JPH10155128A - Image interpolation device, image interpolation method and video printer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image interpolation device, image interpolation method and video printer in which blurred images are avoided and deformation of characters or the like is eliminated. SOLUTION: The image interpolation device is provided with an edge detection section 33 that detects the local edge direction of an image and with a 2-dimension adaptive filter 34 that selects a coefficient depending on the edge direction detected by the edge detection section 33 and filters a signal at a prescribed frequency band with a prescribed gain based on the frequency characteristic of the selected coefficient. Since specific picture elements of an image are interpolated twice in a direction, tilt edges are smoothly interpolated, occurrence of a blurred image is less and deformation of characters is small. Thus, an interpolated image with high image quality is obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image interpolating apparatus and an image interpolating method suitable for, for example, pixel interpolation processing in a video printer.

[0002]

2. Description of the Related Art In recent years, sublimation type color printers have been improved in definition, but the resolution of input images is mainly the same as that of video signals, and high resolution ones are not widely used.
That is, although high-resolution printing is possible on the printer side, the resolution of the input image data is lower than this. Therefore, when performing high-resolution printing, it is necessary to perform predetermined interpolation processing on the image data inside the printer to simulate high-resolution image data. However, since the image quality of the print output greatly depends on the method of the interpolation processing of the image data, a higher-quality interpolation processing method has been demanded.

In the sublimation type color printer, the resolution can be controlled by changing the gradation in the moving direction of the thermal head, that is, the sub-scanning direction, in an analog manner. Therefore, image interpolation in the direction of the heating elements of the thermal head, that is, in the main scanning direction is required. In other words, there has been a demand for a process of interpolating the image in the main scanning direction by a factor of two.
As such an image interpolation method, first, the same pixel is
A simple enlargement method of arranging pixels one by one and simply enlarging the image by a factor of two, an average value interpolation method in which an average value in the vertical direction is used as an interpolation pixel, and the like are mainly used. As an advanced interpolation process, there is a pattern recognition average value interpolation method in which an average value in a diagonal direction is used as an interpolation pixel depending on a pattern of a magnitude relation between neighboring pixels.

[0004]

However, the conventional image interpolation method has the following problems. That is, in the first simple enlargement method, although the pixels at the edges of the image are preserved, the image is not blurred, but the edges in the oblique direction are jagged, and the quality of the image is degraded. In addition, the second average value interpolation method has a disadvantage that the image is blurred because the average value is used as an interpolation pixel, and a slight jagged edge in an oblique direction remains. In the third pattern recognition average value interpolation method, edges in oblique directions are smoothly interpolated, but the image is blurred because the average value is used as an interpolation pixel. Is obliquely interpolated,
There was an inconvenience that characters were deformed.

An object of the present invention is to provide an image interpolation apparatus, an image interpolation method, and a video printer which can eliminate blurring of an image and deformation of characters and the like, in view of the above point. .

[0006]

According to the present invention, there is provided an image interpolating apparatus comprising: an edge detecting means for detecting a local edge direction of an image; and a coefficient which is selected based on the edge direction detected by the edge detecting means. Filter means for filtering a predetermined frequency band with a predetermined gain based on the frequency characteristic of the coefficient, and interpolating a specific pixel of the image twice in one direction.

Further, according to the image interpolation method of the present invention, a local edge direction of an image is detected, a coefficient is selected in accordance with the detected edge direction, and a predetermined frequency band is determined based on a frequency characteristic of the selected coefficient. On the other hand, a filter is applied with a predetermined gain to interpolate a specific pixel of the image twice in one direction.

A video printer according to the present invention performs a predetermined interpolation process on image data of a video image and prints the image data with a print head using the interpolated image data. Edge detection means for detecting, and a filter means for selecting a coefficient by the edge direction detected by the edge detection means and applying a filter with a predetermined gain to a predetermined frequency band according to a frequency characteristic of the selected coefficient, In addition, a specific pixel of the above image is interpolated twice in one direction and printed.

According to the image interpolation apparatus, the image interpolation method and the video printer of the present invention, the following operations are performed. First, the operation of the edge detecting means will be described. The local maximum tilt direction of the image is detected by the maximum tilt direction detecting means. Then, the direction perpendicular to the maximum inclination direction is detected as the edge direction by the vertical direction detecting means. Hereinafter, an edge detection method based on local tilt information will be described. In general, for a function of a smooth analog image, the local maximum inclination direction can be represented by a spatial first derivative.

The direction perpendicular to the vector represented by the spatial first derivative corresponds to the edge direction. Similarly, for digital image data, the local maximum tilt direction can be generally represented by a vector.

A vector perpendicular to this vector is obtained.
That is, in the image block surrounding the interpolation target pixel, there are a pixel having a value of zero for the image data, a pixel having a medium value of the image data, and a pixel having a maximum value of the image data. At this time, the vector indicates the direction from a pixel having an image data value of zero and an intermediate value to a pixel having a maximum value. The direction of the vector perpendicular to this vector corresponds to the edge direction.

The edge direction of the image is divided into eight areas including at least a horizontal direction, a vertical direction, and an oblique direction by the edge direction area specifying means based on the edge direction inclination value thus obtained.

Next, when the oblique direction is detected by the edge detecting means, a right angle corner pattern such as a character is detected by the right angle corner pattern detecting means.
In the image block surrounding the pixel to be interpolated, there are a pixel having a value of image data of zero, a pixel having a medium value of the image data, and a pixel having a maximum value of the image data. At this time, the vector indicates the direction from a pixel having an image data value of zero and an intermediate value to a pixel having a maximum value. Then, a vector perpendicular to this vector is detected as a vector in a direction determined by the influence of surrounding zero pixels, regardless of the direction of a pixel having an intermediate value including the pixel to be interpolated. Then, the direction of this vector corresponds to the edge direction.

However, since the right-angled corner pattern is included not in a natural image but in an artificial image such as a character, if the direction of the vector is detected as an edge direction,
The shape of the character is distorted. Therefore, in order to detect this right-angled corner pattern, the inclination of peripheral pixels is calculated.

At this time, if the absolute value of the inclination of the peripheral pixels is small and has the same sign, it is regarded as a right-angle corner pattern and the horizontal direction is selected. The deformation processing of the artificial image such as a character can be suppressed by the detection processing of the right-angle corner pattern. Thus, the local edge direction of the image is detected.

