KR100463551B1 - System for scaling digital image - Google Patents

Abstract

The present invention relates to an apparatus for scaling a digital image through cubic convolution interpolation, and more particularly, determines four input sample pixels to be used for interpolation from an input image, and determines a distance between a reference input sample pixel of the input image and a pixel position to be interpolated. After obtaining, calculates only three coefficients out of four coefficients by applying a sharpness factor to the distance value and outputs them to the interpolation filter, while subtracting the sum of the three coefficients obtained from constant 1 to obtain the remaining one coefficients. By outputting to the filter, the sum of the four coefficients is always equal to 1, which results in a smaller amount of computation with a flat field response. Therefore, in this case, the coefficient generator may generate DC-invariant coefficients even when implemented in hardware using fixed-point arithmetic.

Description

Digital image scaling device {System for scaling digital image}

TECHNICAL FIELD The present invention relates to a television receiver, and more particularly, to an apparatus for scaling a digital image through cubic convolution interpolation.

Recent advances in technology have enabled traditional TVs (eg NTSC / PAL) to support a variety of formats, including HDTV formats such as 720P (Pregressive) and 1080I (Interlece). In this case, since various formats such as 480P, 720P, and 1080I may be input as a conventional TV output, conversion to an output format is required.

Therefore, a scaler performing format conversion in a ready-to-use TV, a built-in digital TV, or a set-top box is essential.

In general, an inverse-mapping method of mapping a coordinate of an output image to a coordinate of an input image is used to implement a scaling function.

FIG. 1 illustrates such an inverse mapping method. Since an output image coordinate mapped to an input image is not mapped to one input image coordinate, a method of determining a pixel value of an output image using pixels of a neighboring input image coordinate need. Such a method is called interpolation, and the interpolation methods include Nearest Neighbor, Bi-Linear, and Cubic Convolution interpolation.

The Nearest Neighbor interpolation method is the simplest interpolation method in which the input pixel value located nearest to the output pixel value is mapped. However, this method has a problem of inducing a sawtooth effect at the boundary of the output image.

In addition, the Bi-Linear interpolation method obtains an output pixel with the sum of the weights of adjacent input 4 pixels, which is relatively easy to implement and inexpensive, but has a disadvantage of blurring the output image. Have.

Therefore, recently, a scaling method using cubic convolution has been considered to overcome these disadvantages.

2 shows an example of sampling in cubic convolution interpolation.

In FIG. 2, the sample data of the original image, that is, the input function is called f, and the interpolation function interpolated from the sample data of the original image is referred to as f. In this case, an interpolation function for an input having a uniform sample distance may be expressed as Equation 1 below.

In Equation 1, x is an interpolation point of an arbitrary pixel, Is the pixel position or interpolation node to use for interpolation at the current resolution, Is an interpolation kernel, or interpolation filter. And, Is a parameter that varies with input sample data. Is selected to be. In the case of the cubic convolution interpolation filter Becomes

In this case, the cubic convolution interpolation filter is composed of a divisional third-order polynomial only in a section defined by (-2, 1), (-1, 0), (0, 1), (1, 2), The number of input samples used for interpolation at 1 is reduced to 4 as shown in FIG. Also, the cubic convolution interpolation filter must be symmetric Is expressed by Equation 2 below.

According to Equation 2, eight parameters may be represented by only one parameter or two parameters depending on how the constraints are assigned. These equations are referred to as one-parameter models or two-parameter models, respectively.

Equations 3 and 4 below represent a one-parameter model and a two-parameter model, respectively.

In Equation 3, a represents a sharpness factor. And 0≤ | x | <1 is when the pixel to be interpolated is located between x _k and x _{k + 1} , and 1≤ | x | <2 is a case where a pixel to be interpolated is located between x _k and x _k−1 or between x _{k + 1} and x _{k + 2} .

Substituting b = 0 and c = -a in the two-parameter model of Equation 4 is equal to the one-parameter model of Equation 3 described above.

