JP2008167027A - Image processor, image processing method and image processing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processor, image processing method and image processing program for magnifying an image, in such a manner that the jaggies of an edge in an oblique direction is not conspicuous, and overshooting/undershooting of the edge does not occur, while maintaining edge sharpness. <P>SOLUTION: The image processor for magnifying an input image includes an area extracting part 1 for extracting a target area, including a target pixel from the input image; a differentiation calculation part 2 for calculating the derivative values of the pixels in the target area; a pattern classification deciding part 3 for judging an area pattern in the target area, on the basis of the derivative value; an interpolation coefficient calculating part 4 for calculating the interpolation coefficient for calculating the pixel value of an interpolation pixel, in the target area, on the basis of the area pattern and the derivative value; and an interpolation calculating part 5 for calculating the pixel value of the interpolated pixel, on the basis of the interpolation coefficient. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an image processing apparatus, an image processing method, and an image processing program for enlarging an image having a small number of pixels and converting the image into an image having a large number of pixels.

  In order to enlarge an image with a small number of pixels and convert it to an image with a large number of pixels, it is necessary to interpolate the pixels using the pixel value of the original pixel. As the interpolation method, the nearest neighbor method (nearest neighbor method) in which the pixel value of the interpolation pixel is substituted using the pixel value of the original pixel nearest to the interpolation point, four points around the interpolation point (2 × 2) Linear interpolation method (bilinear method) that calculates linearly using the original pixel values, and bicubic interpolation method (bilinear method) that calculates by a cubic function using the original pixel values of 16 points (4 × 4) around the interpolation point. The cubic method) is known.

  The nearest neighbor method has the advantage of being able to perform high-speed processing because it is simple and can sharpen the horizontal and vertical edges, but on the other hand, because the pixels are simply enlarged, the diagonal direction There is a problem that the edge becomes jaggy.

  The linear interpolation method calculates the weighted average value of the original pixel values of the four surrounding points as an interpolation value, and thus has an advantage that the jaggies as in the nearest neighbor method can be made inconspicuous by the smoothing effect. Since the value is set as an interpolation value, there is a problem that the sharpness of the edge is lowered, the fine texture portion is crushed, and the effect of reducing the jaggies in the oblique direction edge cannot be obtained sufficiently depending on the angle.

  The bicubic interpolation method is a weighted average value of 16 surrounding original pixel values, and is approximated by a cubic function. Therefore, the edge sharpness is higher than that of linear interpolation, and the jaggy like the nearest neighbor method is not noticeable. On the other hand, there is a problem that overshoot / undershoot occurs at the edge, and a pseudo contour occurs at the edge of the edge, or the effect of reducing the jaggies in the oblique direction is not sufficient depending on the angle. .

As a conventional interpolation method for solving the above problem, for example, the shape of a partial area of the input image is determined based on comparison with a plurality of shape patterns prepared in advance, and the determined shape pattern is supported. There has been proposed a method of performing interpolation by the interpolation processing (see, for example, Patent Document 1).
As another interpolation method, the mixing ratio between the pixel value of the original pixel closest to the interpolation point (near-position pixel value) and the interpolation value corresponding to the interpolation point in the calculation target pixel section of the original image, that is, There has been proposed a method of calculating a pixel value of an interpolation pixel by determining a mixing ratio of a plurality of interpolation methods based on a contrast amount in the vicinity of a calculation target pixel section (for example, see Patent Document 2).

Japanese Patent Laid-Open No. 9-331447 Japanese Patent Laid-Open No. 10-341334

  However, the interpolation method disclosed in Patent Document 1 has a problem in that when the shape of the input image is determined by pattern matching and there is no optimal shape pattern, the image quality is degraded in the enlarged image. is there. Further, when trying to prevent such image quality deterioration, it is necessary to deal with the shapes of all input images, and there is a problem that the number of types of shape patterns prepared in advance increases as much as possible.

  Further, in the interpolation method disclosed in Patent Document 2, in the region where the edge exists, when calculating the near-position pixel value, the interpolation value, and the contrast amount, the value of the pixel having low correlation is referred to. There is a problem that the sharpness of the edge is lowered.

  Accordingly, an object of the present invention made in view of such a point is an image processing capable of enlarging an image so that diagonal edges are not noticeable, and edge overshoot / undershoot does not occur while maintaining edge sharpness. An apparatus, an image processing method, and an image processing program are provided.