Next, the operation of the filter means will be described.
The edge direction corresponding coefficient selection means of the two-dimensional adaptive filter switches the coefficient according to the edge direction detected by the edge detection means. These filter coefficients are directional and constitute a two-dimensional low-pass filter, which cuts jagged components in the edge direction and preserves the sharpness of edges in the direction perpendicular to this. Therefore, the edges in the oblique direction are smoothly interpolated without being jagged.

Further, since a low-pass filter having better characteristics can be formed as compared with the method using the average value of pixels, blurring of an image can be suppressed. Furthermore, since these filters are applied to a simple double-magnified image, horizontal edges are preserved without blurring as compared with a method using an average value of pixels. In this way, a high-quality interpolated image can be obtained in which oblique edges are smoothly interpolated, the image is less blurred, and characters and the like are less deformed.

[0018]

Embodiments of the present invention will be described below. The image interpolating apparatus according to the present embodiment detects a local edge direction after simply enlarging the image to a double with respect to the digital image, and cuts the jagged frequency component in that direction. By performing image interpolation using a two-dimensional adaptive digital filter having a characteristic, diagonal edges are smoothly interpolated to reduce blurring of the image and obtain a high-quality interpolated image with less deformation of characters and the like. Is what you can do.

First, a color video printer to which the image interpolation device of the present embodiment is applied will be described. The color video printer includes a video input unit 1 for inputting a video signal, and an A / A converter for quantizing an analog video signal input from the video input unit 1 and converting the analog video signal into a digital signal.
A D converter 2, an image memory 4 as a frame memory for storing and holding digital video signal data of one frame of the digital video signal, and a memory control for controlling writing and reading of the digital video signal to and from the image memory 4 Unit 3 and switch unit 5 for performing various settings
A system control unit 6 that controls the operation of the memory control unit based on the setting in the switch unit 5, a mechanism control unit 7 that controls the operation of the mechanical mechanism under the control of the system control unit 6, and a mechanism control unit 7. A mechanical mechanism 8 such as a photographic paper feed motor to be operated, and digital video signal data for each line are received from the image memory 4 in synchronization with the sending of photographic paper by the photographic paper feed motor of the mechanical mechanism 8, and various image processing is performed. It has an image processing section 9 and a thermal head 11 for printing digital video signal data obtained by performing image processing on printing paper.

Here, in this embodiment, in particular, in the image processing section 9, after the image is simply doubled with respect to the digital image, the local edge direction is detected, and the local edge direction is detected. Non-linear interpolation block 1 that performs smooth image interpolation on diagonal edges by performing image interpolation using a two-dimensional adaptive digital filter having directionality that cuts jagged frequency components.
Has zero.

The color video printer thus configured operates as follows. Video input unit 1 to A
The video signal is supplied to the A / D converter 2 and the A / D converter 2
In, the analog video signal supplied from the video input unit 1 is quantized and converted into a digital signal. A
The digital video signal converted by the / D converter 2 is supplied to the memory control unit 3. The writing operation is performed so that the data of the digital video signal for one frame is stored in the image memory 4 from the memory control unit 3.

Data writing or reading operation from the system control unit 6 to the memory control unit is controlled based on various settings set in the switch unit 5.
When an operation command for a mechanical mechanism is supplied from the system control unit 6 to the mechanism control unit 7, the mechanism control unit 7 operates the mechanical mechanism 8 such as a printing paper feed motor. Therefore, an operation command for reading data from the image memory 4 is supplied from the system control unit 6 to the memory control unit.
Then, the digital video signal data for each line is read out from the image memory 4 in synchronization with the feeding of the printing paper by the printing paper feed motor of the mechanical mechanism 8 and supplied to the image processing unit 9. The image processing unit 9 performs various image processing on the digital video signal data for each line. The digital video signal data on which the image processing has been performed is supplied to the thermal head 11. In the thermal head 11, digital video signal data is printed on printing paper with a heat value corresponding to the gradation of the image data.

Here, in this embodiment, in particular, in the non-linear interpolation block 10 of the image processing section 9, after simply enlarging the image to twice the digital image,
By detecting the local edge direction and performing image interpolation using a two-dimensional adaptive digital filter with directionality that cuts jagged frequency components in that direction, the diagonal edges can be smoothly interpolated. In addition, image interpolation is performed so as to reduce blur of the image and obtain a high-quality interpolated image with less deformation of characters and the like.

Next, the image processing section will be described with reference to FIG. The image processing unit shown in FIG. 2 corresponds to the image processing unit 9 shown in FIG. The image processing unit includes a register 12 for storing the control data Dc from the system control unit 6 shown in FIG. 1, and an operation of the image processing unit in the main scanning direction based on the control data Dc and the data from the register 12. And a sub-scanning direction controller 14 for controlling the operation of the image processing unit in the sub-scanning direction based on the control data Dc and the data from the register 12.

The image processing unit is a line buffer for temporarily holding one line of input image data Di supplied from the image memory via the memory control unit 3 shown in FIG. 15a, 15b, 15c; a sub-scanning direction interpolation block 16 for interpolating in a sub-scanning direction the input image data Di for each line temporarily stored in the line buffers 15a, 15b, 15c; And a linear interpolation block 17 for performing Here, in this embodiment, in particular, in each of the field data and the frame data of the image data subjected to the interpolation in the sub-scanning direction, the image is simply doubled, and then the local edge direction is changed. Field / frame interpolation that performs smooth image interpolation of diagonal edges by performing image interpolation using a two-dimensional adaptive digital filter having directionality that cuts jagged frequency components in that direction. Block 10 (FIG. 1)
Corresponds to the non-linear interpolation block 10 shown in FIG. ).

The image processing unit includes a resize block 18 for converting the number of scan lines of the PAL system into a number of scan lines of the NTSC system, and one line of image data whose number of scan lines is converted by the resize block. A line buffer 19 for temporarily holding, an editing function block 20 for adding various editing functions to the image data, a line buffer 21 for temporarily holding one line of image data for sharpness correction, an emphasis process , A thermal storage correction block 25 for performing thermal storage correction in a thermal head, a thermal storage correction memory 23 for storing thermal storage correction data for a short time constant response system, and a thermal storage correction data for a long time constant response system. A heat storage correction memory 24 for storing;
A dither block 26 for adding noise (dither) corresponding to shading dropped at a fixed threshold value when the input signal is binarized to the original signal.