As in Equations 3 and 4, the cubic convolution interpolation filter is used only in a period of -2 <x <2. Since the interpolation continuous function of Equation 1 described above with respect to the cubic convolutional interpolation filter can be expressed as Equation 5 below.

In this case, the pixel resampling point to be resampled for the input having unit distance, that is, the position of the pixel x to be interpolated, is referred to as a reference input pixel point for each input pixel. It can be expressed as shown in Figure 2 using only the distance s from the resampling point x to. That is, when the distance between the input sample pixel and the pixel to be used for interpolation in FIG. 2 is quantified as 1, , , , to be. In other words, to be.

S and (1-s) in the section where 0≤ | x | <1, and (1 + s) and (2-s) in the section where 1≤ | x | <2, the equation of the one-parameter model Solving by substituting to 3, the one-parameter model cubic convolution interpolation function is expressed as Equation 6 below.

In the one-parameter model, coefficients to be multiplied by four input sample pixels are obtained as shown in Equation 7 below.

Reference input pixel position through the above equation (7) Knowing only the distance s from to resampling pixel position x, the coefficients multiplied by each interpolation node, i.e., four peripheral input sample pixels to be used for interpolation, ) Can be obtained.

On the other hand, the full-to-parameter model cubic convolution interpolation function of the above-described two-parameter model of Equation 4 can be obtained as shown in Equation 8 below.

In the two-parameter model, coefficients to be multiplied by four input sample pixels are obtained as in Equation 9 below.

Meanwhile, in US Pat. No. 5,930,407, a coefficient is multiplied by each interpolation node by applying a one-parameter model of Equation 7 by unifying the distance to s for an input pixel having a unit distance. )

FIG. 3 is a block diagram of a coefficient generator disclosed in US Pat. No. 5,930,407 described above, wherein four coefficients are applied by applying a one-parameter model. Equation (7) is implemented in hardware (H / W) as it is using eight multipliers, five subtractors, two adders, and one divider. Through this, it is possible to obtain the effect of reducing the amount of computation as compared to the conventional method.

In this case, the filter must have a flat-field response in order to be used as an interpolation filter. In other words, if the input is constant, the output must be constant. Coefficients obtained through Equation 7 ) Is the coefficient of the interpolation filter, so in this case, the sum of the coefficients multiplied by each interpolation node must be 1 to have a flat field response. In other words, Must satisfy

In addition, since the distance s is a real number, it is expensive to implement floating point operations in H / W. Therefore, it may be implemented in H / W using a fixed point operation in consideration of the final error. In this case, the coefficient generation method of the coefficient generator proposed in US Pat. No. 5,930,407 may be used. The case where this does not become 1 will arise. That is, in the case of US Pat. No. 5,930,407, the fixed point arithmetic operation does not have a flat field response because the step is determined and an error occurs.

In addition, US Pat. No. 5,671,298 proposes a technique for scaling vertically by implementing coefficients in S / W using a two-parameter model. However, this software method has a problem that cannot be used in the case of horizontal scaling such that every clock coefficient must be obtained.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and an object of the present invention is to provide a digital image scaling apparatus for implementing an interpolation filter to have a flat field response when implementing a coefficient generator in hardware using fixed-point arithmetic. .

Another object of the present invention is to provide a digital image scaling apparatus for generating coefficients of an interpolation filter so as to be applicable to both vertical scaling and horizontal scaling of a two-parameter model.

1 is a diagram illustrating a general inverse mapping method

2 is a diagram illustrating an example of sampling in a general cubic convolution interpolation.

3 is a detailed block diagram of a coefficient generator during cubic convolution interpolation in a conventional digital image scaling apparatus.

4 is a block diagram of a digital image scaling device according to the present invention;

5 illustrates a relationship between an input sample pixel and a resampling output pixel when interpolation is performed in a general vertical direction.