An image processing apparatus according to claim 1 that achieves the above object is an image processing apparatus for enlarging an input image.
A region extraction unit that extracts a region of interest including a pixel of interest from the input image;
A differential operation unit for calculating a differential value of a pixel in the region of interest;
A pattern classification determination unit that determines a region pattern in the region of interest based on the differential value;
An interpolation coefficient calculation unit that calculates an interpolation coefficient for calculating a pixel value of an interpolation pixel in the region of interest based on the region pattern and the differential value;
An interpolation calculation unit that calculates a pixel value of the interpolation pixel based on the interpolation coefficient;
It is characterized by having.

The invention according to claim 2 is the image processing apparatus according to claim 1,
The differential operation unit includes a plurality of differential filters that calculate differential values in at least four directions, and a differential absolute value that calculates a differential absolute value sum that is an absolute value sum of the plurality of differential values calculated by the plurality of differential filters. A sum calculator,
The pattern classification determination unit is configured to determine a region pattern in the region of interest based on the plurality of differential values and the sum of differential absolute values.

The invention according to claim 3 is the image processing apparatus according to claim 2,
The differential operation unit calculates the plurality of differential values and the sum of differential absolute values for the pixel of interest.

The invention according to claim 4 is the image processing apparatus according to claim 2,
The differential calculation unit calculates the plurality of differential values and the differential absolute value sum for all pixels in the region of interest,
The pattern classification determination unit determines a region pattern in the region of interest based on statistics of the region pattern determined based on the plurality of corresponding differential values and the sum of differential absolute values for all the pixels. It is a feature.

The invention according to claim 5 is the image processing apparatus according to any one of claims 2 to 4,
The interpolation coefficient calculating unit calculates a plurality of directional interpolation filter coefficients based on the classification code as the interpolation coefficient, and calculates an angle weighting coefficient for each of a plurality of directions based on the plurality of differential values. It is a feature.

The invention according to claim 6 is the image processing apparatus according to claim 5,
The interpolation calculation unit calculates a plurality of direction interpolation values in the region of interest based on the direction interpolation filter coefficient, weights the calculated plurality of direction interpolation values based on the angle weighting coefficient, and It has an outline region interpolation calculation unit for calculating a pixel value of a target pixel as an outline region interpolation value.

Further, the invention of the image processing method according to claim 7 that achieves the above-described object,
Extracting a region of interest including a pixel of interest from the input image;
Calculating a differential value of a pixel within the extracted region of interest;
Determining a region pattern in the region of interest based on the calculated differential value;
Calculating an interpolation coefficient for calculating a pixel value of an interpolation pixel in the region of interest based on the determined region pattern and the calculated differential value;
Calculating a pixel value of the interpolation pixel based on the calculated interpolation coefficient;
It is characterized by including.

Furthermore, the invention of the image processing program according to claim 8 for achieving the above object is as follows:
On the computer,
A function for extracting a region of interest including a pixel of interest from an input image;
A function for calculating the differential value of the pixel within the extracted region of interest;
A function of determining a region pattern in the region of interest based on the calculated differential value;
A function of calculating an interpolation coefficient for calculating a pixel value of an interpolation pixel in the region of interest based on the determined region pattern and the calculated differential value;
A function of calculating a pixel value of the interpolation pixel based on the calculated interpolation coefficient and enlarging the input image;
It is characterized by realizing.

  According to the present invention, the region pattern in the region of interest is determined based on the differential value of the pixel in the region of interest in the input image, and the interpolation coefficient in the region of interest is calculated based on the determined region pattern and differential value. Thus, since the pixel value of the interpolation pixel is calculated, it is possible to accurately determine the region pattern having the contour and perform the interpolation process. Therefore, it is possible to enlarge the input image so that diagonal edges are not noticeable and edge overshoot / undershoot does not occur while maintaining edge sharpness.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(First embodiment)
FIG. 1 to FIG. 14 show a first embodiment of the present invention, and FIG. 1 is a functional block diagram showing the main configuration of the entire image processing apparatus. The image processing apparatus according to the present embodiment classifies a pattern of a target pixel, a region extraction unit 1 that extracts a target region including a target pixel of an input image, a differential operation unit 2 that calculates a differential value of the target region, and the like. The pattern classification determination unit 3 includes an interpolation coefficient calculation unit 4 that calculates an interpolation coefficient for interpolation calculation, and an interpolation calculation unit 5 that performs interpolation calculation.

  The input image is input to the area extraction unit 1. The region extraction unit 1 is connected to the differentiation calculation unit 2 and the interpolation calculation unit 5, and the differentiation calculation unit 2 is connected to the pattern classification determination unit 3 and the interpolation coefficient calculation unit 4. The pattern classification determination unit 3 is connected to the interpolation coefficient calculation unit 4 and the interpolation calculation unit 5, and the interpolation coefficient calculation unit 4 is connected to the interpolation calculation unit 5 and outputs an output image from the interpolation calculation unit 5.

