JPH06113120A - Resolution conversion device for picture processor - Google Patents

Abstract

PURPOSE:To provide a resolution conversion device capable of reducing the occurrence of a picture defect on account of the non-uniformity of MTF caused in a linear interpo lation system, and generating the enlarged picture of satisfactory picture quality. CONSTITUTION:In a picture processor provided with an interpolation arithmetic part 10 which interpolates multilevel picture data by using an interpolation coefficient to be set in accordance with a resolution conversion magnification, a linear filter part 9 which can switch a coefficient 8 in accordance with the resolution conversion magnification as the preprocessing of an interpolation operation is inserted, plural picture elements in a one-dimensional direction are correlated and the interpolation operation between two points is executed. Then, a linear filter part and the interpolation arithmetic part are synthesized and incorporated, and linear interpolation among multiple points is executed. Thus, thinning can be prevented even in the case of reduction at a minimum magnification with such constitution. For reducing the non- uniformity of MTF for respective picture elements, which occurs in a linear interpolation processing between two points, a linear filter processing operation is executed on picture data before the interpolation processing and a gain near a Nyquist frequency Ny being a frequency band where the non-uniformity of MTF is large is reduced so as to improve uniformity.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus such as a fax machine or a copying machine, and in particular, by multiplying and adding an image data of each pixel by an interpolation coefficient representing a positional influence degree between pixels. The present invention relates to a resolution conversion device of an image processing device that performs linear interpolation processing to reduce / enlarge an image.

[0002]

2. Description of the Related Art In a copying machine, an original of an arbitrary size is copied in accordance with a predetermined paper size or a predetermined size so that it can be fitted in a predetermined area. Magnification function is provided. For example, in the case of a device using a CCD line sensor, this reduction / expansion function changes the driving speed of the scanning optical system in the sub-scanning direction and thins out or two points to the image data in the main scanning direction read from the CCD line sensor. Interpolation processing is performed.

As described above, there is proposed a method for obtaining reduced image data by interpolation calculation from original image data according to a magnification, not by simply thinning or increasing the image by image enlargement and reduction. In order to prevent image defects such as jaggies in the edge portion and moire in the periodic structure portion such as halftone dots, the density preserving type such as the linear interpolation method and the projection method is considered to be advantageous. In particular, in terms of hardware feasibility, it is said that the linear interpolation method is particularly advantageous, and many systems use this method.

For example, Japanese Laid-Open Patent Publication No. 62-169278, in which interpolation data is prepared and the interpolation data to be output is read according to the set condition, and the writing or reading start address of the image data to the line memory is used. Japanese Patent Laid-Open No. Sho 62-2, which is changed according to reduction processing.
No. 21272, JP-A-62-62, wherein the substantial reading resolution of the image reading means is changed according to the recording resolution.
No. 221274, JP-A No. 62-212275, the data enlarged by interpolation of the input image data
Japanese Patent Application Laid-Open No. 2-174463 and the like have been proposed in which reduced image data is output by thinning out dot units. However, these include, for example, the problem that the expansion and contraction can be performed only at a rate of 1/2 ⁿ increments, the problem that a clock having a cycle n times as large as the image data transfer clock is required when reducing to 1 / n, etc. was there. These problems are solved by the two-point arithmetic interpolation method described in Japanese Patent Laid-Open No. 2-161872.

FIG. 9 is a diagram for explaining an interpolation algorithm between two points, FIG. 10 is a block diagram for explaining the structure of a reduction / enlargement processing section, and FIG. 11 is a data processing operation at the time of reduction / enlargement. FIG. 12 is a diagram for explaining an example in which gradation changes occur even when the input image data is the same.

The reduction / enlargement algorithm of the two-point arithmetic interpolation method uses interpolation coefficients to synthesize dense data from pixel data of two points before and after, as shown in FIG. The pixel data of is combined and reduced. In the case of reduction processing, the solid line in FIG.
As shown in (a), the read data is written in the image memory as it is, and at the time of reading the written data, data control and interpolation between two points are performed to reduce the image. Also,
In the case of the enlargement processing, as shown by the broken line in FIG. 10 and FIG. 11B, data control and interpolation between two points are performed on the read data, and the result is written in the memory. And
The image is enlarged by reading the written data as it is.

