 1} + 0.25X 2} = 0.8 (0.75X 1 + 0.5X 2).

For _{j = 2 A 2 = 2.72,} A 2 '= 2.625, B 2' (= A 2 '+ q') = 3.875
More, k = 2, a = 0.625 , l = 3, b = 0.875 , and the scaled image signal Y _{2 is, Y 2 = 0.8 {(-} 0.625X 2) + X 2 + 0.875X 3} = 0.8 (0.375X ₂ + 0.875X ₃ ).

When j = 3, k = 4, a = 0, l = 5, b = 0, 25 from A ₃ = 4.08, A ₃ ′ = 4.0, B ₃ ′ (= A ₃ + q ′) = 5.25. Double image signal Y ₃
Becomes Y ₃ = 0,8 (X ₄ + 0.25X ₅ ), and the scaled image signal Y _j is sequentially obtained in the same manner.

Or in the scaling process described, the accuracy of the magnification of the scaled image is determined by the accuracy of the pixel position signal A _j (n-bit precision). To increase the accuracy of A _j , it is only necessary to increase the number of bits of an adder (to be described later) for adding q, so that the scaling ratio can be set with high accuracy with small hardware.

In the present invention, the following equation (2) is used instead of equation (1).
, The same scaling process can be performed.

Here, r ′ is a value obtained by rounding the designated scaling factor r to m bits,
k is the largest integer satisfying k · r + c <j (c is a constant),
a 'is a value obtained by rounding a to m bits, where a is the decimal part of (k + 1) r. Further, l ′ = INT [(1-
a ′) / r ′], b = (1−a ′) − l ′ · r ′, and l =
Let k + 1 '+ 1. Note that INT [] represents the value in [] as an integer part.

The meaning represented by the equation (2) will be described with reference to FIG.

When the designated scaling factor is r = 11/15 = 0.73, if the bit precision m is 3, r '= 5/8. Since the constant c corresponds to the deviation of the origin between the original image and the scaled image, as in the case of the equation (1), it may be appropriately selected depending on what position of the end of each image is considered.

When j = 0 k = -1 (c = 0), a = a '= 0, INT [1 / (5/8)]
= From 1, l = 1, b = 1-5 / 8 = 3/8 _{next, Y 0 = 5 / 8X 0} +
3 / 8X ₁

When j = 1, k = 1, a = 0.46, a '= 3/8, l' = 0, l = 2, b = 5/8
Next, becomes _{Y 1 = 3 / 8X 1 +} 5 / 8X 2, sequentially obtained scaled image signal Y _j is.

The precision of the scaling factor in the scaling process of equation (2) is k · r
Is determined by the accuracy of To improve the accuracy of kr, r
It is only necessary to increase the number of bits of the adder for adding .times.

FIG. 6 is a block diagram showing a specific configuration of an enlargement / reduction circuit to which a scaling process according to equation (1) or (2) is applied. This circuit has two line buffers 104, 10
Has 6, the original image signal with the transfer clock CK ₁ in one line buffer is written, the original signal is read out from the other line buffer read clock CK _2, latches 109 to 111
Is forwarded to The read clock CK ₂ is the transfer clock CK ₁
This signal is synchronized with the internal clock CK ₃ (not shown) which is twice or more faster than the internal clock CK ₃ .

The cell tractors 101 and 108 perform read / write switching of the line buffers 104 and 106, and the selectors 102 and 103 switch input clocks of the address generators 105 and 107 that generate address data to the line buffers 104 and 106.

The original image signals X _{1 to} X ₃ output from the latches 109 to 111 are
After multiplication by the interpolation coefficients P _{1 to} P ₃ output from the controller 117 by the multipliers 112 to 114, they are added by the adder 115.
Then, the output result of the adder 115 is multiplied by the normalization coefficient by the multiplier 116, and further the transfer clock C
By being latched by the latch 119 at K _1, scaled image signal Y corresponding to the formula (1) is obtained. Here, the normalization coefficient is r ′ in equation (1).

The controller 117 calculates the interpolation coefficients P _{1 to} P ₃ according to the equation (1).
And it outputs a read clock CK _2. The value of each interpolation coefficients, P ₃ is b, P ₂ is 1 or 1-a, P ₁ is 1-a or 0
It is. a and b are those used in equation (1). Read clock CK _2, see the position of the pixel of one of the interpolation operation per original image signal is output by the amount of moving (number of pixels).