  FIG. 2 is a flowchart showing a schematic operation of the image processing apparatus shown in FIG. The input image is sequentially input to the region extraction unit 1 in units of pixels. The area extraction unit 1 has a memory. When pixel data of an input image is input, the region extraction unit 1 performs a delay until pixel data for a predetermined line that enables image enlargement processing is completed. When pixel data for a predetermined line capable of image enlargement processing is stored in the memory, the pixel value P (x, y) of the pixel of interest at (x, y) in the image and the surrounding pixel positions (x + i) , Y + j) and the pixel values P (x + i, y + j) of the surrounding pixels are extracted, and the attention area including the extracted attention pixel is output to the differential calculation section 2 and the interpolation calculation section 5 (step S1).

  The differential operation unit 2 calculates a differential value at the target pixel in the target region extracted by the region extraction unit 1 using a differential calculation filter, and the calculated differential value is sent to the pattern classification determination unit 3 and the interpolation coefficient calculation unit 4. Output (step S2).

  Based on the differential value calculated by the differential calculation unit 2, the pattern classification determination unit 3 determines whether the region of interest belongs to a continuous contour region, a texture region, or a flat region. Furthermore, when it is determined that the region of interest is a continuous contour region, the contour direction pattern is determined based on the differential value calculated by the differential calculation unit 2, and the region including this contour direction pattern A classification code is determined from the pattern determination result, and is output to the interpolation coefficient calculation unit 4 and the interpolation calculation unit 5 (step S3).

  The interpolation coefficient calculation unit 4 calculates an angle weighting coefficient for determining the weights of the plurality of interpolation filters based on the differential value calculated by the differentiation calculation unit 2 and also the region pattern determined by the pattern classification determination unit 3. Based on the classification code, a plurality of interpolation filter coefficients corresponding to a plurality of angles for performing interpolation processing along the contour line are calculated, and the calculated angle weighting coefficients and interpolation filter coefficients are output to the interpolation calculation unit 5. (Step S4). In the present embodiment, three types of interpolation filter coefficients corresponding to three types of adjacent angles are calculated.

  For the contour region, the interpolation calculation unit 5 calculates a contour interpolation value based on the plurality of interpolation filter coefficients and the angle weighting coefficient calculated by the interpolation coefficient calculation unit 4 (step S5), and adds a texture region to the texture region. Performs a filtering process on the input image with a texture interpolation filter coefficient prepared in advance to calculate a texture interpolation value (step S6). Further, for a flat region, filtering is performed on the input image with a flat interpolation filter coefficient prepared in advance. Processing is performed to calculate a flat interpolation value (step S7), and a pixel value of an interpolation pixel in the region of interest is calculated. After that, based on the classification code determined by the pattern classification determination unit 3, one of the calculated contour interpolation values, texture interpolation values, and flat interpolation value interpolation pixels is selected, and the interpolation pixel is selected. The output image is output (step S8), and the process ends.

  Next, the configuration and operation of each part of the image processing apparatus shown in FIG. 1 will be described in more detail.

  FIG. 3 is a functional block diagram showing a configuration of a main part of the differential operation unit 2 shown in FIG. The differential operation unit 2 includes a horizontal differential value calculation unit 21, a vertical differential value calculation unit 22, an upper right differential value calculation unit 23, an upper left differential value calculation unit 24, and a differential absolute value sum calculation unit 25. Yes. The input image is input to the horizontal differential value calculation unit 21, the vertical differential value calculation unit 22, the upper right differential value calculation unit 23, and the upper left differential value calculation unit 24 through the region extraction unit 1, respectively.

  The horizontal differential value calculation unit 21 calculates the pixel value P (x + i, y + j) of the pixel of interest in the region of interest extracted by the region extraction unit 1 according to the following equation (1) and either FIG. A horizontal differential value EdgeH is calculated by performing a convolution operation with a known differential filter filterH (i, j) shown in a), and the calculated horizontal differential value EdgeH is supplied to the pattern classification determination unit 3 and the interpolation coefficient calculation unit 4. While outputting, it outputs to the differential absolute value sum calculation part 25.

  The vertical differential value calculation unit 22 calculates the pixel value P (x + i, y + j) of the pixel of interest in the region of interest extracted by the region extraction unit 1 according to the following equation (2) and either FIG. The vertical differential value EdgeV is calculated by performing a convolution operation with the known differential filter filterV (i, j) shown in b), and the calculated vertical differential value EdgeV is sent to the pattern classification determination unit 3 and the interpolation coefficient calculation unit 4. While outputting, it outputs to the differential absolute value sum calculation part 25.