For example, when the 125% enlargement process shown in FIG. 9 is performed, the value of pixel 2.4 is (1-0.4) × (value of pixel 2) + 0.4 × (value of pixel 3) = 0 .6 * (value of pixel 2) + 0.4 * (value of pixel 3). Here, pixel 2 is multiplied by 0.6 and pixel 3
Is multiplied by the positional influence degree of the pixels 2 and 3. The influence of the pixel 2 is large because it is close to the pixel 2 and the influence of the pixel 3 is far from the pixel 3. The impact is getting smaller. In this way, the value of the pixel 2.4 is obtained by multiplying the influence degrees of the pixels 2 and 3 and adding them.

Assuming that the decimal point position of the interpolation coefficient used in this case is 16 bits and the precision of the interpolation coefficient is 13 bits (8192), (100/100) × 8192 = 001 from the calculation when the scaling factor is 1. It becomes 0.0000000000000. Therefore, when 12.5%, such as (100 / 12.5) × 8192 = 1000.0000000000000, 16 bits of the interpolation coefficient are exceeded, so that it cannot be used. Here, the 3 bits above the decimal point are prepared so that they can be used from a magnification of 12.5% or more, and the 13 bits below the decimal point are prepared to improve accuracy. For example, 13%
In case of (100/13) × 8192 = 111.1011000100111, it can be used because it fits in 16 bits of the interpolation coefficient.
The maximum magnification can be set until the interpolation coefficient becomes 0 in calculation.

For example, the method of obtaining the interpolation coefficient when the expansion rate is 125% is as follows: (100/125) × 8192 = 6553 = 1999h = 0001.1001.1001.1001 Eye 6553 = 000 | 1.100 | 1.1
001.1001 = 0.1100 = 1/2 + 1/4 =
0.75 2nd dot 13106 = 001 | 1.001 | 1.0
01.0010 = 0.1001 = 1/2 + 1/16 =
0.5625 3rd dot 19659 = 010 | 0.110 | 0.1
100.1011 = 0.0110 = 1/4 + 1/8 =
0.375 4th dot 26212 = 011 | 0.011 | 0.0
110.0100 = 0.0011 = 0.1815 5th dot 32765 = 011 | 1.111 | 1.1
111.1101 = 0.1111 = 0.9375 6th dot 39318 = 100 | 1.100 | 1.1
It is obtained as follows: 001.0110 = 0.1100 = 0.75.

[0010]

However, in each of the above-mentioned conventional methods, only two or three adjacent original pixels are used in the calculation of the interpolation data, so that reduction of a low magnification is performed, for example. At this time, a large number of original images that are not used in the interpolation calculation occur, which causes a problem such as a thin line being omitted.

Further, depending on the coefficient at the time of interpolation, the MTF
Since the (Modulation Transfer Function) changes significantly, there is a problem that image deterioration occurs. Further, in the linear interpolation method, since the MTF for each pixel also changes due to a change in the interpolation coefficient for each pixel, MTF non-uniformity occurs at a specific cycle according to the expansion / contraction ratio, and compression in the one-dimensional direction occurs. There was a problem that this became apparent as stripes at the time of expansion. In particular, it is remarkable at a scaling ratio close to 100%.

For example, a case of 101% enlargement in only the one-dimensional direction will be described as an example. FIG. 13 shows the interpolation coefficient and MTF at the time of expansion / contraction.
It is a figure which shows the change of. When the linear interpolation between two points is performed,
As shown in FIG. 13, MTF non-uniformity occurs in each pixel,
This is amplified by post-processing such as edge enhancement processing, and as a result, streak-like noise becomes apparent at 100 pixel intervals.

The present invention has been made to solve the above-mentioned problems, and the number of input pixels to the interpolation calculation unit is increased to reduce image defects due to MTF non-uniformity generated by the linear interpolation method, and to reduce good image quality. An object of the present invention is to provide a resolution conversion device of an image processing device that generates a magnified image.

[0014]

To this end, the present invention provides an image for resolution conversion of multi-valued image data, which is provided with an interpolation calculation unit for interpolating multi-valued image data using an interpolation coefficient set according to a resolution conversion magnification. In the processing device, a linear filter unit is inserted as a pre-process of the interpolation calculation.

Further, the linear filter section is characterized in that the coefficient can be switched according to the resolution conversion magnification and is applied only in the one-dimensional direction. Further, the linear filter unit and the interpolation calculation unit are integrated and integrated to perform multipoint linear interpolation, and after performing correlation between a plurality of pixels, a two-point interpolation calculation is performed. To do.