The image processing unit configured as described above operates as follows. The control data Dc from the system control unit 6 shown in FIG. 1 is supplied to the register 12, the main scanning direction controller 13, and the sub-scanning direction controller 14. The control data Dc stored in the register 12 is supplied to the main scanning direction controller 13 and the sub-scanning direction controller 14. In the main scanning direction controller 13, the control data Dc and the register 1
2, the operation of the image processing section in the main scanning direction is controlled, and the sub-scanning direction controller 14 controls the operation of the image processing section in the sub-scanning direction based on the control data Dc and the data from the register 12. Is done. More specifically, from the main scanning direction controller 13 and the sub-scanning direction controller 14, line buffers 15a, 15b, 15c for image interpolation processing, a line buffer 19 for resizing and editing functions, a line buffer 21 for sharpness, a heat storage correction Storage correction memory 23, 24 for
Are supplied with a control signal.

Further, the line buffers 15a, 15b and 15 are transferred from the image memory via the memory control unit 3 shown in FIG.
The input image data Di for one line is sequentially supplied to c. That is, the line buffer 15a temporarily stores the image data Di + 1 of the next line, the line buffer 15b temporarily stores the image data Di of the current line, and the line buffer 15c stores the image data Di of the previous line. Data Di-1 is temporarily stored. Line buffer 1
The input image data Di + 1, Di, Di-1 for each line temporarily stored in 5a, 15b, 15c are supplied to the sub-scanning direction interpolation block 16. In the sub-scanning direction interpolation block 16, input image data Di + 1, Di, D
i-1 is interpolated in the sub-scanning direction. For example, in the sub-scanning direction interpolation block 16, the field data is interpolated, the image data is rotated by 90 degrees, and
Interpolation in the sub-scanning direction is performed so as to obtain a print output of one sheet. The image data interpolated in the sub-scanning direction or the image data passed through without interpolating in the sub-scanning direction is supplied to the field / frame interpolation block 10. When sub-scanning interpolation is performed, the field / frame interpolation block 10 at the subsequent stage
Processing is divided so that frame interpolation is not performed and field / frame interpolation is performed only when sub-scanning interpolation is not performed and a through operation is performed.

Here, in the present embodiment, in particular, in the field / frame interpolation block 10, an image is formed with respect to each of the field data and frame data of the image data passed through without performing interpolation in the sub-scanning direction. After simply enlarging the image by a factor of two, the local edge direction is detected, and image interpolation is performed by a two-dimensional adaptive digital filter having directionality that cuts jagged frequency components in that direction. Image interpolation for smoothing the edge in the direction is performed. Specifically, in the field / frame interpolation block 10, field data or frame data is interpolated twice. The field / frame interpolation block 10 is referred to as a non-linear interpolation block here because an adaptive filter, which will be described later, is a (generally) nonlinear filter. Image data interpolated twice by the field data and the frame data, respectively, is supplied to the linear interpolation block 17. Linear interpolation block 1
At 7, the frame data is passed through and interpolation is performed twice for only the field data. In other words, in each of the non-linear interpolation and the linear interpolation, the frame data is interpolated twice in total by double interpolation and through, and the field data is interpolated in total four times by double interpolation and double interpolation. Here, the filter is called a linear interpolation block 17 because the filter used here is a linear filter.

The image data thus interpolated is supplied to the resize block 18. In the resize block 18, for example, the number of scanning lines of the PAL system is converted into the number of scanning lines of the NTSC system. The image data whose number of scanning lines has been converted is supplied to the line buffer 19. Here, from the input of the input image data Di to the point at which the image data of the current line is supplied to the line buffer 19, it takes time to write and read the image data for one line. In other words, the processing of writing to the line buffer 15, the sub-scanning interpolation block 16, the field / frame interpolation block 10, the linear interpolation block 17, the resizing block 18, and the line buffer 19 is performed in synchronization.

Then, one line of the image data whose number of scanning lines has been converted by the resize block is read from the line buffer 19 and supplied to the editing function block 20. In the editing function block 20, various editing functions are added to the image data. Image data to which various editing functions have been added is supplied to the sharpness block 22. An emphasis process is performed in the sharpness block 22. At this time, one line of image data for sharpness correction is temporarily held in the line buffer 21. The image data that has been subjected to the sharpness correction is supplied to the heat storage correction block 25.

In the heat storage correction block 25, the heat storage in the thermal head is corrected. At this time, the heat storage correction data of the response system of the short time constant is supplied from the heat storage correction memory 23 to the heat storage correction block 25, and the heat storage correction memory 24
The heat storage correction data of the response system having a long time constant is supplied to the heat storage correction block 25, and the two response systems are corrected by different processes.

The heat storage corrected image data is supplied to a dither block 26. In the dither block 26, noise (dither) corresponding to the shading dropped at a fixed threshold value when the input signal is binarized is added to the original signal. Thus, the output image data Do is output from the dither block 26.

Here, the image data is stored in the line buffer 19.
From the time when the output image data Do is output from the dither block 26, it takes time to write and read one line of image data. That is, reading from the line buffer 19, processing of the editing function block 20, the sharpness block 22, the heat storage correction block 25, and the dither block 26 are performed in synchronization.

Next, a specific configuration and operation of the image interpolation apparatus according to the present embodiment will be described. The field /
2 shows a configuration of a frame interpolation block 10. The configuration shown in FIG. 3 corresponds to the field / frame interpolation block 1 shown in FIG.
Corresponds to 0. In FIG. 3, a field / frame interpolation block 10 includes a simple enlargement unit 30 that simply enlarges an image by a factor of two, an edge detection unit 33 that detects a local edge direction of the image, and a jagged frequency in the edge detection direction. It has a two-dimensional adaptive filter 34 having a direction to cut components. Here, G _{i, j and} G ′ _{i, j} represent digital image data, i represents a sub-scanning direction (line increment), and j represents a main scanning direction (line direction).