  The upper right differential value calculation unit 23 calculates the pixel value P (x + i, y + j) of the target pixel in the target region extracted by the region extraction unit 1 according to the following equation (3) and either FIG. 4C or FIG. A convolution operation with a known differential filter filterR (i, j) shown in c) is performed to calculate an upper right differential value EdgeR, and the calculated upper right differential value EdgeR is supplied to the pattern classification determination unit 3 and the interpolation coefficient calculation unit 4. While outputting, it outputs to the differential absolute value sum calculation part 25.

  Similarly, the upper left differential value calculation unit 24 calculates the pixel value P (x + i, y + j) of the target pixel in the target region extracted by the region extraction unit 1 according to the following equation (4), and FIG. A convolution operation with the known differential filter filterL (i, j) shown in FIG. 5D is performed to calculate the upper left differential value EdgeL, and the calculated upper left differential value EdgeL is used as the pattern classification determination unit 3 and the interpolation coefficient calculation. Output to the unit 4 and output to the differential absolute value sum calculation unit 25.

  The differential absolute value sum calculation unit 25 calculates a differential absolute value sum EdgeA, which is the sum of the absolute values of the input differential values in the four directions, according to the following equation (5), and sends it to the pattern classification determination unit 3. Output.

  As described above, the differential calculation unit 2 calculates the differential value at the target pixel position by the differential calculation filter based on the target pixel in the target region extracted by the region extraction unit 1, and uses the calculated differential value as the pattern classification. This is output to the determination unit 3 and the interpolation coefficient calculation unit 4. This makes it possible to control the processing for both the edge shape pattern discrimination and the interpolation filter coefficient weighting based on the differential value.

  FIG. 6 is a functional block diagram showing a configuration of a main part of the pattern classification determination unit 3 shown in FIG. The pattern classification determination unit 3 includes positive / negative determination units 31, 32, 33, and 34 and a pattern determination unit 35. The horizontal direction differential value, the vertical direction differential value, the upper right direction differential value, and the upper left direction differential value output from the differential operation unit 2 are input to the positive / negative determination units 31, 32, 33, and 34, respectively, Input to the pattern determination unit 35.

  The positive / negative determination unit 31 determines whether the horizontal differential value EdgeH output from the horizontal differential calculation unit 21 is positive or negative, and outputs the result to the pattern determination unit 35. Here, as shown in FIG. 7A, the horizontal differential value EdgeH represents a density gradient vector in the normal direction of the horizontal edge, and when EdgeH is positive, the upward differential indicated by the white arrow is shown. A density gradient vector is shown. When EdgeH is negative, it indicates a downward density gradient vector indicated by a black arrow.

  The positive / negative determination unit 32 determines whether the vertical differential value EdgeV output from the vertical differential calculation unit 22 is positive or negative, and outputs the result to the pattern determination unit 35. Here, as shown in FIG. 7B, the vertical differential value EdgeV represents the density gradient vector in the normal direction of the vertical edge. When EdgeV is positive, the vertical differential value EdgeV is indicated by a white arrow. A density gradient vector is shown. When EdgeH is negative, it indicates a left-side density gradient vector indicated by a black arrow.

  The positive / negative determination unit 33 determines whether the upper right direction differential value EdgeR output from the upper right differential calculation unit 23 is positive or negative, and outputs the result to the pattern determination unit 35. Here, the differential value EdgeR in the upper right direction, as shown in FIG. 7C, represents the density gradient vector in the normal direction of the edge in the upper right direction, and when EdgeR is positive, the diagonal value indicated by the white arrow. A concentration gradient vector pointing upwards to the left is shown. When EdgeH is negative, this is a concentration gradient vector pointing diagonally downwards to the right indicated by a black arrow.

  Similarly, the positive / negative determination unit 34 determines whether the upper left direction differential value EdgeL output from the upper left differential calculation unit 24 is positive or negative, and outputs the result to the pattern determination unit 35. Here, the upper left direction differential value EdgeL represents a density gradient vector in the normal direction of the diagonally upper left edge, as shown in FIG. 7D, and when EdgeL is positive, the diagonal value indicated by a white arrow. A density gradient vector pointing in the upper right direction is shown. When EdgeL is negative, it indicates a density gradient vector pointing in the diagonally lower left direction indicated by a black arrow.

  In addition, the pattern determination unit 35 first determines the first region pattern based on the differential absolute value sum EdgeA calculated by the differentiation calculation unit 2 as to whether the region of interest is a flat region or an edge region. Make a decision. In this first determination, when EdgeA is less than a predetermined threshold, it is determined that the attention area is a flat area, and when EdgeA is equal to or larger than the predetermined threshold, the attention area is an area where an edge exists. It is determined that

  Further, in the first determination, when it is determined that the attention area is an area where an edge exists, the first determination is made based on the positive / negative of each differential value determined by the positive / negative determination units 31, 32, 33, and 34. 2 area pattern is determined. In this second determination, it is determined whether or not a condition pattern from 0 to 7 shown below is met, the area pattern is classified based on the outline pattern of the attention area, and the classification code shown in the conceptual diagram of FIG. To decide.