FIG. 7 is a functional block diagram of a portion related to the generation of the interpolation coefficients P _{1 to} P ₃ in the controller 117.
0, first and second coarse pixel position signal generators 201 and 202, and an interpolation coefficient generator 203. The cumulative addition unit 200 is a cumulative addition unit that calculates the pixel position signal A _j of the scaled image signal Y _j shown in Expression (1), and performs the cumulative addition of A _j = j · q + c (c is a constant). The first coarse pixel position signal generator 201 generates a first coarse pixel position signal A _j ′ by rounding the pixel position signal A _j from the accumulator 200 to m-bit precision. Further, the second coarse pixel position signal generation unit 202 calculates B _j ′ = A _j ′ + q ′ from A _j ′ and q ′ rounded to m-bit precision to obtain a second coarse pixel position signal B _j ′. Generate These first and second
The coarse pixel position signals A _j , B _j ′ are input to the interpolation coefficient generation unit 203, where the interpolation coefficients P ₁ , P ₂ , P ₃ are generated.

According to the enlargement / reduction circuit of FIG. 6, it is also possible to perform the scaling process by the equation (2). In that case, the interpolation coefficient P ₁
The to P _3, P ₃ is 1-b, P ₂ is a 'or r', P ₃ is 0 or a ', may be selected an output signal of the adder 115 by the selector 118 as a scaled image signal. In practice, equation (1)
It is sufficient if any one of the scaling processes (2) can be performed.

With the above configuration, it is possible to output a scaled image signal in synchronization with the transfer clock of the original image signal.

In the enlargement / reduction circuit shown in FIG. 6, a reduction ratio of 1/2 or less cannot be achieved, but for a reduction ratio of less than 1/2, the number of latches and multipliers and the number of bits of an adder are increased. Internal clock CK ₃ per cycle of transfer clock CK ₁
Can be made possible by increasing the number of periods.

Next, the reason why the moire noise is extremely unlikely to be generated by the zooming method of the present invention will be described.

In general, it is well known that, when a continuous signal is discretized at a certain discretization interval P, a frequency component of f / 2 or more on the original signal shifts to f / 2 or less, that is, aliasing occurs. f is the discretization frequency, which is the reciprocal of the discretization interval P. Therefore, when a continuous signal is discretized, aliasing is prevented by removing a frequency component of f / 2 or more of the original continuous signal by a low-pass filter before discretization.

However, aliasing often causes a problem. If the original continuous signal has a peak of a high frequency component such as a carrier for modulation, the peak is shifted to a lower frequency band by aliasing, thereby deteriorating the signal quality. The moire phenomenon occurs. For example, in the case of an image, a halftone dot of a print which is hardly visually perceived is shifted to a low-frequency region having high visibility by folding, so that a moire net which does not originally exist on the image is observed, and the image quality is significantly deteriorated. . Therefore, if only the peak of the frequency that causes moiré is removed, the deterioration of the image quality is reduced in most cases.

By the way, the scaling process is, in other words, a conversion process of the discretized density, which can be regarded as a combination of continuity of the original discretized signal, convolution with a certain function, and re-discretization. For example, linear interpolation scheme assumes that continuous signals having values only at discrete grid points as shown in the original discretization signal X ₁ a Figure 8 (a) of the discretization interval (original discrete interval) Px, which The linear interpolation function h having the impulse response shown in FIG.
(T) is convolved to obtain the signal Z shown in FIG. 8 (b).
(T) can be regarded as a signal _Yj shown in FIG. 8 (c), which is re-discretely digitized at a new discretization interval (re-discretization interval) Py.

On the other hand, variable in multiples method considers likewise continuous signal the original discrete signal X _i, Fig. 10 forme function h 'tatami a (t) of this with an impulse response shown in FIG. 9 (b) of the present invention A function h ″ (t) having an impulse response shown in FIG. 9 (c) is convolved with the signal Z ′ (t) shown in FIG. 9A, where the smoothing width q ′ · Px is almost equal to the re-discrete interval Py. The signal Z ″ (t) shown in FIG. 10 (b) is re-discretized at the re-discretization interval Py to calculate Y _j shown in FIG. 10 (c). I have. Here, q 'is a value obtained by rounding the reciprocal of the designated scaling factor r with an appropriate bit precision. This series of processing is shown in the following equation (3).