The simple enlargement section 30 has a double upsampling circuit 31 and a zero-order hold circuit 32. 2
The double up-sampling circuit 31 includes a double up-sampling circuit 31a that samples the digital image data G _{i-1, j} at a double sampling frequency, and a sampler 2 that samples the digital image data G _{i, j} at a double sampling frequency. It has a double upsampling circuit 31b and a double upsampling circuit 31c for sampling the digital image data G _{i + 1, j} at a double sampling frequency. The zero-order hold circuit 32 adds the data before one sampling point to the digital image data G _{i-1, j} sampled at the double sampling frequency, and the zero-order hold circuit 32a at the double sampling frequency. Sampled digital image data G
Zero-order hold circuit 32b for adding data before one sampling point to _{i, j,} and zero-order hold circuit for adding data before one sampling point to digital image data G _{i + 1, j} sampled at twice the sampling frequency Circuit 32c.

FIG. 4 shows functional blocks of the edge detecting section. The functional block shown in FIG. 4 corresponds to the edge detection unit 33 shown in FIG. The edge detection unit includes a maximum inclination detection unit 40 that detects a local maximum inclination of an image, a vertical direction detection unit 41 that detects a direction perpendicular to the maximum inclination as an edge direction, and a detection unit that detects a detected edge direction. An edge direction area designating unit 42 that divides at least into predetermined areas in the horizontal direction 43, the vertical direction 44, and the oblique direction 45 and designates an edge direction area; Having.

FIG. 5 shows functional blocks of the two-dimensional adaptive filter. The functional block shown in FIG. 5 corresponds to the two-dimensional adaptive filter 34 shown in FIG. This two-dimensional adaptive filter has an edge direction corresponding coefficient selecting means 50 and coefficients 51 (H _{1 to} H ₈ ) having frequency characteristics having directionality.

The operation of the field / frame interpolation block 10 configured as described above will be described below. First, the operation of the simple enlargement unit 30 will be described. In the double upsampling circuit 31a, the digital image data G _{i-1, j}
Is sampled at a double sampling frequency, the digital image data G _{i, j} is sampled at a double sampling frequency in a double up-sampling circuit 31b, and the digital image data G _{i + 1, j} is sampled at a double up-sampling circuit 31c. Are sampled at twice the sampling frequency. Then, the values at the time of the original sampling and the value of 0 (zero) at the time of the double sampling are output from the double up-sampling circuits 31a, 31b, 31c, respectively.

Double upsampling circuits 31a, 31
b, 31c, the values at the time of the original sampling,
The value of 0 (zero) at the time of double sampling is supplied to the zero-order hold circuits 32a, 32b, and 32c. In the zero-order hold circuits 32a, 32b, and 32c, the value at the time of the original sampling and the value at the time of the double sampling one time before the sampling are added. In this way, image data that is doubled in the main scanning direction from the simple enlargement unit 30 is obtained.

Next, the operation of the edge detector will be described. First, the local maximum tilt direction of the image is detected by the maximum tilt direction detecting means 40. And the vertical direction detecting means 4
By means of 1, a direction perpendicular to the maximum inclination direction is detected as an edge direction. Hereinafter, an edge detection method based on local tilt information will be described. In general, when a smooth analog image is represented by a function f (x, y), the local maximum inclination direction can be represented by a vector represented by the following formula (1) by first-order spatial differentiation.

[0042]

Grad f (x, y) = (∂f (x, y) /
∂x, ∂f (x, y) / ∂y)

The direction perpendicular to the vector represented by Equation 1 corresponds to the edge direction. Similarly, if the digital image data is set to G _{i, j} , the local maximum inclination direction can be generally represented by a vector (p _{i, j} , q _{i, j} ) of the following equation (2). .

[0044]

(P _{i, j} , q _{i, j} ) = (G _{i + 1, j} −G _{i, j} , G
_{i, j + 1} -G _{i, j} )

When the calculation is performed in a large area in order to match the phases, it can be represented by the vector of Expression 3.

[0046]

## EQU3 ## (p _{i, j} , q _{i, j} ) = ((G _{i + 1, j + 1} + G
_{i + 1, j} + Gi _{+ 1, j-1} )-(Gi _{-1, j + 1} + Gi _{-1, j} + G
_{i-1, j-1} ), (Gi _{+ 1, j + 1} + Gi _{, j + 1} + Gi _{-1, j + 1} )-
(G _{i + 1, j-1} + G _{i, j-1} + G _{i-1, j-1} ))

A vector perpendicular to the vector (p _{i, j} , q _{i, j} ) is obtained as shown in FIG. That is, in the 3 × 3 image block surrounding the pixel to be interpolated, a pixel indicated by a narrow diagonal line has an image data value of zero, a pixel indicated by a wide diagonal line has an image data value of an intermediate value, and a blank pixel has The value of the image data is the maximum value. At this time, the vector (p _{i, j} , q _{i, j} ) indicates the direction from the pixel of the intermediate value to the pixel of the maximum value where the value of the image data is zero. in this case,
In FIG. 6, the lower left three pixels are zero, the three pixels including the pixel to be interpolated from diagonally upper left to the lower right are intermediate values, and the three upper right pixels are maximum values. The vector of the target pixel is in a diagonally upper right direction. And
A vector perpendicular to this vector is a vector (−q _{i, j} , p _{i, j} ) or a vector (q), which is the direction of three pixels of intermediate values including the pixel to be interpolated to the upper left or lower right. _{i, j} , -pi _{, j} ). Then, the vector (−q _{i, j} , p _{i, j} ) or the vector (q _{i, j} , −p
The direction of _{i, j} ) corresponds to the edge direction.

Next, when the image is simply doubled in the j main scanning direction (line direction) by the simple enlarging unit 30, the vector (p _{i, j} , q _{i, j} ) becomes the vector (2p _{i, j).} , Q
_{i, j} ), and a vector perpendicular to the vector (2p _{i, j} , q _{i, j} ) is obtained as shown in FIG. That is,
j In a 3 × 6 image block surrounding a pixel to be interpolated doubled in the main scanning direction, a pixel indicated by a narrow diagonal line has an image data value of zero, and a pixel indicated by a wide diagonal line has an image data value of an intermediate value. In the blank pixels, the value of the image data is the maximum value. At this time, each pixel shown in FIG. 6 is doubled in the j main scanning direction. Vector (2p _{i, j} , q
_{i, j} ) indicates the direction from the pixel having the image data value of zero, the intermediate value to the pixel having the maximum value. In this case, in FIG. 7, the six pixels at the lower left are zero, the six pixels including the interpolation target pixel doubled in the j main scanning direction from the upper left to the lower right are intermediate values, and Since the six pixels have the maximum value, the vector of the pixel to be interpolated is in a diagonally upper right direction.
The vector perpendicular to this vector is the vector (−q _i ,, which is the direction of six pixels of the intermediate value including the interpolation target pixel doubled in the j main scanning direction to the upper left and lower right _{. j} , 2pi _{, j} ) or vector (qi _{, j} , -2)
p _{i, j} ). Then, the vector (−q _{i, j} , 2)
The direction of (pi _{, j} ) or the vector (qi _{, j} , -2pi _{, j} ) corresponds to the edge direction. When this vector is represented by the gradient a _{i, j} of a straight line in this direction, Equation 4 is obtained.