0: (EdgeH> 0, EdgeV> 0, EdgeR <0, EdgeL> 0)
1: (EdgeH> 0, EdgeV> 0, EdgeR> 0, EdgeL> 0)
2: (EdgeH> 0, EdgeV <0, EdgeR> 0, EdgeL> 0)
3: (EdgeH> 0, EdgeV <0, EdgeR> 0, EdgeL <0)
4: (EdgeH <0, EdgeV <0, EdgeR> 0, EdgeL <0)
5: (EdgeH <0, EdgeV <0, EdgeR <0, EdgeL <0)
6: (EdgeH <0, EdgeV> 0, EdgeR <0, EdgeL <0)
7: (EdgeH <0, EdgeV> 0, EdgeR <0, EdgeL> 0)

  In this second determination, as shown in FIG. 8, when the classification code matches any of the above condition patterns 0 to 7, it is determined that the attention area is an area pattern that is an outline area in any direction. Then, a suitable classification code is determined. On the other hand, if the pattern does not match any of the patterns 0 to 7, the attention area is determined to be an area pattern that is a texture area having a complicated shape edge, and the classification code is determined to be 8. If the region of interest is determined to be a region pattern that is a flat region in the first determination, the classification code is determined to be 9.

  The classification codes determined in the first determination and the second determination are output to the interpolation coefficient calculation unit 4 and the interpolation calculation unit 5.

  In this way, the pattern classification determination unit 3 determines the region pattern to determine whether the region of interest is a flat region, a texture region with a complex shape, or a continuous contour region, thereby performing an interpolation calculation. It is possible to select an optimal interpolation process in the unit 5. Further, when the attention area is a continuous outline area, the interpolation calculation unit 5 can perform interpolation along the outline by classifying the outline into eight-direction area patterns. Furthermore, since the determination of the area pattern is performed by relatively simple processing such as threshold processing of the sum of differential absolute values and positive / negative determination of each direction differential value, the processing speed can be increased and the hardware can be simplified.

  FIG. 9 is a functional block diagram showing a configuration of a main part of the interpolation coefficient calculation unit 4 shown in FIG. The interpolation coefficient calculation unit 4 includes a weighting coefficient calculation unit 41 that calculates weights of a plurality of interpolation values, and an interpolation filter coefficient calculation unit 42 that calculates filter coefficients for calculating a plurality of interpolation values. The weighting coefficient calculation unit 41 includes an angle coefficient calculation unit 43, and the interpolation filter coefficient calculation unit 42 includes a horizontal / vertical direction interpolation filter coefficient LUT44, an oblique direction interpolation filter coefficient LUT45, and an intermediate direction interpolation filter coefficient LUT46. It comprises memories 47, 48 and 49 for storing interpolation filter coefficients.

  The horizontal direction differential value, vertical direction differential value, upper right direction differential value, and upper left direction differential value calculated by the differential operation unit 2 are input to the angle coefficient calculation unit 43, and the classification code determined by the pattern classification determination unit 3 is , The horizontal / vertical direction interpolation filter coefficient LUT44, the oblique direction interpolation filter coefficient LUT45, and the intermediate direction interpolation filter coefficient LUT46.

  Based on the horizontal differential value EdgeH, the vertical differential value EdgeV, the upper right differential value EdgeR, and the upper left differential value EdgeL, which are input from the differential operation unit 2, the angle coefficient calculation unit 43 has the following formula (6) and ( 7) The horizontal / vertical edge intensity HV and the oblique edge intensity RL are calculated according to the equation (7), and the calculation results are output to the interpolation calculation unit 5 as angle weighting coefficients. In equations (6) and (7), HV and RL are set to values of 0 or more, and clipped to 0 when smaller than 0. K is an arbitrary coefficient for adjusting the weighting coefficient, and is determined based on the number of bits of the gradation, the magnitude of the differential absolute value sum, and the like.

  As is clear from the above equations (6) and (7), the horizontal / vertical edge strength HV and the oblique edge strength RL are expressed by the absolute value | EdgeH | of the horizontal differential value EdgeH and the absolute value | EdgeV | of the vertical differential value EdgeV. And the difference between the absolute value | EdgeL | of the upper left differential value EdgeL and the absolute value | EdgeR | of the upper right differential value EdgeR.

