JP2001326812A - Resolution converter and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a resolution converter that can suppress occurrence of ringing and increase the resolution of an image without unsharpening the edges of the image. SOLUTION: An over-sample section 1 inserts interpolation pixel data between original pixel data adjacent to each other in image data to increase the number of pixels twice in both horizontal and vertical directions; resulting in 4 times in total. A nonlinear filter section 2 applies nonlinear filter arithmetic operation to output image data from the over-sampling section 1 by summing horizontal and vertical output components in forward/backward directions. A jaggy correction section 3 applies jaggy correction to image data outputted from the nonlinear filter section 2.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a resolution conversion apparatus and method used for enlarging an image in an image output device such as a video printer.

[0002]

2. Description of the Related Art An image output device such as a video printer is equipped with a resolution converter for converting the size (number of pixels) of an input image and outputting the converted image. The resolution converter converts an input image having a first pixel density into an output image having a second pixel density, and is therefore also referred to as a pixel density converter. In recent years, with the improvement in resolution of image output devices, the number of pixels has been increased to enlarge images. For example, a VG
A size (640 horizontal pixels × 480 vertical pixels) image data is converted into SVGA size (1280 horizontal pixels × 96 vertical pixels).
0 pixels).

[0003] A conventional resolution converter is generally constituted by a combination of an upsampler and a band-limiting filter such as a low-pass filter (LPF). Note that an upsampler generally increases the apparent number of pixels by inserting zeros between original pixels, that is, by inserting zeros. In this conventional resolution converter,
The image quality changes depending on the characteristics of the band limiting filter. The filter characteristics of the band limiting filter are determined in consideration of the cost and the amount of calculation of the band limiting filter, required image quality, and the like.

[0004]

The above-mentioned conventional resolution converter has the following two problems. First, as shown in FIG. 18, the first problem is that ringing is likely to occur at an edge portion (contour portion) of an image with an increase in resolution. In FIG. 18, (A) shows a desirable state in which ringing does not occur, and (B)
Indicates a state in which ringing has occurred. In particular, if a higher-order filter is used as the band limiting filter, ringing is likely to occur. The second problem is that, as shown in FIG. 19, as the resolution increases, the image is blurred at the edge portion of the image, and the sharpness deteriorates. FIG.
(A) shows a desirable state in which the image is not blurred, and (B) shows a state in which the image is blurred and the sharpness is deteriorated. Particularly, in a filter such as a linear interpolation filter in which ringing does not occur, sharpness is greatly deteriorated.

As described above, in the conventional resolution converter, when the resolution is increased (enlarged image), the sharpness is degraded when the occurrence of ringing is suppressed, and the ringing is frequently generated when the degradation of the sharpness is suppressed. There is a problem that it is difficult to achieve both in a desirable state.

SUMMARY OF THE INVENTION The present invention has been made in view of such a problem, and a resolution conversion apparatus and method capable of suppressing occurrence of ringing and increasing resolution without blurring an edge portion of an image. The purpose is to provide.

[0007]