[0016]

In the resolution converting apparatus of the image processing apparatus according to the present invention, since the linear filter section is inserted as the preprocessing of the interpolation calculation, the interpolation calculation can be performed after the correlation between the pixels is taken in the linear filter section. It is possible to prevent thinning even in the case of reduction at the minimum magnification. Further, in order to reduce the MTF non-uniformity for each pixel that occurs in the two-point linear interpolation process, the image data before the interpolating process is subjected to a linear filter processing operation, and the frequency band near the Nyquist frequency Ny, which is a large frequency band with the MTF non-uniformity, It is possible to reduce the gain of the above, improve the uniformity, and suppress the occurrence of image defects.

[0017]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing an embodiment of a resolution conversion device of an image processing apparatus according to the present invention, FIG. 2 is a diagram showing a configuration example of a filter applied to the resolution conversion device of an image processing apparatus according to the present invention, FIG. Is a diagram showing an example of filter coefficients according to the conversion magnification, FIG. 4 is a diagram showing frequency characteristics of a linear filter when the coefficients of FIG. 3 are set, and FIG.
It is a figure which shows the change of F.

In FIG. 1, the register 2 sets the reciprocal value of the scaling factor for the scaling process and the interpolation between two points, and the register 8
Sets a linear filter coefficient according to the scaling factor. These scaling factors and linear filter coefficients are stored in the CPU 1
Generated and set by. The adder 3 is the register 2
The reciprocal value of the magnification set to is added every pixel clock. The register B4 stores the interpolation coefficient B obtained as a value below the decimal point at that time.
The register A11 stores the interpolation coefficient A obtained by the calculation of 1-B. The read pattern generation circuit 5 generates a read pattern of the image input signal stored in the line memory 7 from the reciprocal value of the magnification, and the address control circuit 6 controls the read address of the line memory 7 in advance according to the pattern. Is. The filter processing unit 9 performs a filter process by a convolution (sum of products) operation using the image input signal read from the line memory 7 and the coefficient set in the register 8. The two-point interpolator 10 multiplies the interpolation coefficients A and B and the two-point filtered signal by multipliers 13 and 14, respectively.
After being multiplied by, the sum is added by the adder 15 and is sent as an interpolation signal.

The interpolation signal sent from the two-point interpolation unit 10 is address-controlled by the address control circuit 17 according to the pattern generated by the write pattern generation circuit 16 from the reciprocal value of the magnification, and is written in the line memory 8, and finally. It is output as an image output signal after the various scaling conversion processes.

As described above, the resolution conversion apparatus of the image processing apparatus according to the present invention is roughly divided, and the interpolation coefficients A and B are obtained for each pixel according to the scaling factor, and the interpolated pixel is sandwiched between the two. It is composed of a two-point interpolating unit that performs scaling processing by multiplying the pixel data of the points, and a linear filter processing unit that is located in front of it and that sets a weighting coefficient according to the scaling ratio.

In the present invention, the linear filter section is inserted as preprocessing for interpolation as described above, MTF nonuniformity occurs in each pixel as in the conventional case, and this is amplified by postprocessing such as edge enhancement processing. As a result, the occurrence of streak-like noise at 100 pixel intervals is improved.

When the sub-scanning speed can be changed according to the scaling ratio, interpolation processing in the sub-scanning direction becomes unnecessary, so that the filter configuration is, for example, as shown in FIG.
A one-dimensional filter applied only in the main scanning direction of × 3 (4) is used. This filter is not limited to one dimension and may be two dimensions. Further, the size of the filter is not limited to that of this embodiment, and a filter with a larger size may be used.

FIG. 2 applied only to the 1 × 4 main scanning direction.
FIG. 3 shows an example of a set of one-dimensional filter coefficients of, and FIG. 4 shows these frequency characteristics. In this example, a coefficient having a relatively low pass degree is selected at a magnification ratio in which streak noise of 51% to 120% is relatively conspicuous, and noise of 25% to 50% is inconspicuous, but lines or the like disappear. For easy magnification, a coefficient with a stronger low-pass degree is selected. On the other hand, 121%
At a magnification of up to 200%, a coefficient that slightly attenuates the high-frequency component is selected, and at a coefficient of 201% to 400%, a through coefficient is selected to prevent deterioration of the sharpness and definition of the image due to the deterioration of the high-frequency component. We have selected.

Regarding the filter size, 2
Considering prevention of line disappearance at the time of 5% reduction, a size of 1 × 4 is required, but 12.5%
When it comes to preventing the disappearance of the lines up to, it is necessary to have about 1 × 8 even if it is simply considered.