6 is a detailed block diagram of the coefficient generator of FIG. 4.

7 is a detailed block diagram of the interpolation filter of FIG.

Explanation of symbols for main parts of the drawings

401: line buffer 402: filter

403: interpolation filter 404: image resampling unit

405: coefficient generator

According to an aspect of the present invention, a digital image scaling apparatus according to the present invention determines four input sample pixels to be used for interpolation from an input image, and determines a distance between a reference input sample pixel of the input image and a pixel position to be interpolated. And calculating four coefficients of an interpolation filter using the distance values, and the interpolation filter outputs the interpolated pixels by applying corresponding coefficients to the four input sample pixels. Obtaining only three coefficients out of the four coefficients by applying a sharpness factor to the output to the interpolation filter, at the same time subtracting the sum of the three coefficients obtained from the constant 1 to obtain the remaining one coefficients to output to the interpolation filter Characterized in that it comprises a coefficient generator.

The coefficient generation unit implements hardware by applying the one-parameter model to the distance value s, thereby generating three coefficients. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( ) To obtain the output.

The coefficient generation unit implements hardware by applying the one-parameter model to the distance value s, thereby generating three coefficients. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( ) To obtain the output.

The coefficient generation unit implements hardware by applying the one-parameter model to the distance value s, thereby generating three coefficients. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( ) To obtain the output.

The coefficient generation unit implements hardware by applying the one-parameter model to the distance value s, thereby generating three coefficients. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( ) To obtain the output.

The coefficient generation unit implements hardware by applying the two-parameter model to the distance value s, thereby performing three coefficients. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( ) To obtain the output.

The coefficient generation unit implements hardware by applying the two-parameter model to the distance value s, thereby performing three coefficients. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( ) To obtain the output.

The coefficient generation unit implements hardware by applying the two-parameter model to the distance value s, thereby performing three coefficients. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( ) To obtain the output.

The coefficient generation unit implements hardware by applying the two-parameter model to the distance value s, thereby performing three coefficients. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( ) To obtain the output.

The coefficient generator generates four coefficients by applying the two-parameter model to the distance value s. Obtaining and outputting.

According to an aspect of the present invention, a digital image scaling apparatus includes a plurality of line buffers for storing and outputting pixels of an input image in units of lines according to size information of an input image and an output image, and interpolating the pixels output from the line buffers. A filter for outputting four input sample pixels in the vertical direction to be used for the image, and the coordinates of the input image mapped to the coordinates of the output image according to the size information of the input image and the output image, and by using the interpolation in the vertical direction An image resampling unit which determines four input sample pixels of a required input image and outputs a distance of an output pixel to be interpolated from a reference input sample pixel of the input image, and a one-parameter to a distance value output from the image resampling unit. Implement hardware with either model or two-parameter model to obtain three of four coefficients And a coefficient generator which calculates and outputs the other coefficient by subtracting the sum of the three coefficients obtained by the constant 1, and the coefficient generator is output to each of the four input sample pixels output from the filter. A vertical cubic convolution interpolation unit is configured as an interpolation filter which multiplies corresponding coefficients and adds all of them to output interpolated pixel values.

The digital image scaling apparatus according to the present invention includes a plurality of pixel delays for sequentially storing and outputting pixels of a line of input image, and four horizontal directions for filtering and interpolating the pixels output from the respective pixel delayers. The filter outputs the input sample pixels, and determines the coordinates of the input image mapped to the coordinates of the output image according to the size information of the input image and the output image, and uses the four inputs of the input image required for interpolation in the horizontal direction. The image resampling unit determines the sample pixels and outputs a distance of an output pixel to be interpolated from a reference input sample pixel of the input image, and a one-parameter model or a two-parameter model is used for the distance value output from the image resampling unit. Implement the applied hardware to obtain 3 coefficients out of 4 coefficients A coefficient generator which calculates and outputs the other coefficient by subtracting the sum of the three coefficients obtained by multiplying each of the four input sample pixels output from the filter and corresponding coefficients output from the coefficient generator, A horizontal cubic convolution interpolation unit is configured as an interpolation filter for outputting the pixel value.