[0049]

## EQU4 ## a _{i, j} = tan θ _{i, j} = -2 p _{i, j} / q _{i, j} The value of a _{i, j} can represent the local edge direction of the image.

Next, details of such edge direction detection will be described. If the image data G _{i, j} obtained by enlarging the image data G _{i, j} by a factor of 2 in the j main scanning direction is g _{i, k} , Equation 5 is obtained.

[0051]

G _{i, k} = G _{i, k / 2} (k: even) = G _{i, (k−1) / 2} (k: odd)

The spatial first derivative (p _{i, j} , q
_{i, j} ) is equivalent to the image data g _{i, k}
And (r _{i, k} , s _{i, k} ) can be obtained by equation (6).

[0053]

(R _{i, k} , s _{i, k} ) = (max [(g _{i + 1, k} + g _{i + 1, k−2−} g _{i, k} −g _{i, k−2} ) / 2, (Gi _{, k} + gi _{, k-2} -gi _{-1, k-} gi _{-1, k-2} ) / 2], (gi _{+ 1, k} + gi _{, k} + gi _{-1, k-} g _{i + 1, k−2} −g _{i, k−2} −g _{i−1, k−2} ) / 3) (k: even) (max [(g _{i + 1, k} + g _{i + 1, k + 2−} g _{i, k} −g _{i, k + 2} ) / 2, (g _{i, k} + g _{i, k + 2} −g _{i−1, k} −g _{i−1, k + 2} ) / 2], ( gi _{+ 1, k + 2} + gi _{, k + 2} + gi _{-1, k + 2-} gi _{+ 1, k-} gi _{, k-} gi _{-1, k} ) / 3) (k: odd)

Here, max [*, *] indicates the one with the larger absolute value. Thus, it can be seen that (ri _{, k} , si _{, k} ) takes into account the phase in the double-magnified image.

Further, the second derivative in the i and k directions is calculated. If the absolute value is small, it is determined that the curved surface of the image is small, and the equation (7) is adopted.

[0056]

(7) (ri _{, k} , si _{, k} ) = ((gi _{+ 1, k} + gi _{+ 1, k-2} -gi _{-1, k-} gi _{-1, k-2} ) / 2, (g _{i + 1, k + 2} + g _{i, k + 2} + g _{i-1, k + 2} -g _{i + 1, k-2} -g _{i, k-2} -g _{i-1, k-2} ) / 3) (k: even) ((gi _{+ 1, k} + gi _{+ 1, k-2} -gi _{-1, k-} gi _{-1, k-2} ) / 2, (gi _{+ 1) , k + 2} + gi _{, k + 2} + gi _{-1, k + 2} -gi _{+ 1, k-2} -gi _{, k-2} -gi _{-1, k-2} ) / 3) (k: odd)

As described above, when the secondary differential value in the i and k directions is small, erroneous detection can be prevented by calculating the inclination vector in a large area. Vector (r
_{i, k} , s _{i, k} ), the amount corresponding to the gradient a _{i, j} that previously expressed the local edge direction of the image by Equation 4 is the gradient b
It is expressed by Expression 8 as _{i, k} .

[0058]

## EQU8 ## b _{i, k} = −2r _{i, k} / s _{i, k}

According to the values of the inclinations b _{i, k} obtained in this way, the edge direction of the image is specified by the edge direction area specifying means 42 as shown in FIG. 8 including at least the horizontal direction 43, the vertical direction 44, and the oblique direction 45. Into eight regions. Since the same applies to the case where the edge direction of the image is divided by the value of the inclination a _{i, j} obtained earlier, FIG. 8 shows an example in which the area is divided into eight. The region is when the value of the gradient a _{i, j} is positive and is not more than ／. The area is slope a
_{This is when} the value of _{i, j} is larger than 3/4 and equal to or smaller than 3/2. In the area _, the value of the slope a _{i, j} is larger than 3/2 and 4
It is the following time. The region is when the absolute value of the slope a _{i, j} is greater than 4. The region is when the value of the gradient a _{i, j} is negative and is −3/4 or more. The area is when the value of the gradient a _{i, j} is smaller than −3/4 and is −3/2 or more.
In the region _, the value of the slope a _{i, j} is smaller than −3/2 and
That is the time. The region is when the absolute value of p _{i, j} is equal to or less than a constant. That is, the region is the horizontal direction 43, the region is the vertical direction 44, and the region, the region, the region, the region, the region, and the region are the oblique directions 45.

Next, in FIG. 8, when the edge direction designating unit 42 detects the oblique direction 45 such as a region, a region, a region, a region, a region, and a region, the right-angle corner pattern detecting unit 46 returns to FIG. The detection of a right-angle corner pattern such as a character to be shown is performed. In FIG. 9, in a 3 × 3 image block surrounding a pixel to be interpolated, a pixel indicated by a narrow diagonal line has an image data value of zero, a pixel indicated by a wide diagonal line has an image data value of an intermediate value, and is blank. The pixel has the maximum value of the image data. At this time, the vector (p _{i, j} , q _{i, j} ) indicates the direction from the pixel of the intermediate value to the pixel of the maximum value where the value of the image data is zero.
In this case, in FIG. 9, the lower left two pixels are zero, the two pixels including the interpolation target pixel are intermediate values, and the remaining five pixels extending from top to right are maximum values.
The vector of the pixel to be interpolated is detected as a diagonally upper right direction. The vector perpendicular to this vector is a vector (−q _{i, j} , p _{i, j} ) or a vector (q _{i, j} ,
-Pi _{, j} ). Then, the vector (−q _{i, j} , p _{i, j} ) or the vector (q _{i, j} , −p
The direction of _{i, j} ) corresponds to the edge direction.