According to the present invention, there is provided a resolution conversion apparatus for increasing the number of pixels of input image data. An oversampling section that inserts interpolated pixel data between original pixel data to double the number of pixels in the horizontal direction and doubles in the vertical direction to increase the number of pixels by four, and a horizontal direction output from the oversampling section. , And sequentially performs a filter operation in the forward and reverse directions in the horizontal direction as a set, adds the result of the filter operation in the forward direction and the result of the filter operation in the reverse direction, and sets a plurality of pixel data in the vertical direction. As one set, sequentially performing filter operations in the forward and reverse directions in the vertical direction, and adding the result of the filter operation in the forward direction and the result of the filter operation in the reverse direction. A resolution conversion device comprising: a non-linear filter unit that performs a non-linear filter operation on image data output from the oversampling unit. (2) In the resolution conversion method for increasing the number of pixels of input image data, the number of pixels is doubled in the horizontal direction and the vertical direction by inserting interpolated pixel data between adjacent original pixel data of the image data. An oversampling step of doubling and increasing by four times, and a plurality of horizontal pixel data output by the oversampling step as a set, and sequentially performing a filter operation in the forward and reverse directions in the horizontal direction. The result of the filter operation in the direction and the result of the filter operation in the reverse direction are added, and a plurality of sets of pixel data in the vertical direction are used as a set, and the filter operation is sequentially performed in the forward and reverse directions in the vertical direction. And the result of the filter operation in the opposite direction, thereby obtaining image data output in the oversampling step. There is provided a resolution converting method characterized by comprising a non-linear filtering step for performing nonlinear filter operation.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A resolution conversion apparatus and method according to the present invention will be described below with reference to the accompanying drawings. FIG. 1 is a block diagram showing an embodiment of the overall configuration of a resolution conversion apparatus according to the present invention, FIG. 2 is a diagram showing an operation of oversampling in an oversampling section 1 in FIG. 1, and FIG. 3 is a nonlinear filter in FIG. FIG. 4 is a diagram for explaining an operation in the unit 2, FIG. 4 is a conceptual diagram showing an operation in the nonlinear filter unit 2 in FIG. 1, FIG. 5 is a block diagram showing a specific configuration example of the nonlinear filter unit 2 in FIG. FIG. 6 shows filters 202 and 2 in FIG.
FIG. 7 is a block diagram showing a specific configuration example of FIG. 3, FIG. 7 is a diagram for explaining the operation of FIG. 6, and FIG. 8 is a function unit 2003 in FIG.
FIGS. 9 to 11 are diagrams for explaining the operation of the function unit 2003 in FIG. 7, and FIG. 12 is a diagram for explaining the operation of the nonlinear filter unit 2 in FIG. FIG. 13 is a diagram for explaining the effect of the resolution conversion apparatus and method of the present invention in comparison with a conventional example, and FIGS. 14 to 16 are diagrams for explaining the operation of the jaw correction unit 3 in FIG. FIG. 17 is a diagram showing a program list as an example of a jaggy correction calculation in the jagged correction unit 3 in FIG.

In FIG. 1, image data to be subjected to resolution conversion is input to an oversampling section 1 and oversampled (oversampling step).
The image data is R (red), G (green), B (blue) or Y (yellow), M (magenta), C (cyan), K
It is input as a primary color signal such as (black) or a luminance signal and a color difference signal. In the present invention, the oversampling in the oversampling unit 1 is doubled in the horizontal direction and 2 times in the vertical direction.
The number of pixels per area is quadrupled.

As shown in FIG. 2, the oversampling in the oversampling section 1 is the same as the original pixel data between two adjacent original pixel data indicated by solid circles as interpolated pixel data indicated by a broken circle. This is done by inserting. FIG. 2 shows either the horizontal direction or the vertical direction.

The image data output from the oversampling unit 1 is input to a non-linear filter unit 2 and subjected to a non-linear filter operation described later in detail (non-linear filtering step). Hereinafter, the operation in the nonlinear filter unit 2 will be described in order. In FIG. 3, S is one screen represented by image data to be subjected to resolution conversion, and the horizontal direction is the horizontal direction and the vertical direction is the vertical direction. If the present invention is applied to a printing device such as a video printer, the horizontal direction is the main scanning direction, and the vertical direction is the sub-scanning direction.

The non-linear filter unit 2 performs two filter operations in the horizontal direction and also performs two filter operations in the vertical direction. That is, the non-linear filter unit 2 performs a horizontal filter operation with the function fh0 in the forward direction from the horizontal start end (left end in FIG. 3) to the end (right end in FIG. 3) of the screen S, A horizontal filter operation of a function fh1 is performed in the reverse direction from the horizontal end to the start end of the screen S. The nonlinear filter unit 2 performs a vertical filter operation of a function fv0 in the forward direction from the vertical start (upper end in FIG. 3) to the end (lower end in FIG. 3) of the screen S, and A vertical filter operation of a function fv1 is performed in the reverse direction from the vertical end to the start end of the screen S.

FIG. 4 conceptually shows the operation of the nonlinear filter unit 2. In FIG. 4, image data of the screen S is subjected to a horizontal filter operation by a horizontal filter unit 2H, and a vertical filter operation is performed by a vertical filter unit 2V. The image data is represented by the function fh0
And the filter 22 of the function fh1. The adder 23 adds the outputs of the filters 21 and 22. The output of the adder 23 is input to the filter 24 of the function fv0 and the filter 25 of the function fv1. The adder 26 adds the outputs of the filters 24 and 25.