Other objects, features and advantages of the present invention will become apparent from the following detailed description of embodiments taken in conjunction with the accompanying drawings.

Hereinafter, with reference to the accompanying drawings illustrating the configuration and operation of the embodiment of the present invention, the configuration and operation of the present invention shown in the drawings and described by it will be described as at least one embodiment, By the technical spirit of the present invention described above and its core configuration and operation is not limited.

4 is a block diagram illustrating an embodiment of a cubic convolution interpolation device in a vertical direction.

Referring to FIG. 4, a plurality of line buffers 401 storing input pixels in line units, and four input sample pixels in a vertical direction to be used for interpolation by filtering pixels output from each line buffer 401 ( A filter 402 for outputting the input image and the size information of the input image and the output image to determine the coordinates of the input image mapped to the coordinates of the output image, and use the pixels of the input image necessary for interpolation ( ), The reference pixel of the input image ( ) After image resampling unit 404 outputs the distance s _y of the output pixel being resampled from receives the video material away from the sampling portion 404 s _y input the obtained three coefficients of the four coefficients of the interpolation filter The coefficient generator 405 outputs the remaining coefficients by subtracting the sum of the three coefficients obtained by the constant 1, and outputs the coefficient generator 405 to each input sample pixel output from the filter 402. It is composed of an interpolation filter 403 for determining the value of the output pixel, that is, the interpolated pixel by multiplying the corresponding coefficients.

FIG. 5 illustrates a relationship between input sample pixels and resampling output pixels to be used for interpolation when performing cubic convolution interpolation in a vertical direction. Is the input sample pixels to use for interpolation, Is the reference input sample pixel to use for interpolation, Is the pixel interpolated by the cubic convolution interpolation method, and s _y is the reference input pixel position. Resampling point from Distance.

In this case, the cubic convolution interpolation may be performed only in the vertical direction according to the position of the pixel to be interpolated (for example, when the pixel position to be interpolated is located between two pixels in the vertical direction) or may be performed only in the horizontal direction. (E.g., when the pixel position to be interpolated is located between two pixels in the horizontal direction), both may be performed for the horizontal and vertical directions (e.g., the pixel position to be interpolated may be Located between four pixels). Here, when interpolation is performed in both the horizontal and vertical directions, the interpolation order in the horizontal and vertical directions may be changed.

In the present invention, a case of performing cubic convolution interpolation in the vertical direction is described as an embodiment.

In the present invention configured as described above, the image resampling unit 404 receives the size information of the input image and the output image, determines the coordinates of the input image mapped to the coordinates of the output image, and uses the same to determine the input image required for interpolation. Pixels ), The reference pixel of the input image ( ), The distance s _y of the output pixel to be resampled is obtained, and output to the coefficient generator 405.

On the other hand, a plurality of pixels, that is, eight line buffers 401 are stored with eight lines of pixels according to write control signals, and eight line buffers 401 according to read control signals. Eight pixels are output to the filter 402 at the same time. At this time, the read / write control signal of the line buffer 401 is determined by the size of the input image and the output image.

The filter 402 performs filtering on the input image read from the line buffer 401, and then performs pixels of the input image determined by the image resampling unit 404. ) Is output to the interpolation filter 403.

The interpolation filter 403 may include each input sample pixel output from the filter 402. Corresponding coefficients output from the coefficient generator 405 ), Then multiply each and add them together to determine the value of the output pixel, ) Is determined and printed.

7 is a detailed block diagram of the interpolation filter 403, in which Equation 10 below is implemented in hardware.

That is, FIG. 7 is composed of four multipliers 701 to 704 and one adder 705. In the multiplier 701, an input pixel is provided. And its coefficients Multiply by the multiplier 702 In the multiplier 703 In the multiplier 702 After the addition is added through the adder 705.