However, since the right-angled corner pattern as shown in FIG. 9 is not included in a natural image but is often included in an artificial image such as a character, as shown in FIG.
_{If the direction of i, j} , p _{i, j} ) or the vector (q _{i, j} , −p _{i, j} ) is detected as the edge direction, the shape of the character is distorted. Therefore, in order to detect the right-angled corner pattern, the inclination of the peripheral pixels is calculated as in equation (9).
This is a case where the region, the region, and the oblique direction 45 of the region are selected in FIG. 8 with k: odd number (odd).

[0062]

Dm _{i, j} = g _{i + 2, k} −g _{i + 2, k−2} d _{n i, k} = g _{i, k + 2} −g _{i−2, k−2} dh1 _{i, k} = g _{i, k} −g _{i + 2, k} dh2 _{i, k} = g _{i, k−2} −g _{i + 2, k−2}

At this time, if the absolute values of dm _{i, k} and dn _{i, k} are small and dh1 _{i, k} and dh2 _{i, k} have the same sign,
Considering a right angle corner pattern, in FIG. 8, the horizontal direction 43 of the area is selected. In other cases, similar processing is performed, and deformation of artificial images such as characters can be suppressed by the detection processing of the right-angle corner pattern. Thus, the local edge direction of the image is detected.

Next, the operation of the two-dimensional adaptive filter will be described. In accordance with the edge direction of the area shown in FIG. 8 detected by the edge detection unit, the edge direction corresponding coefficient selection means 50 of the two-dimensional adaptive filter converts the coefficient 51 into H ₁ (Z _i) shown in FIG. 10 and FIG. , Z _j ) -H
₈ Switch to (Z _i , Z _j ). In FIG. 10, H ₁
(Z _i , Z _j ) is 0, 0, 0, +5/16, +5/1 in the j main scanning direction with respect to the first line in the i sub scanning direction.
6, -3/32, -1/32, 0,0,0,0,0,0,0 in the j main scanning direction for the second line, and j for the third line. -1/32 in the main scanning direction,
−3/32, +5/16, +5/16, 0, 0, 0. H ₂ (Z _i , Z _j ) is 0, 0, 0, 0, +5/1 in the j main scanning direction with respect to the first line in the i sub scanning direction.
6, −3/32, −1/32, and 0, 0, 0, +5/8, 0, 0, 0 in the j main scanning direction with respect to the second line.
And -−１ in the main scanning direction with respect to the third line.
2, −3/32, +5/16, 0, 0, 0, 0.
H ₃ (Z _i , Z _j ) is 0, 0, 0, 0, +5/32, −5 in the j main scanning direction with respect to the first line in the i sub scanning direction.
3/32, -1/32, and for the second line, j
0, 0, +5/32, +5/8, +5/3 in the main scanning direction
2, 0, 0, and -1/32, -3/32, +5/32, 0, 0, 0,
0. H ₄ (Z _i , Z _j ) is the first in the i sub-scanning direction.
With respect to the line, 0, 0, 0, 0, 0,
0, 0, and −
1/32, -3/32, +5/16, +5/8, + 5 /
16, −3/32, −1/32, and 0,0,0,0,0,0,0 in the j main scanning direction with respect to the third line.

In FIG. 11, H ₅ (Z _i ,
Z _j ) is -1/32, -3/32, +5/16, + 5 / in the j main scanning direction with respect to the first line in the i sub scanning direction.
16, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 in the j main scanning direction with respect to the second line, and j in the j main scanning direction with respect to the third line. 0,0,0, + 5/1
6, +5/16, -3/32 and -1/32. H ₆
(Z _i , Z _j ) is -1/32, -3/32, +5/1 in the j main scanning direction with respect to the first line in the i sub scanning direction.
6,0,0,0,0, and for the second line, j
0, 0, 0, +5/8, 0, 0, 0 in the main scanning direction, and 0, 0, 0, 0 in the j main scanning direction with respect to the third line.
0, +5/16, -3/32, -1/32. H ₇
(Z _i , Z _j ) is -1/32, -3/32, +5/3 in the j main scanning direction with respect to the first line in the i sub scanning direction.
2,0,0,0,0, and for the second line, j
0, 0, +5/32, +5/8, +5/3 in the main scanning direction
2, 0, 0, and 0, 0, 0, 0, +5/32, −3/32, − ／ in the j main scanning direction with respect to the third line.
2. H ₈ (Z _i , Z _j ) is the first in the i sub-scanning direction.
0, 0, 0, +5/3 in the j main scanning direction with respect to the line
2, 0, 0, 0, and -1/32, -3/32, +5/32, + 5 /
8, +5/32, -3/32, -1/32.
0, 0, 0, +5/3 in the j main scanning direction with respect to the line
2,0,0,0.

Each filter operation by these coefficients is j
Since it is performed in the main scanning direction, an image memory of about three lines may be provided inside. In addition, since the filter coefficient is a multiple of two, hardware can be implemented relatively inexpensively.

The coefficient H of such a two-dimensional adaptive filter
_{1 (Z i, Z j)} ~H 8 (Z i, Z j) the frequency characteristic is shown in FIGS. 12 to 16. This frequency characteristic is represented by X
The gain for two dimensions of frequency and Y frequency is shown.
FIG. 12 shows the frequency characteristics of H ₅ (Z _i , Z _j ). Here, the frequency characteristic of H ₁ (Z _i , Z _j ) is obtained by inverting left and right at X = 0. FIG. 13 shows the frequency characteristics of H ₆ (Z _i , Z _j ). Here, the frequency characteristic of H ₂ (Z _i , Z _j ) is obtained by inverting left and right at X = 0. H in FIG.
₇ shows the frequency characteristic of (Z _i , Z _j ). Where H
The frequency characteristic of ₃ (Z _i , Z _j ) is obtained by inverting left and right at X = 0. Thus, H ₁ (Z _i , Z _j ), H
The frequency characteristics of ₂ (Z _i , Z _j ) and H ₃ (Z _i , Z _j ) are H ₅ (Z _i , Z _j ) and H ₆ (Z _i ,
The frequency characteristics of Z _j ) and H ₇ (Z _i , Z _j ) are inverted in the i sub-scanning direction. FIG. 15 shows H ₄ (Z _i , Z _j )
FIG. FIG. 16 shows the frequency characteristics of H ₈ (Z _i , Z _j ).

