JPH06309452A - Resolution converting processor - Google Patents

Abstract

PURPOSE:To provide a picture resolution converting processor without the generation of moire by making only a character image definite and smoothing a dot picture. CONSTITUTION:An edge judging part 5 judges whether an attentional picture element is in the vicinity of a linear edge or not from the input picture data of plural picture elements around the attentional picture and a 1st interpolating processing part 2 calculates a picture element value on an optional picture element position by interpolating the input picture data. A high band emphasis part 3 executes the clearing processing of the input picture data an a 2nd interpolating processing part 4 inputs and interpolates the output of the emphasis part 3 and calculates a picture element value on the same picture element position as that of the 1st interpolating processing part 2. An adder 8 executes the weighting operation of the results of the 1st and 2nd interpolating processing parts 2, 4 in accordance with the degree of existence of the remarked picture element in an edge area by means of the output of the edge judging part 5.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image resolution conversion processing device in an image input / output device or an image processing device.

[0002]

2. Description of the Related Art Image resolution conversion processing is a technique required to obtain an output image of the same size as an input original image when the resolutions of an input device and an output device are different, or an image called zooming. This is also a technique used in the enlarging / reducing process. Conventionally, this type of processing apparatus has been described in the literature: Image Analysis Handbook (The University of Tokyo Press) pp.
441 to 444 (the nearest neighbor interpolation method,
A method using a colinear linear interpolation method and a cubic convolution interpolation method) is generally used. These techniques will be described below.

First, FIG. 7 shows a principle diagram of the nearest neighbor interpolation method. Here, it is assumed that the given original image is I, the pixel position in the main scanning direction is i, and the pixel position in the sub scanning direction is j. That is, I _{i, j} represents a pixel value indicating the luminance or density of the original image at the pixel (i, j). In addition, the coordinates of the point to be interpolated are (i + Δi, j + Δ
j). Δi and Δj are 0 ≦ Δi <1 and 0 ≦, respectively.
It is an arbitrary real number given by Δj <1. As shown in FIG. 7, the nearest-neighbor interpolation method is to obtain the image data of the original image closest to the point to be interpolated, and is expressed by the following equation. I _{i +} Δ _{i, j +} Δ _j = I _{m, n} where m = [i + Δi + 0.5], n = [j + Δj +
0.5]. Here, [] represents a Gaussian symbol and indicates that only an integer part is taken by truncation. In the case of FIG. 7, since (i, j + 1) is the closest pixel, I _{i +} Δ _{i, j +} Δ _j = I _{i, j + 1} . Since the nearest neighbor interpolation method is configured only by calculating the address of the pixel to refer to, the hardware scale is small,
It has the advantage of being easy to implement. However, the output image using this method is rough and the image quality is degraded. In particular, there is a drawback in that the deterioration in the vicinity of a portion (edge) in which the gradation changes sharply, which is included in a character image or the like, becomes remarkable.

Next, FIG. 8 shows a principle diagram of the bilinear interpolation method. In the bilinear interpolation method, the image data to be obtained is obtained by using original image data of four points surrounding the point to be interpolated. The calculation formula is as follows. _{I i + Δ i, j +} Δ j = (1-Δi) {(1-Δj) I i, j + ΔjI i, j + 1} + Δi {(1-Δj) I i + 1, j + ΔjI i + 1, _As is clear from the above equation, the _{j + 1} } co-linear interpolation method is a weighted average using the distance between the point to be interpolated and four surrounding points as a weight, and the output image is a smoothed version of the original image. It will be what you did.
Therefore, although the image quality of the output image is high in the region where the gradation change is gradual, such as a photographic image, in the edge region, even when this method is used, the edges are blurred due to the smoothing effect and the image quality is deteriorated. There are drawbacks.

Finally, FIG. 9 shows the principle of the cubic convolution interpolation method. In the third-order convolution interpolation method, the image data to be obtained is obtained using 16 points of original image data surrounding the point to be interpolated. The calculation formula is as follows.