Meanwhile, the coefficient generator 405 receives a distance s _y from the image resampling unit 404 and a coefficient to be multiplied by the input pixels in the interpolation filter 403. In the present invention, three coefficients of the four coefficients of the interpolation filter are obtained through the one-parameter model of Equation 7 or the two-parameter model of Equation 9 described above, and then obtained by the constant 1 The sum of three coefficients is subtracted to obtain the other coefficient.

That is, the coefficients of the coefficient generator 405 are determined through one of the following methods.

1) After obtaining three coefficients by applying the one-parameter model of Equation 7, the remaining coefficients are obtained by subtracting the sum of the three coefficients from the constant 1.

2) After obtaining three coefficients by applying the two-parameter model of Equation 9, the remaining coefficients are obtained by subtracting the sum of the three coefficients from the constant 1.

3) Four coefficients by applying the two-parameter model of Equation 9 Is determined.

Typically, the filter must have a flat-field response in order to be used as the interpolation filter 403. this is It means to satisfy. E.g, Knowing three coefficient values, the values of the remaining coefficients can be determined. If the coefficients are obtained through this method, the sum of the four coefficients is always 1, so that the amount of computation is further reduced with a flat field response. Accordingly, DC-invariant coefficients are generated even when the coefficient generator 405 obtaining coefficients as described above is implemented in hardware using fixed-point arithmetic.

FIG. 6 is a detailed block diagram illustrating an embodiment of a coefficient generator according to the present invention. After obtaining three coefficients by applying a one-parameter model, the other coefficient is obtained by subtracting the sum of three coefficients from the constant 1. Yes.

Especially, The one-parameter model is implemented in hardware using seven multipliers (601 to 607), five subtractors (608 to 611,613), three adders (614 to 616), and one divider (612). It is composed of

That is, the first multiplier 601 squares the distance s, the second multiplier 602 multiplies the sharpness factor a by the distance s, and the third multiplier 603 multiplies (2a + 3) by The output of the first multiplier 601 is multiplied. The fourth multiplier 604 multiplies the distance s by the output of the first multiplier 601, the fifth multiplier 605 multiplies (a + 2) by the output of the fourth multiplier 604, and the sixth multiplier 606 multiplies the sharpness factor a by the output of the fourth multiplier 604. The seventh multiplier 607 multiplies 2a by the output of the first multiplier 601, the first subtractor 608 subtracts the output of the fifth multiplier 605 from the output of the third multiplier 603, The second subtractor 609 subtracts the output of the second multiplier 602 from the output of the first subtractor 609 and counts the coefficients. Outputs

The third subtractor 610 subtracts the output of the second subtractor 609 from a constant 1 The fourth subtractor 611 subtracts the output of the seventh multiplier 607 from the output of the sixth multiplier 606, and the divider 612 divides the output of the seventh multiplier 607 by two. . The fifth subtractor 613 subtracts the output of the sixth multiplier 606 from the output of the divider 612 and counts. Outputs The first adder 614 adds the output of the second multiplier 602 and the output of the fourth subtractor 611 to form a coefficient. Outputs The second adder 615 is a coefficient And coefficient And a third adder 616 outputs the third subtractor 610. And output of second adder 615 Add up coefficient Obtain

Here, Figure 6 In this embodiment, a one-parameter model is implemented in hardware by using. The selection of three coefficients obtained first and the coefficients when obtaining the other one using the three coefficients may be changed by the designer. In addition, hardware implemented according to the three coefficients to be selected will also vary.

E.g, Can be used to implement a one-parameter model in hardware, or Can be used to implement a one-parameter model in hardware, or You can also implement a one-parameter model in hardware using.

On the other hand, unifying the distance to s for an input with unit distance results in the coefficients in the two-parameter model. ) Is obtained as shown in Equation 9 above, Since the value of three coefficients is known, the values of the remaining coefficients can be determined. Here too, , , , Either of the two-parameter model may be implemented in hardware.