These filter coefficients have directionality, and constitute a two-dimensional low-pass filter, thereby cutting jagged components in the edge direction and preserving the sharpness of the edge in the direction perpendicular to the edge. With such characteristics, oblique edges are smoothly interpolated without being jagged.

Further, since a low-pass filter having better characteristics can be configured as compared with the method using the average value of pixels, blurring of an image can be suppressed. Furthermore, since these filters are applied to a simple double-magnified image, there is an advantage that horizontal edges are preserved without blurring as compared with a method using an average value of pixels.

As described above, according to the present embodiment, it is possible to obtain a high-quality interpolated image in which oblique edges are smoothly interpolated, the blur of the image is small, and the deformation of characters and the like is small.

The present invention is not limited to the above example, and may be constituted by printing a low-resolution image at a high resolution in a digital color printer, or by software for increasing the image resolution on a computer.

Also, after interpolating to twice as many pixels using this embodiment, resolution conversion may be performed by the free-magnification resolution conversion means. In this case, the free-magnification resolution conversion means may be used alone. An output image with higher image quality can be obtained as compared with the case where the resolution conversion is performed.

[0073]

According to the image interpolation apparatus of the present invention, an edge detecting means for detecting a local edge direction of an image, and a coefficient selected based on the edge direction detected by the edge detecting means are selected. Filter means for filtering a predetermined frequency band with a predetermined gain according to a frequency characteristic of the image, and a specific pixel of the image is interpolated twice in one direction, so that a jagged component in an edge direction is cut. By preserving the steepness of the edge in the direction perpendicular to the direction, the oblique edge is smoothly interpolated, the blur of the image is reduced, and the deformation of characters and the like is reduced, so that a high-quality interpolated image can be obtained. To play.

In the image interpolation apparatus of the present invention, the edge detecting means is detected by the maximum tilt direction detecting means for detecting a local maximum tilt direction of the image and the maximum tilt direction detecting means. A vertical direction detecting means for detecting a direction perpendicular to the maximum inclination direction as an edge direction; and an edge direction for dividing the edge direction detected by the vertical direction detecting means into a plurality of regions and specifying a region to which the edge direction belongs Since the image processing apparatus includes the area designating means, an effect is provided that a local edge direction of an image can be detected and a filter can be applied by changing a coefficient according to the edge direction.

Further, in the image interpolation apparatus of the present invention, the filter means is a two-dimensional digital filter, and comprises a coefficient having a directional frequency characteristic, and the edge detection means among the coefficients. Since there is provided an edge-direction-corresponding coefficient selecting means for selecting a coefficient corresponding to the detected edge direction, there is an effect that a filter having a different frequency characteristic can be applied by changing the coefficient in accordance with the edge direction.

In the image interpolation apparatus according to the present invention, the edge detecting means includes a maximum tilt direction detecting means for detecting a local maximum tilt direction of the image, and a maximum tilt direction detected by the maximum tilt detecting means. Vertical direction detecting means for detecting the vertical direction as an edge direction, and an edge direction area specifying means for dividing the edge direction detected by the vertical direction detecting means into a plurality of areas, and specifying an area to which the edge direction belongs, In the case where the edge direction is detected by the edge direction area designating means to be an oblique direction,
When detecting that the pixel to be interpolated is a right-angled corner in the image, the image processing apparatus has a right-angled corner pattern detecting means for horizontally selecting an edge direction by the edge direction area specifying means. This has an effect that deformation of an image such as a character can be suppressed.

Further, the image interpolation method of the present invention detects a local edge direction of an image, selects a coefficient in accordance with the detected edge direction, and sets a coefficient in a predetermined frequency band based on a frequency characteristic of the selected coefficient. On the other hand, a filter is applied with a predetermined gain to interpolate a specific pixel of the above image twice in one direction, so the jagged component in the edge direction is cut and the steepness of the edge in the direction perpendicular to this is preserved. By doing so, it is possible to obtain an effect that a high-quality interpolated image can be obtained in which oblique edges are smoothly interpolated, the image is less blurred, and characters and the like are less deformed.

In the image interpolation method according to the present invention, the edge detection includes detecting a local maximum inclination direction of the image, and detecting a direction perpendicular to the detected maximum inclination direction as an edge direction. Since the detected edge direction is divided into a plurality of regions and the region to which the edge direction belongs is specified, by detecting the local edge direction of the image, the coefficient is determined in accordance with the edge direction. It has the effect of being able to apply a filter by changing it.

Further, in the image interpolation method of the present invention, in the above, the filter is a two-dimensional digital filter, and the filter has a directional frequency characteristic in an edge direction detected by the edge detection. Since the corresponding coefficient is selected, there is an effect that a filter having different frequency characteristics can be applied by changing the coefficient according to the edge direction.

In the image interpolation method according to the present invention, the edge detection detects a local maximum tilt direction of the image, and sets a direction perpendicular to the detected maximum tilt direction as an edge direction. Detecting and dividing the detected edge direction into a plurality of regions having at least a horizontal direction, a vertical direction, and a diagonal direction, specifying a region to which the edge direction belongs, and the edge direction belongs to a diagonal direction. When it is detected that the pixel to be interpolated is a right-angled corner of the image, the edge direction is selected to be horizontal, so that deformation of an image such as a character having a right-angled corner pattern can be suppressed. This has the effect that it can be performed.

The video printer according to the present invention performs a predetermined interpolation process on image data of a video image, and prints the image data with a print head using the interpolated image data. And a filter means for selecting a coefficient in accordance with the edge direction detected by the edge detection means and filtering the predetermined frequency band with a predetermined gain based on the frequency characteristic of the selected coefficient. Since the image is printed by interpolating a specific pixel of the image twice in one direction, the jagged component in the edge direction is cut, and the steepness of the edge in the direction perpendicular to the edge direction is saved. High-quality interpolated images with smooth interpolation of diagonal edges, less blurring of images and less deformation of characters An effect that can be obtained printout.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a video printer to which an image interpolation device according to an embodiment is applied.

FIG. 2 is a block diagram of an image processing unit according to the embodiment.

FIG. 3 is a block diagram illustrating a configuration of a field / frame interpolation block according to the present embodiment.

FIG. 4 is a functional block diagram of an edge detection unit according to the embodiment.