Here, for convenience, filters 21, 22, 2
Although the names of the functions 4 and 25 are distinguished from each other as fh0, fh1, fv0, fv1, they may all be the same function. The functions fh0, fh1, fv0, fv1 may be different from each other, but this is not necessary. Functions fh0, fh1, fv0,
It is preferable that all the fv1s be the same function for simplification of the configuration or processing. FIG. 4 is a conceptual diagram in which a horizontal filter operation is first performed by the horizontal filter unit 2H, and then a vertical filter operation is performed by the vertical filter unit 2V. The processing may be reversed, and the order of the processing is not limited.

The nonlinear filter unit 2 is specifically configured as shown in FIG. 5 as an example. In FIG. 5, the image data of the screen S is input to the image memory 201 and written. The image memory 201 stores image data for one screen of the screen S. Image data is read from the image memory 201 in the forward direction, and is input to the filter 202. The filter 202 performs a filter operation process of a function f, and the operation result is written to the line memory 204. Note that the filter 202 outputs two signals, Out0 and Out1.

After the processing in the forward direction, the image memory 201
Image data is read out from the
3 is input. The filter 203 performs filter operation processing of a function f and outputs the result. Note that the filter 203 also outputs two signals out0 and out1. Filter 2
02, 203 have the same function f, and the functions fh0, fh1,
It is a generic term for fv0 and fv1. Thus, the filter 20
Reference numeral 2203 outputs a signal every two pixels.

At the same time as the processing in the filter 203, the image data is read from the line memory 204 in the forward direction as described above, and the signals out0 and out1 subjected to the filter operation processing by the filter 202 are read in the reverse direction. This is because the image data is read out in the reverse direction, and the filter 203 performs filter operation processing on the signals out0 and out1 to match the phases of the pixels.

The adder 205 adds the signal out0 output from the line memory 204 and the signal out0 output from the filter 203. Here, the addition processing is to add the two input signals and divide by two, that is, to take an average. The adder 206 includes a line memory 204
The signal out1 output from the filter 203 and the signal out1 output from the filter 203 are added. The addition process here is also
That is, the two input signals are added and divided by 2, that is, an average is obtained. The outputs of the adders 205 and 206 are written to the image memory 201.

The above processing is first performed for each row in the horizontal direction. The image data of each row output from the adders 205 and 206 is overwritten on the image data of the original row in the image memory 201. When the processing for all the rows is completed, the above processing is performed for each column in the vertical direction. When the processing for all the columns is completed, the filter calculation processing for all the image data of the screen S is completed. FIG.
The filters 202 and 203 in FIG.
1, 22, 24, 25.

Here, a specific configuration example of the filters 202 and 203 and a function f will be described. Filter 202,
203 has the same configuration, and is configured as shown in FIG. 6 as an example. In FIG. 6, image data is input to a delay unit 2001 for two pixels and a function unit 2003. The output of the delay unit 2001 is a delay unit 2002 for one pixel,
It is input to the function unit 2003 and the adder 2005. The one pixel and two pixels here are one pixel and two pixels of the image data oversampled by the oversampling unit 1. The output of the delay unit 2002 is the adder 2
004. The output of the adder 2004 is the function unit 2
003 is input.

The incoming image data,
The pixel data input to 1 is collectively referred to as c, the pixel data output from the delay unit 2001 is referred to as b, and the pixel data output from the adder 2004 is referred to as a. In FIG. 7, a solid circle represents original pixel data, and a broken circle represents interpolated pixel data. The pixel data a1, b1, and c1 shown in FIG.
This is the first pixel data from which data has begun to be input from the image memory 201. The pixel data a1, b1, and c1 have a positional relationship as shown in FIG. When data is read in the forward direction, pixel data a1, b1, and c1 are:
The three original pixel data from the horizontal or vertical start end of the screen S. When data is read in the reverse direction, the pixel data a1, b1, and c1 are the horizontal or vertical end of the screen S. Are the three original pixel data.

The function unit 2003 calculates a function f (a, b, c) based on a set of three pieces of input pixel data a, b, and c.
As a result, two correction data cor0 and cor1 are output. The correction data cor0 is correction data for the image data b. The result of adding the correction data cor0 and the image data b by the adder 2005 becomes the output signal out0. The correction data cor1 is correction data for image data b and c (interpolated pixel data inserted between image data b and c due to oversampling). The result of adding the correction data cor1 and the interpolated pixel data between the image data b and c by the adder 2004 becomes the output signal out1. Therefore, the output signal ou as the output 1 based on the pixel data a1, b1, c1
t0 and out1 are output to the positions shown in FIG.