[0006]

[Equation 1] However, f (t) = sin (πt) / (πt) ≈1-2 | t | ² + | t | ³ (0 ≦ | t | <1) 4-8 | t | +5 | t | ² − | t | ³ (1 ≦ | t | <2) 0 (2 ≦ | t |) At this time, x ₁ = 1 + Δi y ₁ = 1 + Δj x ₂ = Δi y ₂ = Δj x ₃ = 1−Δi y ₃ = 1− Δj x ₄ = 2-Δi y ₄ = 2-Δj. Since the cubic convolutional interpolation method uses a sin function as an interpolation function, it has an advantage that an image is smoothed and sharpened at the same time, but the amount of calculation required for one interpolation processing is reduced. Many are not easy to implement with hardware. Further, the sharpening effect of this method does not function to the extent that edges are preserved, and even in the case of using this method, the edges are blurred due to the smoothing effect in the edge region, resulting in deterioration of image quality.

[0007]

In the resolution conversion method using the interpolation method described above, there is a problem in reproducibility of edges included in a character image or the like regardless of which method is used. In order to improve this problem, a method in which original image data is subjected to enhancement processing in advance and this image data is interpolated may be considered. This can be easily realized, for example, by applying a high-frequency emphasis type filter as shown in FIG. 10, whereby the reproducibility of the edge of the character image can be improved. The filter shown in FIG. 10 is a filter generally called a Laplacian type filter, which performs weighted addition shown in the figure on image data of (3 × 3) pixels and sets the result as the central pixel data. Since this filter has a characteristic of emphasizing high frequency components, it contributes to sharpening of an image. However, a new problem occurs by performing the above processing. That is, the above-described high-frequency emphasis type filter also emphasizes a halftone dot printed image in the same manner, which causes a problem that moire occurs in the halftone dot printed image and the image quality is extremely deteriorated. That is, even with any of the methods described above, there is no image resolution conversion method that does not cause moire by sharpening only the character image and smoothing the halftone image.

The present invention solves the problems described above,
An object of the present invention is to provide an image resolution conversion processing device that does not generate moire by sharpening only a character image and smoothing a halftone image.

[0009]

In order to solve the above-mentioned problems, the present invention provides a resolution conversion processing apparatus for interpolating a pixel value at an arbitrary position on an image using a multi-valued image as an input image to obtain an input image signal. Edge determining means for determining whether the pixel of interest is near a linear edge, and first interpolation processing means for interpolating the input image signal to calculate a pixel value at an arbitrary pixel position, High-frequency emphasis processing means for performing sharpening processing on the input image signal, and pixels at the same pixel position as the first interpolation means by inputting the output of the high-frequency emphasis processing means and interpolating and interpolating the output. A second interpolation processing means for calculating a value; and means for performing weighted addition on the results of the first and second interpolation processing using the output of the edge determination means. And

In implementing the present invention, as the edge determining means, the filter coefficient of the pixel adjacent to the target pixel in the main scanning and sub-scanning directions is set to 0, and the filter coefficient of the pixel diagonally adjacent to the target pixel is a negative value. Then, using a filter in which the filter coefficient of the pixel of interest is set so that the total sum of the filters is 0 is suitable for the determination of the linear edge region which is the feature of the character image region.

[0011]

According to the resolution conversion processing apparatus of the present invention, the edge determining means uses the image data of a plurality of pixels centering on the target pixel as input data to determine whether the target pixel is near a linear edge from the input image data. A determination is made, and the first interpolation processing means interpolates the input image data to calculate a pixel value at an arbitrary pixel position. The high-frequency emphasizing unit performs a sharpening process on the input image data, and the second interpolation processing unit inputs the output of the high-frequency emphasizing unit and interpolates it, which is the same as the first interpolation unit. The pixel value at the pixel position of is calculated.
The adding means performs a weighting calculation on the results of the first and second interpolation processing using the output of the edge determining means, and the interpolation result of the first interpolation means and the second interpolation processing. The interpolation result after the sharpening process, which is the output of the inserting unit, is weighted according to the degree to which the pixel of interest exists in the edge region.
Therefore, it is possible to obtain a clear image that has been subjected to the emphasis processing for a character image and a smooth gradation image that has not been subjected to the emphasis processing for a photographic image or a halftone dot printed image with a small amount of calculation.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows the outline of the configuration of a resolution conversion processing apparatus in an embodiment of the present invention. In FIG. 1, reference numeral 1 is an input terminal for inputting image information of an original document decomposed into pixel units by a scanner or the like at a multi-valued level, from which image data of a plurality of pixels centering on a target pixel is input. Reference numeral 2 is a first interpolation processing unit for interpolating pixel values at designated positions from original multi-valued image data input from the input terminal 1, and 3 is original multi-valued image data input from the input terminal 1. A high-frequency emphasizing unit for performing high-frequency emphasizing processing, 4 is a second high-frequency emphasizing image data output from the high-frequency emphasizing unit 3 by interpolating pixel values at the same position as the first interpolation processing unit 2 Interpolation processing section 5, an edge determination section for determining whether or not the target pixel is near an edge from the original multi-valued image data input from the input terminal 1, and 6 is a first output from the first interpolation processing section 2. A first multiplier that multiplies the first interpolated image data and the first weighting coefficient output by the edge determination unit 5, and 7 is the second interpolated image data output by the second interpolation processing unit 4. And a second weighting coefficient output from the edge determination unit 5
, 8 is the first interpolated image data weighted by the edge amount output from the first multiplier 6 and the second interpolated image data weighted by the edge amount output by the second multiplier 7. An adder for calculating the sum with the inserted image data, and 9 is an output terminal for outputting the multi-valued image data subjected to the resolution conversion by the present resolution conversion processing device to the outside, and the output of the adder 8 is output from here. .