Z (t) = h (t) * X (t) Z '(t) = h' (t) * X (t) Z "(t) = h" (t) * Z '(t) (* (Representing convolution integral) (3) FIGS. 11 (a) and 11 (b) show the frequency characteristics of the linear interpolation and the scaling process according to the present invention, respectively. These are the Fourier transforms H (f) and h '(t) of h (t), respectively.
* Fourier transform H ′ (f) · H ″ (f) of h ″ (t).

Here, fx and fy are reciprocals of the original discretization interval Px and the re-discreteization interval Py, respectively. The former has zeros only at the frequency fx, while the latter has zeros very close to the frequencies fx and fy. Now, it is assumed that the original discretized signal has a peak at a frequency of fy + Δf (Δf is a frequency close to 0). This can occur not only when the original continuous signal has the frequency component but also when the low frequency component (fx-fy-Δf) of the original signal is convoluted by discretization and shifted to a lower frequency Δf. These frequency components are convoluted by re-discretization and shifted to a lower frequency, and moire noise occurs. In the linear interpolation, the frequency component of fx + Δf is reduced to some extent by the interpolation processing, but since the response at the frequency fy increases depending on the magnification, it cannot be completely eliminated, and moire noise still occurs.

On the other hand, in the scaling process of the present invention, the frequency response of the interpolation process is almost 0 at fy, and therefore, the frequency component of fy + Δf which generates Moire noise in the immediate vicinity is almost removed before the re-discretization. Does not occur.

Next, the reason why grazing noise does not occur even when the bit precision is reduced according to the present invention will be described. Here, the grazing noise refers to grid-like noise caused by fluctuation of the sum of interpolation coefficients when obtaining a scaled image.

FIGS. 12 (a) to 12 (c) show various scaling methods in a case where each pixel of the original image is divided into 3 bits (8 divisions). Here, the designated scaling factor is 11/15. Also,
Both ends of the reference range of each pixel of the scaled image with respect to the original image are represented by a _j and b _j , respectively.

The scaling method shown in FIG. 12 (a) is a method in which the interpolation calculation accuracy of the conventional projection method is reduced, and the reciprocal q of the designated scaling factor is set to
= 10.9 / 8, and a scaled image is obtained as a _j = b _j −1 = fn (j · q). However, fn () means rounding with 3 bits after the decimal point. In this scaling method, one of the scaled images is
The original pixel reference area per pixel is not always constant.
For this reason, grazing noise occurs.

The scaling method shown in FIG. 12 (b) is a method of approximating the scaling factor by 2 ⁿ / k (n and k are integers), approximating the reciprocal q ′ of the scaling factor by 10/8, and a _j = A scaled image is obtained as b _j −1 = j · q.
According to this method, since the original pixel reference area per pixel of the scaled image is constant (q ′), no grazing noise is generated. However, since the scale itself is rounded, the scale cannot be set with high accuracy. Further, in order to set the scaling ratio with high precision, the value of the number of division bits n between pixels must be increased, so that the circuit scale is increased.

The scaling method shown in FIG. 12 (c) is a method according to the embodiment of the present invention, in which the reciprocal q of the designated scaling factor r is represented by 15/11 = 10.9 / 8, and q '= fn (q), a _j = fn (j · q), b j = a j + q '
To obtain a scaled image. Therefore, according to the method of the present invention, since the reference region of the original pixel per one pixel of the scaled image is constant (q '), no grazing noise occurs. Moreover, the re-discrete position (resampling position) a _{j of} the scaled image
Since the calculation itself is performed using high-precision q, a high-precision scaling ratio can be set without increasing the circuit scale so much.

In the above-described embodiment, only the conversion of the one-dimensional discretized density is performed as the scaling process, but the present invention is not necessarily limited to this. For example, the original image signal defined on the zy axis is first enlarged / reduced in the x-axis direction, and then enlarged / reduced in the y-axis direction.
Two-dimensional enlargement / reduction can be performed. In this case, zooming associated with movement of the carriage need not be performed. Of course, the scaling process by the scaling method and the scaling operation by moving the carriage may be used together.