In FIG. 3, if, for example, the coefficient is selected, a filter having a low-pass frequency characteristic as shown in FIG. 4 can be constructed. This filter drops the Nyquist frequency Ny band completely to zero, half of that Ny /
Even in the band of 2, the gain is reduced to 1/2. Therefore, the total MTF characteristic of the processing by this filter and the interpolation processing between two points becomes a characteristic as shown in FIG. 5B, which is compared with the MTF characteristic of the single interpolation processing between two points shown in FIG. As a result, the MTF nonuniformity can be greatly improved. As a result, even if a post-processing such as edge enhancement is performed, the generation of stripe noise can be significantly reduced.

However, if such a filter having a strong low-pass characteristic is uniformly applied to all scaling factors, it adversely affects the sharpness and the definition reproduction of the image.
The present invention employs a configuration in which the filter coefficients are switched according to the scaling factor as shown in representative examples 1 to 3 in FIG.

As described above, according to the present invention, by introducing the linear filter in which the coefficient according to the expansion / contraction ratio is set before the two-point linear interpolation section, the MTF non-uniformity caused by the two-point linear interpolation in the subsequent stage is introduced. This has the effect of reducing the level and suppressing the occurrence of image defects.

That is, FIG. 13 shows the change of the interpolation coefficient and MTF for each pixel at the time of 101% enlargement, and A and B show the interpolation coefficients of each of two points. MTF at this time
Is based on MTF (f) = | A + B · e ^{j * f} |, the Nyquist frequency Ny (f = 0.5) and its half frequency Ny /
At 2 (f = 0.25), ignoring the phase, it is described as follows.

MTF (Ny) = (A ² −2AB + B ² ) ^1/2 MTF (Ny / 2) = (A ² + B ² ) ^{1/2 From} these relationships, the MTF as shown in FIG.
Characteristics are required. As can be seen from this, the MTF (N
y) = 0, MTF (Ny / 2) = 0.5, while
At AorB = 1.0, MTF (Ny) = 1.0, MT
Since F (Ny / 2) = 0.71, the MTF nonuniformity of 0.3 at the maximum of Ny / 2 frequency and 1.0 at the maximum of Ny frequency in the image after the expansion / conversion is 101% enlarged. Occurs at 100 pixel intervals. This is emphasized by post-processing or the like, and becomes a streak when the image is expanded or contracted in the one-dimensional direction.

As a method for improving this, in the present invention, a linear filter is provided before the two-point linear interpolation section. As a result, the MTF nonuniformity level can be reduced as shown in FIG. That is, the gain in the vicinity of Ny, which is a frequency band having a large non-uniformity of MTF, is reduced to improve the uniformity, and the occurrence of image defects is suppressed.

FIG. 6 is a diagram showing a concrete configuration example of the filter processing section. In this embodiment, the 1 × 4 one-dimensional filter processing unit inserted as the pre-processing of the interpolation unit between two points is
As shown in FIG. 6, the latches 21-1 to 21-4 hold the input image data for four adjacent pixels, and the input image data is stored in the CPU.
The correlation coefficient between the four points set in the filter coefficient register 8 from 1 is multiplied by the multipliers 22-1 to 22-4, and then added by the adder 23 to perform the convolution operation, and the interpolator between the two points. Send to. On the other hand, in the two-point interpolating unit, the reciprocal of the scaling factor is set from the CPU 1 in the reciprocal scale register 2, and the adder 3 and the interpolating coefficient generating unit 4'provide 2
When the inter-point interpolation coefficient is generated, the multipliers 13 and 14 and the adder 15 perform a predetermined interpolation calculation.

When the coefficient set in the filter processing section is as shown in FIG. 3, the coefficient of A is assigned to the pixel of interest, the coefficient of B is assigned to the two pixels before and after, and the coefficient of C is assigned to the pixel immediately before. To be Therefore, if the coefficient of A is set to 1 and the other coefficients are set to 0 as in the example of 201% to 400%, the pixel data of the pixel of interest is sent to the 2-point interpolation unit as it is, and the same result as the conventional case is obtained. To be Since thinning by interpolation between two points does not occur at the time of expansion, there is almost no problem even with such coefficient setting. However, in the case of 50% or less,
Since thinning occurs, the coefficient A,
A setting of 0.25 is required for each of B and C.