The signal out1 is circulated as pixel data a of an input signal to the function unit 2003, and becomes pixel data a2 in the next processing. The next three pixel data of the three pixel data a1, b1, c1 processed by the function f (a, b, c) are called a2, b2, c2, and the next three pixel data are a3, b3, It will be referred to as c3.
Outputs based on the pixel data a2, b2, and c2 are referred to as output 2, and outputs based on the pixel data a3, b3, and c3 are referred to as output 3. Output signals out0 and ou as outputs 2 and 3 are provided.
t1 is output to the position shown in FIG.

Thus, the filters 202 and 203
In succession, two output signals out0 and out1 are generated based on one set of the inputted three pixel data a, b and c. As can be understood from the above description, the first three pixel data a1,
Only b1 and c1 are original pixel data, and have a positional relationship separated by two pixels. The next pixel data a2
In b2 and c2, pixel data b2 and c2 are original pixel data, pixel data a2 is interpolation pixel data, and a2, b
The two pixels are separated by one pixel, and the pixels b2 and c2 are separated by two pixels. Subsequent pixel data a3, b3, c3 are the same as pixel data a2, b2, c2.

The function unit 2003 performs an operation using two adjacent original pixel data b and c and one interpolated pixel data a except for the first three pixel data a1, b1 and c1. One piece of interpolated pixel data a is not interpolated pixel data between two pieces of original pixel data b and c, but is interpolated pixel data adjacent to only one of the two pieces of original pixel data b and c. is there. After the pixel data a2, b2, and c2, the function unit 2003 sequentially shifts the pixel data a, b, and c used for the function f (a, b, c) by two pixels.

The filters 202 and 203 have the same configuration as the FIR filter in which the shift unit is two pixels. However, since the pixel data a circulates through the output signal out1,
It has the same configuration as the IIR filter. The filters 202 and 203 used in the present invention are an FIR filter and an IIR filter.
Modified 3-tap FI with features of both IR filter
It can be considered an R filter.

FIG. 8 shows a program list of the function f (a, b, c) in C language. This function f (a, b, c)
Is to select one of two operation methods according to the internal variable gain. According to the equation shown in FIG.
One sign is determined. The internal variable gain is determined by the values of the pixel data a, b, and c, and the sign thereof is determined by the magnitude relation of the pixel data a, b, and c. From the formula, a
The sign of the internal variable gain is positive when <b <c or a>b> c, and negative or 0 otherwise. Internal variable ga
The correction data cor1 when in is negative is an expression, and the internal variable
Correction data cor0 when gain is negative is obtained from the equation.
At other times (when the internal variable gain is not negative), the correction data cor0 and cor1 are obtained from the equations (in {}).

FIG. 9 shows the result obtained by the function f (a, b, c). In FIG. 9, (A) is an internal variable ga
FIG. 9 shows the case where in is positive and (B) the internal variable gain is negative.
Solid lines in (A) and (B) indicate input image data,
The broken line indicates the output image data (computed result by the function f (a, b, c)). The definition of the solid circle and the broken circle is the same as in FIGS. As explained in FIG.
The pixel data a is original pixel data indicated by a solid circle in the first pixel data a1, and the following pixel data a2 is interpolated pixel data indicated by a broken circle as shown in (a). . The case where the internal variable “gain” is positive indicates a state where the pixel data “b” is between pixel data “a” and “c”. At this time, the correction data cor0 and cor1 have the same absolute value and opposite signs from each other, from within {} of the equation. Internal variable gain
Is negative when the pixel data b is not between the pixel data a and c. Correction data cor0, cor1 at this time
Is as shown in FIG. 9 (B).

Further, the operation when the internal variable gain is positive will be described. In FIG. 10, (A) shows ac (a
(B) shows the case where a-c is small, when minus c) is large. The definitions of the solid and broken lines, the solid circle and the broken circle in FIG. 10 are the same as in FIG. This is the same in FIGS. 11, 13 and 16 described later.
If ac is large as shown in FIG.
Becomes larger, and when ac is small as shown in FIG. 10B, the internal variable gain becomes smaller. The fact that a-c is large means that there is a high possibility that the part is an edge part of the image data. When the internal variable gain increases, the correction data cor0 and cor1 increase according to the equation.