In this case, the two-parameter model can also obtain coefficients in hardware as in FIG. 6 described above, and thus can be applied to both horizontal and vertical scaling.

That is, in the present invention, vertical scaling is described as an embodiment, but horizontal scaling can be easily implemented by simply replacing a plurality of line buffers with a plurality of pixel delays. Here, the pixel retarder sequentially stores and delays pixels of one line.

The cubic convolution interpolation method of the present invention may be applied to a TV system, a computer graphics system, or the like to scale an input image.

As described above, according to the digital image scaling apparatus according to the present invention, three coefficients of four coefficients are first obtained by using a one-parameter model or a two-parameter model, and then one coefficient is a constant of one to three coefficients. By subtracting, the sum of the four coefficients is always equal to 1, which reduces the amount of computation with a flat field response. Therefore, in this case, the coefficient generator may generate DC-invariant coefficients even when implemented in hardware using fixed-point arithmetic. And by implementing hardware using a two-parameter model, it can be applied to both horizontal and vertical scaling.

Those skilled in the art will appreciate that various changes and modifications can be made without departing from the spirit of the present invention.

Therefore, the technical scope of the present invention should not be limited to the contents described in the embodiments, but should be defined by the claims.

Claims (15)

Determine four input sample pixels to be used for interpolation from an input image, obtain a distance between a reference input sample pixel of the input image and a pixel position to be interpolated, and then obtain four coefficients of the interpolation filter using the distance values; An interpolation filter is a digital image scaling device that outputs interpolated pixels by applying corresponding coefficients to the four input sample pixels. After applying only a sharpness factor to the distance value, only three coefficients of four coefficients are obtained and outputted to the interpolation filter, and the remaining one coefficient is obtained by subtracting the sum of the three coefficients obtained from the constant 1 and then performing the interpolation filter. And a coefficient generator for outputting the digital image scaling apparatus. The method of claim 1, wherein the coefficient generator Implement the hardware by applying the one-parameter model represented by the following equation to the distance value s to obtain and output three coefficients, and obtain the other coefficient by subtracting the sum of the three coefficients from the constant 1 Digital image scaling device, characterized in that. here, Is the coefficient of the four interpolation filters output from the coefficient generator, and a is the sharpness factor. The method of claim 2, wherein the coefficient generator Three coefficients are implemented by implementing the hardware by applying the one-parameter model to the distance value s. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( Digital output scaling device characterized in that the output to obtain. The method of claim 2, wherein the coefficient generator Three coefficients are implemented by implementing the hardware by applying the one-parameter model to the distance value s. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( Digital output scaling device characterized in that the output to obtain. The method of claim 2, wherein the coefficient generator Three coefficients are implemented by implementing the hardware by applying the one-parameter model to the distance value s. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( Digital output scaling device characterized in that the output to obtain. The method of claim 2, wherein the coefficient generator Three coefficients are implemented by implementing the hardware by applying the one-parameter model to the distance value s. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( Digital output scaling device characterized in that the output to obtain. The method of claim 2, wherein the coefficient generator A first multiplier that takes a distance s and squares it, A second multiplier that multiplies the sharpness factor a by the distance s, A third multiplier that multiplies (2a + 3) by the output of the first multiplier, A fourth multiplier that multiplies the distance s by the output of the first multiplier, a fifth multiplier for multiplying (a + 2) by the output of the fourth multiplier, A sixth multiplier that multiplies the sharpness factor a by the output of the fourth multiplier, A seventh multiplier that multiplies 2a by the output of the first multiplier, A first subtractor for subtracting the output of the fifth multiplier from the output of the third multiplier, A coefficient of subtracting the output of the second multiplier from the output of the first subtractor A second subtractor for outputting Subtract the output of the second subtractor from constant 1 With the third subtractor, A fourth subtractor for subtracting the output of the seventh multiplier from the output of the sixth multiplier, A divider for dividing the output of the seventh