The reduction / expansion interpolation method has a drawback that the image quality is deteriorated by the magnification. This depends on the interpolation coefficient of the two-point interpolation unit. For example, when the coefficient is 1, the input image is output as it is, but when the coefficient is 0.5, the average value of two pixels before and after is output. Therefore, an output slightly different from the input image can be obtained.

In the case of normal expansion / contraction such as 120%, 150%, and 75%, the interpolation coefficient changes from time to time, so the image quality is not so affected, but it is close to the same magnification of 100.1%, for example. In the case of reduction / expansion, the interpolation coefficient changes gradually for several pixels to several tens of pixels continuously, so if the calculation accuracy in the two-point interpolation unit is poor, when inputting image data having uniform density, Similar to the case of the deterioration of the MTF, uneven density may occur in a cycle of a specific number of pixels, and streak-shaped image defects may occur. Therefore, in the embodiment of FIG.
Each of the multipliers 13 and 14 is configured to add each 8 bits obtained by cutting off the decimal part from the multiplication value of 12 bits, but for example, if each multiplication value of 12 bits is added as it is,
It is possible to prevent the influence of unevenness due to this calculation accuracy. The coefficient setting example of 51% to 200% shown in FIG. 3 is for correcting the change in the MTF of the interpolating unit between two points by the four-point correlation calculating unit (filter processing unit) in the preceding stage.

FIG. 7 is a diagram showing another embodiment of the resolution conversion apparatus of the image processing apparatus according to the present invention. In the above-described embodiment, the linear filter unit and the inter-point interpolation unit are separated, but FIG. 7 shows an example in which these two processing units are integrated. Now, if the coefficients of the 1 × 3 linear filter are a, b, and a, and the interpolation coefficient between two points is A, B, the processing of both can be integrated by convolution calculation, and the coefficients aA, bA + aB, aA + bB, a.
It becomes four 1 × 4 filters of B. The filter coefficient calculation circuit 32 obtains four coefficients aA, bA + aB, aA + bB, and aB from the interpolation coefficients A and B and the filter coefficients a and b for each pixel of interest. The processing is performed by the filter processing unit 3
It is 1. Note that the filter coefficient calculation circuit 32 and the filter processing unit 31 may be integrated into a calculation table that receives the conversion magnification, the interpolation coefficient, and the signals of a plurality of adjacent pixels.

FIG. 8 is a diagram for explaining the effect of this embodiment in an example of generating 25% reduced data. When a low-magnification reduction is performed by the 2-point interpolation method, thinned-out original pixels occur. For example, when the reduction is performed by about 25%, four pixels are reduced to one pixel. Conventionally, this output one pixel is generated by interpolating from two original pixels before and after as shown in FIG. To be done. That is, two pixels that are not used for interpolation are thinned out. The number of thinned pixels increases as the magnification decreases. According to the present invention, as shown in FIG. 8 (b), at the time of reduction at the minimum magnification, correlation between pixels is obtained by filtering from a number of pixels that do not cause thinning, and then expansion / contraction processing is performed by a two-point interpolation method. Therefore, it is possible to reflect the signal of the pixel, which has been thinned out in the past, in the signal after the reduction. In the second embodiment of FIG. 7, the correlation calculation by such filter processing and the interpolation calculation between two points are made to be the same as in FIG.
As shown in (c), the reduced and expanded data is directly interpolated from the data of n points.

The present invention is not limited to the above embodiment, but various modifications can be made. For example, in the above-described embodiment, an example in which two-point interpolation is performed according to the scaling factor is shown, but not limited to scaling, multi-valued image data is interpolated using the interpolation coefficient set according to the resolution conversion factor. It goes without saying that the same may be applied to the case.

[0038]

As is apparent from the above description, according to the present invention, since the linear filter unit is inserted as the preprocessing of the interpolation calculation, the interpolation calculation is performed after the correlation between the pixels is taken in the linear filter unit. Therefore, it is possible to prevent thinning even when the image is reduced at the minimum magnification. Also, in order to reduce the MTF non-uniformity for each pixel that occurs in the linear interpolation processing between two points at a magnification close to 1 ×, linear filtering processing operation is performed on the image data before the interpolation processing, and a frequency with a large MTF non-uniformity is applied. It is possible to reduce the gain near the Nyquist frequency Ny, which is the band, to improve the uniformity and suppress the occurrence of image defects.