11A shows a case where the pixel data b is located at the center between the pixel data a and c, and FIG. 11B shows a case where the pixel data b is shifted from the center between the pixel data a and c. Is shown. When the pixel data b is located at the center between the pixel data a and c as shown in FIG.
b) = (bc) = (ac) / 2, the internal variable ga
in becomes the maximum, and the correction data cor0 and cor1 become the maximum.
As shown in FIG. 11B, the pixel data b is replaced with the pixel data a and c.
The further away from the center, the smaller the internal variable gain, and the smaller the correction data cor0 and cor1.

As described above, the amplitude a-c of the edge portion is obtained.
Are the same, the closer the pixel data b is to the center between the pixel data a and c (that is, the center of the edge portion), the larger the internal variable gain and the larger the correction data cor0 and cor1 are. , The intermediate pixel data b is the pixel data a,
This is because the edge portion looks duller as it approaches the center between c. In the present invention, the amplitude a of the edge portion is a
Even if −c is the same, the correction amount increases as the edge portion looks duller visually.

On the other hand, the operation when the internal variable gain is negative is as described with reference to FIG. From the equation in FIG. 8, when the internal variable gain is negative, the internal variable gain is the pixel data a,
It changes depending on how much the pixel data b projects upward or downward in FIG. 9B with respect to c (projection degree). As the degree of protrusion of the pixel data b increases, the absolute value of the internal variable gain increases. The correction data cor0 changes according to the internal variable gain, and the correction data cor1 changes the internal variable ga.
It is constant without being affected by in. Therefore, the degree of protrusion in the pixel data b is adaptively increased as indicated by a broken line according to the original degree of protrusion, and the sharpness is increased.

As described above, the function unit 2003 determines, based on the function f (a, b, c) shown in FIG. 8, the positional relationship between the pixel data a, b, and c, that is, whether or not the pixel data a, b, and c can be considered as an edge. ,
If it is determined that the data is an edge, the correction data cor0, cor0, is determined according to the state of the edge.
This means that cor1 is adaptively changed. Correction data
Cor0 and cor1 adaptively change, and filter 2
The output signals out0 and out1 from 02 and 203 also change adaptively.

That is, the nonlinear filter unit 2 (function unit 200)
3) means for detecting the state of the image data output from the oversampling section 1, specifically, edge detecting means for detecting whether or not an edge is present, and detecting whether or not a protruding part is present. That is, the projection detecting means is provided. Further, a correction characteristic varying means for varying the correction characteristic for the edge portion or the protruding portion according to the amplitude of the edge portion or the degree of protrusion of the protruding portion is provided. The function f shown in FIG.
Since (a, b, c) is not a higher-order filter, no ringing occurs.

Here, the effect of the cyclic operation of the pixel data a in the filters 202 and 203 will be described.
In FIG. 12, (A) shows a case where the tour is not performed, and (B)
Is the case where the tour is performed. When the tour is not performed, as shown in FIG. 12A, the image data is reversed at the portion surrounded by the dashed line, and the image data is discontinuous. On the other hand, when the tour is performed, as shown in FIG. 12B, the inversion phenomenon of the image data does not occur in the portion surrounded by the one-dot chain line, and the image data is continuous.

According to the present invention, the filters 202 and 203 perform a cyclic operation so that discontinuous portions do not occur in the image data output from the nonlinear filter unit 2. Thereby, deterioration of the image quality is avoided.
As described with reference to FIG. 2, the nonlinear filter unit 2
By performing filter operations in the forward and reverse directions and adding the results of both operations, the details of the image data output from the nonlinear filter unit 2 become smoother.

The operation from the oversampling section 1 to the nonlinear filter section 2 will be described in comparison with the operation according to the conventional example. FIG. 13 shows the processing at the edge portion of the image data. In FIG. 13, (B) shows the original image, that is, the image before being doubled in the horizontal or vertical direction by the oversampling unit 1. The original image shown in FIG. 13B originally has twice the horizontal and vertical directions (horizontal and vertical) in FIG.
It is assumed that the image shown in FIG. This is referred to as a virtual double image. It is assumed that, when the virtual double image shown in FIG. 13A is reduced to と, the original image shown in FIG. 13B is obtained. That is, conversely, ideally, FIG.
When the original image shown in FIG. 3 (B) is enlarged twice, FIG.
It is to become.

However, when an original image is reduced to one half, image data having an edge portion as shown in FIG. 13B is generated because the original image shown in FIG. Not only if. The image data as shown in FIG. 13B is generated because the original image can be in various states.
It can be seen that it is impossible to reproduce the virtual double image shown in FIG. This is obvious when the amount of information of the image data is considered.