  FIG. 10 is a graph showing the magnitude of the absolute value of the differential value in each direction with respect to the angle of the contour line. The solid line in the graph indicates the size of | EdgeH |, the double line indicates the size of | EdgeV |, the dotted line indicates the size of | EdgeR |, and the alternate long and short dash line indicates the size of | EdgeL |.

  As is apparent from FIG. 10, when the angle of the contour line is close to the horizontal direction (around 0 ° and 180 ° in FIG. 10), | EdgeH | approaches the maximum value, and | EdgeV | approaches the minimum value, | EdgeL | and | EdgeR | are almost the same value. When the angle of the contour line is close to the vertical direction (near 90 ° in FIG. 10), | EdgeV | approaches the maximum value, | EdgeH | approaches the minimum value, and | EdgeL | and | EdgeR | It becomes the same value. Therefore, when the angle of the contour line is close to the horizontal direction or the vertical direction, the absolute value of the difference between | EdgeH | and | EdgeV | is maximum, and the absolute value of the difference between | EdgeL | and | EdgeR | Therefore, the value of HV shown in the above equation (6) is maximal, and the value of RL shown in the above equation (7) is a negative value, so it is clipped to 0.

  When the angle of the contour line is close to a 45 ° diagonally upward direction (near 45 ° in FIG. 10), | EdgeR | approaches a maximum value, | EdgeL | approaches a minimum value, and | EdgeH | and | EdgeV | Are almost the same value. Further, when the angle of the contour line is close to a 45 ° diagonally upward direction (around 135 ° in FIG. 10), | EdgeL | approaches a maximum value, | EdgeR | approaches a minimum value, and | EdgeH | | EdgeV | becomes substantially the same value. Therefore, when the angle of the contour line is close to a 45 ° diagonally upward or leftward direction, the absolute value of the difference between | EdgeR | and | EdgeL | becomes maximal, and the absolute value of the difference between | EdgeH | and | EdgeV | Since the value is close to 0, the RL shown in the above equation (7) becomes a maximum, and the HV value shown in the above equation (6) becomes a negative value, so it is clipped to 0.

  Further, when the angle of the contour line is close to the intermediate direction between horizontal / vertical and 45 ° obliquely (around 22.5 °, 67.5 °, 112.5 °, and 157.5 ° in FIG. 10), The absolute value of the difference between EdgeH | and | EdgeV | and the absolute value of the difference between | EdgeR | and | EdgeL | are substantially equal. Therefore, the HV shown in the above equation (6) and the above equation (7) are shown. Both RL values are close to zero.

  On the other hand, the horizontal / vertical direction interpolation filter coefficient LUT44, the diagonal direction interpolation filter coefficient LUT45, and the intermediate direction interpolation filter coefficient LUT46 refer to the lookup table (LUT) based on the classification code determined by the pattern classification determination unit 3. The horizontal / vertical direction interpolation filter coefficient filterHV (i, j), the diagonal direction interpolation filter coefficient filterRL (i, j), and the intermediate direction interpolation filter coefficient filterMid (i, j) are calculated, respectively. They are stored in the corresponding memories 47, 48, 49 and output to the interpolation calculation unit 5.

  Here, in the lookup table referred to by the horizontal / vertical direction interpolation filter coefficient LUT44, the diagonal direction interpolation filter coefficient LUT45, and the intermediate direction interpolation filter coefficient LUT46, the interpolation filter coefficient is determined in advance so as to perform interpolation processing along the contour line. It is a table that has been. For example, in the image diagram of FIG. 11, when an interpolation pixel is generated at a position indicated by a black circle using the density value of the original pixel in the attention area indicated by a white circle, as shown in FIG. Filter HV (i, j) that matches the classification code is selected from the filter coefficients corresponding to the horizontal or vertical contour, and the diagonal 45 ° contour as shown in FIG. Filter RL (i, j) that matches the classification code is selected from the filter coefficients corresponding to, and the contour line in the middle direction between horizontal / vertical and 45 ° diagonal as shown in FIG. FilterMid (i, j) that matches the classification code is selected from the corresponding filter coefficients. These filter coefficients are defined in advance so as to be weighted linearly based on the distance from the interpolation pixel position. In FIG. 12, a filter that generates an interpolation pixel at the upper right position of the region of interest is shown. However, by rotating the filter, interpolation at upper left, lower left, and lower right positions can be supported.