Moreover, since the coefficient of the linear filter portion can be switched according to the resolution conversion magnification, the reproducibility of the image can be controlled according to the magnification / reduction magnification, the occurrence of image defects can be suppressed, and the image of the magnification / reduction magnification can be suppressed. It is also possible to prevent deterioration of sharpness and definition. Further, since the one-dimensional filter is used, it is possible to reduce the image defect that has occurred at the time of reduction and expansion by the conventional interpolation method with a simple configuration.

Further, the signals of the target pixel and a plurality of pixels adjacent to the target pixel on the scanning line are input, and the product of each signal of the plurality of pixels and the coefficient given corresponding to each pixel is calculated. A linear filter unit that performs a sum-of-products operation and outputs a signal of the pixel of interest, a unit that generates an interpolation coefficient according to the position of the pixel of interest, a constant preset according to the resolution conversion magnification, and the interpolation coefficient Since a means for giving a coefficient of the filter section based on the interpolation coefficient generated by the generating means is provided, pixels to be simply thinned out at the time of reduction, thin line omissions and image defects can be reduced. Since the uniformity of the MTF is maintained, a reduced / enlarged image with uniform image quality can be obtained.

[Brief description of drawings]

FIG. 1 is a diagram showing an embodiment of a resolution conversion apparatus of an image processing apparatus according to the present invention.

FIG. 2 is a diagram showing a configuration example of a filter applied to a resolution conversion device of an image processing device according to the present invention.

FIG. 3 is a diagram showing an example of a filter coefficient according to a conversion magnification.

FIG. 4 is a diagram showing frequency characteristics of a linear filter when the coefficients shown in FIG. 3 are set.

FIG. 5 is a diagram showing changes in an interpolation coefficient and MTF during expansion / contraction.

FIG. 6 is a diagram illustrating a specific configuration example of a filter processing unit.

FIG. 7 is a diagram showing another embodiment of the resolution conversion apparatus of the image processing apparatus according to the present invention.

FIG. 8 is a diagram illustrating an effect of the present embodiment in an example of generating 25% reduced data.

FIG. 9 is a diagram for explaining an interpolation algorithm between two points.

FIG. 10 is a block diagram for explaining a configuration of a reduction / enlargement processing unit.

FIG. 11 is a diagram for explaining a data processing operation at the time of reduction / enlargement.

FIG. 12 is a diagram for explaining an example in which a gradation change occurs even when the input image data is the same.

FIG. 13 is a diagram showing changes in interpolation coefficient and MTF during expansion / contraction.

[Explanation of symbols]

1 ... CPU, 2 ... Reciprocal magnification register, 3, 15 ... Adder, 4, 11 ... Register, 5 read pattern generating circuit, 6, 17 ... Address control circuit, 7, 18 ... Line memory, 8 ... Filter coefficient register, 9 ... Filter processing unit, 10 ... Interpolation unit between two points, 12 ... Latch, 13, 14 ...
Multiplier, 16 ... Writing pattern generation circuit

Front page continued (72) Inventor Hiroyuki Oyabu 2274 Hongo, Ebina City, Kanagawa Prefecture Fuji Xerox Co., Ltd.

Claims (4)

[Claims] 1. An image processing apparatus for converting the resolution of multi-valued image data, comprising an interpolation calculation section for performing interpolation calculation of multi-valued image data using an interpolation coefficient set according to a resolution conversion magnification, and linearly in the preceding stage of the interpolation calculation. A resolution conversion apparatus for an image processing apparatus, comprising a filter section. 2. The resolution of the image processing apparatus according to claim 1, wherein the linear filter unit has means for changing a frequency characteristic according to a set coefficient and switching the coefficient according to a resolution conversion magnification. Converter. 3. The linear filter unit performs a product-sum operation for summing products of respective image signals of a plurality of pixels adjacent to each other on a scanning line and coefficients corresponding to the respective pixels of the plurality of pixels. The resolution conversion device of the image processing apparatus according to claim 1. 4. A signal of a target pixel and a plurality of pixels adjacent to the target pixel on a scanning line is input, and each product of each image signal of the plurality of pixels and a coefficient given corresponding to each pixel. A linear filter unit for performing a sum-of-products operation to output the signal of the pixel of interest, a means for generating an interpolation coefficient according to the position of the pixel of interest, a constant preset according to the resolution conversion magnification, and the interpolation A resolution conversion apparatus for an image processing apparatus, comprising: means for providing a coefficient of the filter unit based on the interpolation coefficient generated by the coefficient generation means.