Further, in the above embodiment, the scaling processing (discrete density conversion processing) is performed by the signal processing circuit. However, the same effect can be obtained even if the processing is executed by software using the CPU according to the same conversion formula. Needless to say. In addition, the present invention can be variously modified and implemented without departing from the gist.

[Effects of the Invention] According to the present invention, it is possible to perform scaling processing while suppressing the occurrence of moiré noise and glare noise with a configuration having a small number of operations suitable for real-time processing of a signal, and to achieve high-precision scaling. Can be set to

[Brief description of the drawings]

FIG. 1 is a diagram for explaining an operation method in an embodiment of a scaling method according to the present invention, FIG. 2 is a diagram showing the principle of an interpolation method,
FIG. 3 is a block diagram showing a configuration of an image input device to which the present invention is applied, FIG. 4 is a block diagram showing a configuration of an image reading unit in FIG. 3, and FIG. FIG. 6 is a block diagram showing the configuration of the enlargement / reduction circuit in FIG. 3 according to the present invention, and FIG. 7 is a diagram showing the generation of interpolation coefficients in the controller in FIG. Functional block diagram of the part related to
FIG. 8 is a diagram showing the process of linear interpolation, FIG. 9 is a diagram showing the impulse response of the linear interpolation and the scaling method according to the present invention, FIG. 10 is a diagram showing the process of the scaling method according to the present invention, FIG. 11 is a diagram showing the frequency response of the linear interpolation and the scaling method according to the present invention, and FIG. 12 is a diagram showing the relationship between the scaled image and the reference area of the original image in various scaling methods. 14 enlargement / reduction circuit (magnification processing device) 112 to 114 first multiplier 115 adder 116 second multiplier 200 cumulative adder 201 first coarse pixel Position signal generator 202... Second coarse pixel position signal generator 203... Interpolation coefficient generator

Claims (3)

(57) [Claims] 1. A method for scaling a digital image signal according to a designated scaling factor, comprising the steps of: calculating a pixel position of the scaled image signal with n-bit precision; and calculating the calculated pixel position with m bits (m <n). A) obtaining a first coarse pixel position signal by rounding to an accuracy; and a coarse regression obtained by rounding the first coarse pixel position signal and the specified variable magnification, or a coarse reciprocal obtained by rounding a reciprocal of the specified variable magnification. Calculating a second coarse pixel position signal from the first and second coarse pixel position signals; and calculating the scaled image signal in accordance with the first and second coarse pixel position signals. Double processing method. 2. A cumulative addition means for cumulatively adding a constant determined by a designated scaling factor to a predetermined initial value to output a pixel position signal with n-bit accuracy; (M <n) means for obtaining a first coarse pixel position signal by rounding to an accuracy; and rounding the first coarse pixel position signal and the designated magnification, or a reciprocal of the designated magnification. Means for obtaining a second coarse pixel position signal from the coarse reciprocal, and first multiplication for multiplying an original digital image signal by an interpolation coefficient determined by the first and second coarse pixel position signals to obtain a plurality of multiplication results Means for adding the plurality of multiplication results; and multiplying the addition result obtained by the addition means by a normalization coefficient determined by the coarse scaling factor to output a digital image signal subjected to scaling processing. To the power of Scaling processing apparatus in a digital image signal, characterized by and means. 3. A cumulative addition means for cumulatively adding a constant determined by a designated scaling factor to a predetermined initial value to output a pixel position signal with n-bit accuracy, and an m-bit pixel position signal output from said cumulative addition means. (M <n) means for obtaining a first coarse pixel position signal by rounding to an accuracy; and a second coarse pixel position signal from the first coarse pixel position signal and the coarse scaling factor obtained by rounding the designated scaling factor. Obtaining means, multiplying means for multiplying an original digital image signal by an interpolation coefficient determined by the first and second coarse pixel position signals and the coarse scaling factor to obtain a plurality of multiplication results, and adding the plurality of multiplication results. And an adder for outputting a digital image signal subjected to a scaling process.