  FIG. 13 is a functional block diagram showing a configuration of a main part of the interpolation calculation unit 5 shown in FIG. The interpolation calculation unit 5 includes an outline region interpolation calculation unit 51, a texture region interpolation calculation unit 52, a flat region interpolation calculation unit 53, and an interpolation output selection unit 54. The input image is input to the contour region interpolation calculation unit 51, the texture region interpolation calculation unit 52, and the flat region interpolation calculation unit 53 via the region extraction unit 1, and the three types of interpolation filters calculated by the interpolation coefficient calculation unit 4 are used. The coefficient and the angle coefficient are input to the contour line area interpolation calculation unit 51. The outputs of the contour area interpolation calculation unit 51, texture area interpolation calculation unit 52, and flat area interpolation calculation unit 53 are input to the interpolation output selection unit 54. The classification code determined by the pattern classification determination unit 3 is input to the interpolation output selection unit 54, and an output image that is a final output value is output from the interpolation output selection unit 54.

  The contour region interpolation calculation unit 51 calculates the calculated horizontal / vertical direction interpolation filter coefficient filterHV (i, j), the diagonal direction interpolation filter coefficient filterRL (i, j), the intermediate direction interpolation filter coefficient filterMid (i, j), On the basis of the horizontal and vertical edge strengths HV and the diagonal edge strengths RL that are angle weighting coefficients, interpolation processing of the input image along the contour line is performed to calculate the contour region interpolation value, and the calculated contour region interpolation value Is output to the interpolation output selector 54. A more detailed configuration and operation of the contour line area interpolation calculation unit 51 will be described in detail with reference to FIG.

  The texture region interpolation calculation unit 52 calculates a texture region interpolation value by performing an interpolation process by a known interpolation method (for example, bicubic interpolation) that preserves a fine texture structure using a predefined interpolation filter. The texture area interpolation value thus output is output to the interpolation output selection unit 54.

  The flat area interpolation calculation unit 53 calculates a flat area interpolation value by a known interpolation method (for example, bilinear interpolation) that is simple, high-speed, and not easily affected by noise in the flat area, using a predefined interpolation filter, and calculates the calculated value. The flat area interpolation value thus obtained is output to the interpolation output selection unit 54.

  Based on the classification code determined by the pattern classification determination unit 3, the interpolation output selection unit 54 selects an outline area interpolation value when the attention area is an outline area, and texture area interpolation when the attention area is a texture area. A value is selected, and if it is a flat region, a flat region interpolation value is selected and output as a final output value.

  Here, the interpolation output selection unit 54 may be installed before the contour region interpolation calculation unit 51, the texture region interpolation calculation unit 52, and the flat region interpolation calculation unit 53. In this case, the interpolation output selecting unit 54 performs contour line region interpolation when the attention region is a contour region based on the classification code determined by the pattern classification determination unit 3 in the same manner as shown in FIG. Select the processing of the calculation unit 51, select the processing of the texture region interpolation calculation unit 52 if it is a texture region, select the processing of the flat region interpolation calculation unit 53 if it is a flat region, and select the selected interpolation calculation By turning on the processing of the part and turning off the processing of the unselected interpolation calculation part, the interpolation value to be output corresponding to the area indicated by the classification code is selected. In this embodiment, since the calculation is not performed for a process that is not selected among the contour region interpolation calculation unit 51, the texture region interpolation calculation unit 52, and the flat region interpolation calculation unit 53, the power consumption can be further suppressed. It becomes possible.

  Next, details of the interpolation processing along the contour line in the contour region interpolation calculation unit 51 will be described with reference to FIG.

  FIG. 14 is a functional block diagram illustrating a configuration of a main part of the contour line area interpolation calculation unit 51. The contour line area interpolation calculation unit 51 includes convolution calculation units 61, 62, and 63 and a weighted addition unit 64. The input images are input to the convolution operation units 61, 62, and 63 via the region extraction unit 1, respectively. The horizontal / vertical direction interpolation filter coefficient calculated by the interpolation coefficient calculation unit 4 is input to the convolution operation unit 61, the oblique direction interpolation filter coefficient is input to the convolution operation unit 62, and the intermediate direction filter coefficient is input to the convolution operation unit. 63. The outputs of the convolution operation units 61, 62, and 63 are input to the weighted addition unit 64. Further, the angle coefficient calculated by the interpolation coefficient calculation unit 4 is input to the weighting addition unit 64, and the contour line region interpolation value is output from the weighting addition unit 64 to the interpolation output selection unit 54.

The convolution operation unit 61, according to the following equation (8), the horizontal / vertical interpolation filter coefficient filterHV (i, j) input from the horizontal / vertical interpolation filter coefficient LUT44 and the input image P (x + i, y + j) The horizontal / vertical interpolation value _PHV is calculated, and the calculated horizontal / vertical interpolation value _PHV is output to the weighting addition unit 64.

The convolution operation unit 62 performs a convolution operation between the oblique direction interpolation filter coefficient filterRL (i, j) input from the oblique direction interpolation filter coefficient LUT 45 and the input image P (x + i, y + j) according to the following equation (9). Then, the diagonal interpolation value _PRL is calculated, and the calculated diagonal interpolation value _PRL is output to the weighted addition unit 64.