multiplier by two; The output of the divider is subtracted from the output of the sixth multiplier A fifth subtractor for outputting Coefficient of the second multiplier plus the output of the fourth subtractor With a first adder for outputting Coefficient output from the first adder And coefficients output from the fifth subtractor With a second adder that adds Output from the third subtractor And the output from the second adder Add up coefficient And a third adder for outputting the digital image processing apparatus. The method of claim 1, wherein the coefficient generator Implement the hardware by applying the two-parameter model represented by the following equation to the distance value s to obtain and output three coefficients, and obtain the remaining one coefficient by subtracting the sum of the three coefficients from the constant 1 Digital image scaling device, characterized in that. here, Is the coefficient of the four interpolation filters output from the coefficient generator, and a is the sharpness factor. The method of claim 8, wherein the coefficient generator Three coefficients are implemented by implementing the hardware by applying the two-parameter model to the distance value s. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( Digital output scaling device characterized in that the output to obtain. The method of claim 8, wherein the coefficient generator Three coefficients are implemented by implementing the hardware by applying the two-parameter model to the distance value s. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( Digital output scaling device characterized in that the output to obtain. The method of claim 8, wherein the coefficient generator Three coefficients are implemented by implementing the hardware by applying the two-parameter model to the distance value s. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( Digital output scaling device characterized in that the output to obtain. The method of claim 8, wherein the coefficient generator Three coefficients are implemented by implementing the hardware by applying the two-parameter model to the distance value s. At the same time, the sum of the three coefficients is subtracted from the constant 1 and the other coefficient ( Digital output scaling device characterized in that the output to obtain. The method of claim 8, wherein the coefficient generator Four coefficients are implemented by applying the two-parameter model to the distance value s. Obtaining and outputting a digital image scaling device. A plurality of line buffers storing and outputting pixels of the input image in units of lines according to size information of the input image and the output image; A filter for filtering the pixels output from the line buffers and outputting four input sample pixels in a vertical direction to be used for interpolation; The coordinates of the input image mapped to the coordinates of the output image are determined according to the size information of the input image and the output image, and four input sample pixels of the input image required for interpolation in the vertical direction are determined using the input image. An image resampling unit for outputting a distance of an output pixel to be interpolated from a reference input sample pixel; Implement the hardware by applying the one-parameter model or the two-parameter model to the distance value output from the image resampling unit to obtain and output three coefficients out of four coefficients, the sum of the three coefficients obtained from the constant 1 A coefficient generator which calculates and outputs the other coefficient by subtracting the subtraction; And A digital image comprising a vertical cubic convolution interpolation unit configured as an interpolation filter for multiplying each of the four input sample pixels output from the filter by corresponding coefficients output from the coefficient generator and adding all the interpolated pixel values. Scaling device. A plurality of pixel delayers sequentially storing and outputting pixels of a line of an input image; A filter for filtering the pixels output from the pixel delay units and outputting four input sample pixels in a horizontal direction to be used for interpolation; The coordinates of the input image mapped to the coordinates of the output image are determined according to the size information of the input image and the output image, and four input sample pixels of the input image necessary for interpolation in the horizontal direction are determined using the input image. An image resampling unit for outputting a distance of an output pixel to be interpolated from a reference input sample pixel; Implement the hardware by applying the one-parameter model or the two-parameter model to the distance value output from the image resampling unit to obtain and output three coefficients out of four coefficients, the sum of the three coefficients obtained from the constant 1 A coefficient generator which calculates and outputs the other coefficient by subtracting the subtraction; And A horizontal image cubic convolution interpolation unit is configured as an interpolation filter that multiplies each of the four input sample pixels output from the filter by corresponding coefficients output from the coefficient generator and adds all the interpolated pixel values. Scaling device.