FIG. 2 shows the relationship between the pixel position of the input image and the pixel position of the output image. When calculating the pixel value at the position (i + Δi, j + Δj) existing in the shaded area in the drawing as the output image, the position (i,
j) image data of 9 neighboring pixels, that is, pixels (i-1, j-1), (i, j-1), (i + 1, j)
−1), (i−1, j), (i, j), (i + 1,
j), (i-1, j + 1), (i, j + 1), (i +
The image data at 1, j + 1) is used. However, −0.5 ≦ Δi <0.5 and −0.5 ≦ Δj <
It is 0.5. The operation of the embodiment of the resolution conversion processing apparatus of the present invention configured as described above will be described below.

First, the image data of the neighboring 9 pixels centered on the pixel of interest (i, j) input from the input terminal 1 is the first
Is input to the interpolation processing unit 2. The first interpolation processing unit 2 selects the image data necessary for the interpolation processing according to the designated interpolation position and performs the interpolation processing on these data. The image data input from the input terminal 1 is also input to the high frequency enhancing section 3 at the same time. Further, the high-frequency emphasizing unit 3 performs high-frequency area emphasizing processing of an image for emphasizing edges such as character images. At the same time as these processes, the image data input from the input terminal 1 is input to the edge determination unit 5,
It is determined whether the pixel of interest exists in the edge area.

Next, the second interpolation processing section 4 is a high-frequency emphasis section 3
The emphasized image data which is the output of is input, and the same interpolation processing is performed on the same interpolation position as the first interpolation processing unit 2.

The output of the first interpolation processing section 2 and the output of the edge determination section 5 are input to the first multiplier 6 and weighted according to the edge determination result. Similarly, in the second multiplier 7, the output of the second interpolation processing unit 4 and the output of the edge determination unit 5 are input and weighted according to the edge determination result. The two determination results output by the edge determination unit 5 are set so that the sums are always equal, and when the pixel of interest exists near the edge, the output to the second multiplier 7 becomes large, and conversely. When the pixel of interest does not exist near the edge, the output of the edge determination unit changes so that the output to the first multiplier 6 becomes large.

Finally, the adder 8 calculates the sum of the output of the first multiplier 6 and the output of the second multiplier 7, and the result is output from the output terminal 9 to complete the processing for one pixel. To do. Repeat the above processing for the number of pixels of the output image to 1
The screen processing is completed.

The outline of the operation of the resolution conversion processing apparatus according to the present invention has been described above. The operations of the first and second interpolation processing units 2 and 4, the high-frequency emphasis unit 3 and the edge determination unit 5 will be described in detail below. explain. 3 and 4 are explanatory diagrams of the interpolation processing used in this embodiment. First interpolation processing unit 2 and second
The interpolation processing unit 4 selects the image data required for the interpolation processing by dividing it into four areas shown in FIG. 3 according to the designated interpolation position. The necessary image data are three data that are the vertices of a triangle surrounding each area. That is, image data as shown in Table 1 below is used for the interpolation process.