Therefore, conventionally, by using a band limiting filter, only the information on the frequency of the image data shown in FIG. 13B is left, and the aliasing component (unwanted harmonic component) generated by oversampling is eliminated. It is common to remove it. FIG. 13D shows image data obtained when a 2-tap LPF is used as a band-limiting filter when the conventional configuration doubles the size. Note that the result obtained by the 2-tap LPF is exactly the same as the result obtained by the linear interpolation.

On the other hand, the image data output from the nonlinear filter unit 2 in the configuration of FIG.
3 (C). In the present invention, as described above, the state of the edge of the image data output from the non-linear filter unit 2 changes adaptively depending on the state of the edge of the original image. At the edge portion with a large amplitude, the state of the edge portion becomes sharper as it approaches FIG. At the edge portion having a small amplitude, the state of the edge portion becomes gentler as it approaches FIG. 13D. Considering this from the viewpoint of frequency, the aliasing component generated by the oversampling unit 1 is emphasized at the edge portion having a large amplitude, and the aliasing component is limited at the edge portion having a small amplitude. Nonlinear filter unit 2
Can be said to be a filter that adaptively varies the aliasing component according to the state of the edge portion of the original image.

Here, returning to FIG. 1, the image data output from the nonlinear filter unit 2 is input to the jaw correction unit 3 and subjected to jag correction (jag correction step). The enlarged image data output from the non-linear filter unit 2 has good characteristics in the horizontal and vertical directions. However, good characteristics are not obtained in other directions (oblique directions). When a two-dimensional image filter is configured, as described in the present embodiment, processing is generally performed separately in the horizontal direction and the vertical direction. This is because the amount of calculation and circuit scale required for the processing are reduced. In this case, since the characteristics of the two-dimensional image filter are determined by the product of the filter functions in the horizontal and vertical directions, the characteristics in the oblique direction are limited. The jaw correction unit 3 is for correcting characteristics in an oblique direction.

In FIG. 14, (A) shows an original image before oversampling by the oversampling unit 1, and (B) shows an image oversampled by the oversampling unit 1 and filtered by the nonlinear filter unit 2. In FIG. 14, black circles indicate black pixels, and white circles indicate white pixels. Although FIG. 14B includes the original pixel data and the interpolated pixel data, they are not distinguished from each other by the solid circle and the broken circle as shown in FIG. In FIG. 14B and FIG. 15 to be described later, the original pixel data and the interpolated pixel data are not distinguished from each other, but are distinguished between black pixels and white pixels.

The original image shown in FIG. 14A has an edge portion shown by a broken line. Since an aliasing component in an oblique direction remains in the output of the non-linear filter unit 2, FIG.
As shown in (B), the edge portion is jagged, and so-called jaggies occur. 15A shows a state before jaw correction, and FIG. 15B shows a state after jaw correction. As shown in FIG. 15, the jaw correction unit 3 corrects jaggies and smoothes oblique edges.

Using FIG. 16 and FIG.
Will be described in detail. The pixel to be jagged by the jagged correction unit 3 is
Only pixels based on the interpolated pixel data inserted by oversampling. As described above, the solid circle in FIG. 16 is a pixel based on the original pixel data, and the broken circle is a pixel based on the interpolation pixel data. Here, it is not illustrated to distinguish whether the colors of the original pixel data and the interpolated pixel data are black pixels or white pixels.

As shown in FIG. 16, the jaw correction unit 3
Nine pixels of 3 pixels in the horizontal direction × 3 pixels in the vertical direction are processed as one group. Each of the nine pixels is referred to as D0 to D8.
Pixels D0, D2, D6, and D8 are pixels based on original pixel data, and pixels D1, D3, D5, and D7 are pixels based on interpolated pixel data. With the pixel D4 at the center of the nine pixels as the center, the interpolation axes for jagged correction are an axis L0 connecting the pixels D0 and D8, an axis L1 connecting the pixels D1 and D7, and a pixel D.
There are four directions, axis L2 connecting pixel 2 and pixel D6, and axis L3 connecting pixel D3 and pixel D5. The jaw correction unit 3 selects an interpolation axis for actually performing jaw correction from these four interpolation axes L0 to L3.