FIG. 5 is a functional block diagram of a two-dimensional adaptive filter according to the present embodiment.

FIG. 6 is a diagram illustrating a tilt vector of an image according to the present embodiment.

FIG. 7 is a diagram illustrating a tilt vector of an interpolation image according to the present embodiment.

FIG. 8 is a diagram illustrating a region in an edge direction of an image according to the present embodiment.

FIG. 9 is a diagram showing a right-angled corner pattern according to the present embodiment.

FIG. 10 is a diagram illustrating coefficients of a two-dimensional adaptive filter according to the present embodiment.

FIG. 11 is a diagram illustrating coefficients of a two-dimensional adaptive filter according to the present embodiment.

FIG. 12 is a diagram showing a frequency characteristic of a coefficient H ₅ (Z _i , Z _j ) of the present embodiment (a frequency characteristic of H ₁ (Z _i , Z _j ) is obtained by inverting left and right at X = 0); .

FIG. 13 is a diagram showing a frequency characteristic of a coefficient H ₆ (Z _i , Z _j ) according to the present embodiment (a frequency characteristic of H ₂ (Z _i , Z _j ) is a right and left inversion at X = 0); .

FIG. 14 is a diagram showing a frequency characteristic of a coefficient H ₇ (Z _i , Z _j ) according to the present embodiment (a frequency characteristic of H ₃ (Z _i , Z _j ) is obtained by inverting left and right at X = 0); .

FIG. 15 is a diagram showing a frequency characteristic of a coefficient H ₄ (Z _i , Z _j ) of the present embodiment.

FIG. 16 is a diagram illustrating frequency characteristics of a coefficient H ₈ (Z _i , Z _j ) according to the present embodiment.

[Explanation of symbols]

Reference Signs List 1 video input unit, 2 A / D converter, 3 memory control unit, 4 image memory, 5 switch unit, 6 system control unit, 7 mechanism control unit, 8 mechanism, 9 image processing unit, 10
Non-linear interpolation block (field / frame interpolation block), 11 thermal head, 12 registers,
13 main scanning direction controller, 14 sub-scanning direction controller, 15 line buffer, 16 sub-scanning interpolation block, 17 linear interpolation block, 18 resize block, 19 line buffer, 20 editing function block, 21 line buffer, 22 sharpness block, 23 heat storage Correction memory (short time constant), 24 heat storage correction memory (long time constant), 25 heat storage correction block, 26
Dither block, Dc system control unit control data, Di input image data, Do output image data, 30 simple enlargement unit, 31 double upsampling circuit, 32 zero-order hold circuit, 33 edge detection unit, 34 two-dimensional adaptive filter, 40 Maximum inclination direction detection means, 41 vertical direction detection means, 42 edge direction area designation means, 43 horizontal direction areas, 44 vertical direction areas,
45 oblique direction area, 46 right angle corner pattern detecting means, 50 edge direction corresponding coefficient selecting means, 51 coefficient (has frequency characteristics having directionality)

Claims (9)

[Claims] An edge detecting means for detecting a local edge direction of an image, a coefficient is selected based on the edge direction detected by the edge detecting means, and a predetermined frequency band is determined by a frequency characteristic of the selected coefficient. An image interpolating apparatus, comprising: filter means for applying a filter with a predetermined gain to a specific pixel of the image, wherein the specific pixel of the image is doubled in one direction. 2. The image interpolation apparatus according to claim 1, wherein said edge detecting means is detected by said maximum tilt direction detecting means for detecting a local maximum tilt direction of said image. A vertical direction detecting means for detecting a direction perpendicular to the maximum inclination direction as an edge direction, and an edge for dividing the edge direction detected by the vertical direction detecting means into a plurality of regions and specifying a region to which the edge direction belongs An image interpolating apparatus, comprising: a direction area designating unit; 3. The image interpolation device according to claim 1, wherein said filter means is a two-dimensional digital filter, and comprises a coefficient having a directional frequency characteristic, and said edge detection means among said coefficients. An edge direction corresponding coefficient selecting means for selecting a coefficient corresponding to the edge direction detected by the following. 4. The image interpolation apparatus according to claim 1, wherein said edge detecting means is detected by said maximum tilt detecting means for detecting a local maximum tilt direction of said image. Vertical direction detecting means for detecting the maximum inclination direction and the vertical direction as an edge direction, and an edge direction area for dividing the edge direction detected by the vertical direction detecting means into a plurality of areas and designating an area to which the edge direction belongs When the edge direction is detected by the specifying means and the edge direction area specifying means to be an oblique direction, and when it is detected that a pixel to be interpolated is a right-angled corner of an image, And a right-angle corner pattern detecting means for selecting an edge direction in the horizontal direction by the direction area specifying means. 5. A local edge direction of an image is detected, a coefficient is selected based on the detected edge direction, and a predetermined frequency band is filtered with a predetermined gain by a frequency characteristic of the selected coefficient. An image interpolation method, wherein a specific pixel of the image is interpolated twice in one direction. 6. The image interpolation method according to claim 5, wherein in the edge detection, a local maximum inclination direction of the image is detected, and a direction perpendicular to the detected maximum inclination direction is detected as an edge direction. And dividing the detected edge direction into a plurality of regions and designating a region to which the edge direction belongs. 7. The image interpolation method according to claim 5, wherein the filter is a two-dimensional digital filter, and an edge direction detected by the edge detection among coefficients having directional frequency characteristics. An image interpolation method, wherein a coefficient corresponding to is selected. 8. The image interpolation method according to claim 5, wherein the edge detection detects a local maximum inclination direction of the image, and sets an edge direction perpendicular to the detected maximum inclination direction. The detected edge direction is divided into a plurality of regions having at least a horizontal direction, a vertical direction, and a diagonal direction, and a region to which the edge direction belongs is specified, and the edge direction belongs to a diagonal direction. Detecting a pixel to be interpolated at a right-angled corner of the image, selecting the horizontal direction as the edge direction. 9. A video printer which performs a predetermined interpolation process on image data of a video image and prints the image data by a print head using the interpolated image data, wherein edge detection means for detecting a local edge direction of the image. Filter means for selecting a coefficient according to the edge direction detected by the edge detection means and filtering the predetermined frequency band with a predetermined gain based on a frequency characteristic of the selected coefficient, and specifying the image. A video printer characterized in that the pixels are interpolated twice in one direction and printed.