[0019]

[Table 1]

In the example of FIG. 3, since the interpolation position is included in the area IV, the image data required for the interpolation processing is (i, j), (i
It is data of 3 pixels of +1, j) and (i, j + 1).
FIG. 4 shows the relationship between the data of these three pixels and the interpolation position. At this time, the interpolation processing is performed according to the following equation. However, I _{m, n} (i−1 ≦ m ≦ i + 1, j−1 ≦ n
≦ j + 1) is an input signal to the interpolation processing unit in the pixel (m, n), and I _{i +} Δ _{i, j +} Δ _j is an output signal. I _{i +} Δ _{i, j +} Δ _j = (1−Δi−Δj) I _{i, j} + ΔiI
_{i + 1, j} + ΔjI _{i, j + 1} Interpolation processing can be similarly performed in other regions.

Next, the high frequency emphasizing section 3 will be described. In this processing unit, a high-frequency region is emphasized in order to enhance the reproducibility of an edge region such as a character image. The input signals of 9 pixels centered on the pixel of interest are I _{m, n} (i−1 ≦ m ≦ i + 1, j−
1 ≦ n ≦ j + 1) and the output signal is J _{m, n} (i−1 ≦ m
≦ i + 1, j−1 ≦ n ≦ j + 1), the calculation in this processing unit is given by the following equation. J _{i-1, j-1} = I _{i-1, j-1} J _{i, j-1} = 3 I _{i, j-1} − (I _{i-1, j-1} + I _{i + 1, j-1} ) J _{i + 1, j-1 =} I i + 1, j-1 J i-1, j = 3I i-1, j - (I i-1, j-1 + I i-1, j + 1) J i _{, j} = 5I _{i, j} − (I _{i-1, j} + I _{i, j-1} + I
_{i + 1, j} + I _{i, j + 1} ) J _{i + 1, j} = 3 I _{i + 1, j} − (I _{i + 1, j-1} + I _{i + 1, j + 1} ) J _{i-1, j + 1} = I _{i-1, j + 1} J _{i, j + 1} = 3 I _{i, j + 1} − (I _{i-1, j + 1} + I _{i + 1, j + 1} ) J _{i + 1, j + 1} = I _{i + 1, j + 1}

As is apparent from the above equation, four pixels ((i-1, j-1), (i + 1,
j-1), (i-1, j + 1), (i + 1, j + 1))
Then, the input signal becomes an output signal as it is, and four pixels ((i, j-
1), (i-1, j), (i + 1, j), (i, j +)
In 1)), the one-dimensional Laplacian filter process shown in FIG. 5 is performed, and the pixel of interest itself is subjected to the two-dimensional Laplacian filter shown in FIG. The signals of these 9 pixels are sent to the second interpolation processing unit 4.

Finally, the edge determination section 5 will be described. The edge determination unit aims to detect a linear edge included in a character image or the like. That is, a filter may be used that has a large edge determination value in a character image and a small determination value in other photographic images or halftone dot printed images. In this embodiment, the filter shown in FIG. 6 is used. In this filter, the filter coefficient of the pixel adjacent to the target pixel in the main scanning direction and the sub-scanning direction is set to 0, the filter coefficient of the pixel diagonally adjacent to the target pixel is set to a negative value -1, and the total sum of the filters is set to 0. The filter coefficient of the pixel of interest is set to. The absolute value of the output of this filter becomes the output W ₁ from the edge determination unit 5 to the second multiplier 7. The output W ₂ from the edge determination unit 5 to the first multiplier 6 is a constant minus the first output value. The above process is expressed by the following formula. However, the signals are normalized so that the total sum is 1. W ₁ = | 4I _{i, j} − (I _{i-1, j-1} + I _{i + 1, j-1} + I _{i-1, j + 1}
+ I _{i + 1, j + 1} ) | W ₂ = 1−W ₁ Therefore, the output value of the first interpolation processing unit 2 is I _{i +} Δ
_{i, j +} Δ _j , the output value of the second interpolation processing unit 4 is J _{i +} Δ _{i, j +} Δ _j
Then, the image data Do output from the output terminal 9
Becomes Do = W ₁ · J _{i +} Δ _{i, j +} Δ _j + W ₂ · I _{i +} Δ _{i, j +} Δ _j .