FIG. 17 shows the operation of the jaw correction unit 3 as a program list in C language. The jaw correction unit 3 determines the interpolation axes L0 to L3, determines which one is the most appropriate as the interpolation axis to be actually jagged, and selects an interpolation axis, and calculates an interpolation value using the selected interpolation axis. I do. In FIG. 17, the jaw correction unit 3 obtains the absolute value of the data difference between two pixels forming the four interpolation axes L0 to L3, as indicated by de [0] to de [3]. What should be selected as an interpolation axis for actually performing jagged correction is basically that the absolute value of the data difference between two pixels is the smallest. However, since erroneous interpolation will occur by itself, erroneous interpolation is prevented as follows.

As shown in FIG. 17, the axis with the smallest absolute value of the data difference between the interpolation axes L0 to L3 and the axis with the second smallest are obtained. Then, a difference between the two data differences is obtained, and if the difference is less than a predetermined threshold, no interpolation is performed. In FIG. 17, the threshold value is 16. No interpolation means that the pixel D4 in FIG. 16 is used as it is. If the second smallest axis is L0 or L2, that is, if the second smallest axis is an oblique axis, no interpolation is performed. This is because if both the smallest axis and the second smallest axis are oblique, the possibility of erroneous interpolation is extremely high.

After the interpolation axis is determined from the interpolation axes L0 to L3 in this manner, an interpolation value is generated by averaging two pixels on the interpolation axis. This interpolation value is used as new interpolation pixel data. The jaw correction unit 3 performs the processing shown in FIG. 17 on all the interpolated pixel data inserted by the oversampling. As a result, jagged correction is performed as shown in FIG.
Is corrected smoothly.

The present invention is not limited to the embodiment described above. The present invention can be configured by hardware or software by a computer program. Further, it can be configured by combining hardware and software. In the present invention, the resolution is converted to four times (two times vertically and twice) the number of pixels, but the present invention can be applied to conversion of an arbitrary multiple. For example, in order to enlarge the image by 9 times (3 times vertically and horizontally), it is necessary to enlarge the image by 4 times according to the present invention, and then enlarge the image by 1.5 times using a conventional configuration or the like. In this way, image quality degradation due to the conventional configuration can be minimized.
Further, in order to enlarge the image by 16 times (4 times vertically and horizontally), the present invention may be implemented in two stages.

[0050]

As described above in detail, the resolution conversion apparatus and method of the present invention double the number of pixels in the horizontal direction by inserting interpolation pixel data between adjacent original pixel data of image data.・ Double in the vertical direction and increase by a factor of four, and perform a filter operation in the forward and reverse directions in the horizontal direction as a set of a plurality of pixel data in the horizontal direction. The filter operation results are added, and a plurality of pixel data in the vertical direction are set as one set, and the filter operation is sequentially performed in the forward and reverse directions in the vertical direction, and the result of the forward filter operation and the result of the reverse filter operation are calculated. By performing the addition, a non-linear filter operation is performed on the image data, so that occurrence of ringing can be suppressed, and the resolution can be improved without blurring the edge portion of the image. It can be increased.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of the overall configuration of the present invention.

FIG. 2 is a diagram illustrating an operation of oversampling in an oversampling unit 1 in FIG.

FIG. 3 is a diagram for explaining an operation in the nonlinear filter unit 2 in FIG. 1;

FIG. 4 is a conceptual diagram showing an operation of the nonlinear filter unit 2 in FIG.

FIG. 5 is a block diagram illustrating a specific configuration example of a nonlinear filter unit 2 in FIG. 1;

FIG. 6 is a block diagram illustrating a specific configuration example of filters 202 and 203 in FIG.

FIG. 7 is a diagram for explaining the operation of FIG. 6;

8 is a diagram illustrating a program list as an example of a filter operation in a function unit 2003 in FIG. 7;

FIG. 9 is a diagram for explaining the operation of a function unit 2003 in FIG. 7;

FIG. 10 is a diagram for explaining the operation of a function unit 2003 in FIG. 7;

FIG. 11 is a diagram for explaining the operation of a function unit 2003 in FIG. 7;

FIG. 12 is a diagram for explaining the operation of the nonlinear filter unit 2 in FIG.

FIG. 13 is a diagram for explaining an effect of the present invention in comparison with a conventional example.

14 is a diagram for explaining the operation of the jaw correction unit 3 in FIG.

15 is a diagram for explaining the operation of the jaw correction unit 3 in FIG.

16 is a diagram for explaining the operation of the jaw correction unit 3 in FIG.

FIG. 17 is a diagram showing a program list as an example of a jaggy correction calculation in the jaw correction section 3 in FIG. 1;

FIG. 18 is a diagram for explaining a problem of the conventional example.

FIG. 19 is a diagram for explaining a problem of a conventional example.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 Over-sampling part 2 Nonlinear filter part 2H Horizontal filter part 2V Vertical filter part 3 Jaw correction part 21,22,24,25,202,203 Filter 23,26,205,206,2004,2005 Adder 201 Image memory 204 line Memory 2001, 2002 Delay unit 2003 Function part

 ──────────────────────────────────────────────────続 き Continued on the front page F-term (reference) PQ12 RR19

Claims (14)

[Claims] 1. A resolution conversion apparatus for increasing the number of pixels of input image data, wherein the number of pixels is doubled in the horizontal direction by inserting interpolation pixel data between adjacent original pixel data of the image data.
An oversampling unit that increases the number of pixels by two in the vertical direction and increases by a factor of four; The forward filter operation result and the backward filter operation result are added, and a plurality of vertical pixel data are sequentially processed as a set in the forward and reverse directions in the vertical direction. A resolution converter configured to include a non-linear filter unit that performs a non-linear filter operation on the image data output from the oversampling unit by adding the operation result and the filter operation result in the reverse direction. . 2. The image processing apparatus according to claim 1, wherein the non-linear filter unit detects an edge in the image data output from the oversampling unit, and varies a correction characteristic for the edge in accordance with an amplitude of the edge. 2. The resolution conversion device according to claim 1, further comprising a correction characteristic changing unit. 3. The resolution conversion apparatus according to claim 2, wherein said correction characteristic changing means increases the correction amount so that the edge portion becomes steeper as the amplitude of the edge portion becomes larger. 4. The plurality of pixel data used in the filter operation by the non-linear filter unit are two adjacent original pixel data and one interpolated pixel data adjacent to only one of the two original pixel data. The resolution conversion device according to claim 1, wherein 5. The correction characteristic varying means, wherein even if the amplitude of the edge portion is the same, intermediate pixel data among three pixel data of the two original pixel data and the one interpolation pixel data. 5. The resolution conversion apparatus according to claim 4, wherein the correction amount is increased so that the edge portion becomes steeper as it approaches the center of the edge portion. 6. The resolution conversion device according to claim 4, wherein said non-linear filter unit has a cyclic filter for circulating said one interpolated pixel data. 7. The resolution conversion device according to claim 1, further comprising a jagged correction section for performing jagged correction on image data output from said non-linear filter section. 8. A resolution conversion method for increasing the number of pixels of input image data, wherein the number of pixels is doubled in the horizontal direction by inserting interpolated pixel data between adjacent original pixel data of the image data.
An oversampling step of doubling and increasing the number of pixels in the vertical direction to four times, and sequentially performing a filter operation in the forward and reverse directions in the horizontal direction as a set of a plurality of pixel data in the horizontal direction output by the oversampling step; The forward filter operation result and the backward filter operation result are added, and a plurality of vertical pixel data are sequentially processed as a set in the forward and reverse directions in the vertical direction. A non-linear filtering step of performing a non-linear filter operation on the image data output in the oversampling step by adding the operation result and the filter operation result in the reverse direction. 9. The non-linear filtering step includes: an edge detection step of detecting an edge in the image data output in the oversampling step; and a correction characteristic for the edge according to an amplitude of the edge. 9. The resolution conversion method according to claim 8, comprising a step of changing a correction characteristic. 10. The resolution conversion method according to claim 9, wherein in the correction characteristic changing step, the correction amount is increased so that the edge portion becomes steeper as the amplitude of the edge portion becomes larger. 11. The plurality of pixel data used in the filter operation in the non-linear filtering step are two adjacent original pixel data and one interpolated pixel data adjacent to only one of the two original pixel data. 11. The resolution conversion method according to claim 8, wherein: 12. The correction characteristic changing step, wherein even if the amplitude of the edge portion is the same, intermediate pixel data among three pixel data of the two original pixel data and the one interpolation pixel data. 12. The resolution conversion method according to claim 11, wherein the correction amount is increased so that the edge portion becomes steeper as the edge portion approaches the center of the edge portion. 13. The nonlinear filtering step according to claim 13,
13. The resolution conversion method according to claim 11, further comprising a circulating step of circulating the one interpolated pixel data. 14. The resolution conversion method according to claim 8, further comprising a jagging correction step of performing jagging correction on the image data output by said non-linear filtering step.