The operation of the filter shown in FIG. 6 will be further described. This filter is a high-pass filter inclined at 45 °, and has a characteristic that a high frequency component in a diagonal direction of a pixel of interest in the center is attenuated. Therefore, diagonal 45 °
Does not react to the image data of 1-pixel alternating having a power spectrum in the high frequency band, and the filter output value becomes small. Generally, a dot print image has a screen angle inclined at 45 °, and therefore the filter output value for the dot print image can be suppressed to be small for the above reason. Furthermore, since this filter amplifies high-frequency components in the main scanning and sub-scanning directions, its output value becomes large for image data of alternating one line. That is, the weight W1 is large for image data having a linear edge such as a character image, and the weighting for the image subjected to the emphasis processing is large.

Although the above method is used for the interpolation processing in this embodiment, it is easy to use another interpolation method such as the co-linear interpolation method shown in the conventional example. In addition, in the present embodiment, the two interpolation processing units are treated as independent units for ease of explanation, but since the same processes are performed, it is easy to realize them in one.

Further, in this embodiment, the input signal is (3 ×
3) Although the description has been limited to pixels, if a larger range of input is possible, the high-frequency emphasizing unit 3 can perform two-dimensional Laplacian processing on all pixels. Similarly, it is needless to say that since a wider range of filters can be applied to the edge determination unit 5, highly accurate edge determination can be performed.

[0027]

As described above in detail, in the resolution conversion processing apparatus of the present invention, the image data of a plurality of pixels centered on the pixel of interest is used as input data, and the pixel of interest is input from the input image data in the vicinity of a linear edge. Edge determining means for determining whether or not, and first interpolation processing means for interpolating the input image data to calculate a pixel value at an arbitrary pixel position,
A high-frequency emphasis unit that performs a sharpening process on the input image data and an output of the high-frequency emphasis unit are input and interpolated to calculate a pixel value at the same pixel position as the first interpolation unit. By providing second interpolation processing means and means for performing weighting calculation on the results of the first and second interpolation processing using the output of the edge determination means, the first interpolation processing means is provided. The simple interpolation result which is the output of the means and the interpolation result after the sharpening processing which is the output of the second interpolation means are weighted according to the degree to which the pixel of interest exists in the linear edge region. Therefore, it is possible to obtain a clear image that has been subjected to the emphasis processing for a character image and a smooth gradation image that has not been subjected to the emphasis processing for a photographic image or a halftone dot printed image.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of a resolution conversion processing device according to an embodiment of the present invention.

FIG. 2 is a diagram showing a relationship between a pixel position of an input image and a pixel position of an output image.

FIG. 3 is a diagram showing a relationship between a pixel position of an input image and a pixel position of an output image in the interpolation processing.

FIG. 4 is an explanatory diagram of interpolation processing.

FIG. 5 is an example of a high-frequency emphasis filter.

FIG. 6 is an example of a filter used for edge determination.

FIG. 7 is a principle diagram of the nearest neighbor interpolation method.

FIG. 8 is a principle diagram of a bilinear interpolation method.

FIG. 9 is a principle diagram of a cubic convolution interpolation method.

FIG. 10 is an example of a high-frequency emphasis filter.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Input terminal 2 1st interpolation processing part 3 High-frequency emphasis part 4 2nd interpolation processing part 5 Edge determination part 6 1st multiplier 7 2nd multiplier 8 Adder 9 Output terminal

Claims (2)

[Claims] 1. A resolution conversion processing apparatus for interpolating a pixel value at an arbitrary position on an image using a multi-valued image as an input image to determine whether a pixel of interest is near a linear edge from an input image signal. Edge determining means, first interpolation processing means for interpolating the input image signal to calculate a pixel value at an arbitrary pixel position, and high-frequency emphasis processing for performing sharpening processing on the input image signal Means for inputting the output of the high-frequency emphasis processing means, interpolating the output, and calculating a pixel value at the same pixel position as the first interpolation means; And a means for performing weighted addition using the output of the edge determination means with respect to the results of the first and second interpolation processing. 2. The resolution conversion processing device according to claim 1, wherein the edge determination means sets a filter coefficient of a pixel adjacent to the pixel of interest in the main scanning and sub-scanning directions to 0, and filters a pixel diagonally adjacent to the pixel of interest. A resolution conversion processing device using a filter in which a filter coefficient of a target pixel is set so that a coefficient has a negative value and a total sum of filters is 0.