The convolution operation unit 63 performs a convolution operation between the intermediate direction interpolation filter coefficient filterMid (i, j) input from the intermediate direction interpolation filter coefficient LUT 46 and the input image P (x + i, y + j) according to the following equation (10). The intermediate interpolation value P _Mid is calculated, and the calculated intermediate interpolation value P _Mid is output to the weighted addition unit 64.

Based on the horizontal / vertical edge strength HV and the oblique edge strength RL input from the angle coefficient calculation unit 43 of the interpolation coefficient calculation unit 4, the weighting addition unit 64 performs the horizontal / vertical interpolation value _PHV according to the following equation (11). Then, the weighted addition of the diagonal interpolation value P _RL and the intermediate interpolation value P _Mid is performed to calculate the contour line area interpolation value P _line, and the calculated contour line area interpolation value P _line is output to the interpolation output selection unit 54.

As apparent from the above equation (11), the coefficient applied to the intermediate interpolation value P _Mid is always 1, whereas the weight of the horizontal / vertical interpolation value P _HV increases in proportion to the horizontal / vertical edge strength HV. The weight of the diagonal interpolation value _PRL increases in proportion to the diagonal edge strength RL. This is because, when the horizontal / vertical edge strength HV is large, that is, for a contour line close to horizontal / vertical, the contribution of the horizontal / vertical interpolation value P _HV is large, and at the same time, the contribution of the intermediate interpolation value P _Mid is small. Is meant to be. Further, when the oblique edge strength RL is large, that is, for a contour line close to 45 °, it means that the contribution of the oblique interpolation value P _RL is increased and at the same time, the contribution of the intermediate interpolation value P _Mid is reduced. is doing. Further, when both the horizontal / vertical edge strength HV and the oblique edge strength RL are small, that is, for the contour line having an angle close to the middle between the horizontal / vertical and the oblique 45 °, the contribution of the intermediate interpolation value P _Mid is large. Is meant to be.

  According to the present embodiment, it is possible to perform pattern classification of a pixel of interest based on a differential value, and to calculate an optimal interpolation value by an interpolation method corresponding to any one of a contour line area, a texture area, and a flat area. Become. Furthermore, in the contour region, a filter coefficient is calculated based on the classification code that is the determination result of the pattern classification, and interpolation processing is performed without referring to pixels having low correlation along the contour. A reduction in sharpness and jaggies can be reduced, and overshoot / undershoot of edges near the contour line can be prevented, and a high-quality output image can be obtained.

  Further, in the contour region, three types of interpolation values corresponding to the three directions are weighted and added based on the differential values, whereby interpolation processing corresponding to fine contour line angle changes can be performed. For example, for an arc-like contour line in which the contour angle gradually changes, the horizontal / vertical direction interpolation value, the intermediate direction interpolation value, and the diagonal 45 ° direction interpolation value are used as the change in the contour angle. Since the weighting is gradually changed according to the addition, it is possible to prevent an unnatural contour such as a seam due to a pattern change and to prevent the occurrence of jaggies and smoothly interpolate the contour line. Can be done.

  Furthermore, by interpolating by combining pattern classification, filter coefficient calculation, and weighting coefficient calculation based on the differential value, a high-quality output image can be obtained by a combination of relatively simple methods, and processing Speed and hardware scale can be reduced.

(Second Embodiment)
15 and 16 are diagrams for explaining the image processing apparatus according to the second embodiment of the present invention. FIG. 15 is a functional block diagram showing the configuration of the main part of the pattern classification determination unit. FIG. It is a figure for demonstrating operation | movement.

  The image processing apparatus according to the present embodiment is different from the image processing apparatus according to the first embodiment in the configuration of the pattern classification determination unit 3. Therefore, the same constituent elements as those in the first embodiment are denoted by the same reference numerals, description thereof is omitted, and only different points will be mainly described.

  In this embodiment, when performing pattern classification determination of a region of interest, not only the pixel of interest but also all pixels in the region of interest are subjected to pattern determination, and correction processing for pattern classification determination of the region of interest is performed based on the statistics. Do. For this reason, as shown in FIG. 15, the pattern classification determination unit 3 includes the memory 36 and the configuration in the case of the first embodiment having the positive / negative determination units 31, 32, 33, 34 and the pattern determination unit 35. A pattern correction unit 37 is provided.

  Here, the classification code determined by the pattern determination unit 35 is input to the memory 36 in the same manner as the processing described with reference to FIG. In this embodiment, the classification codes determined by the pattern determination unit 35 are stored in the memory 36 without being sequentially output as they are, and are delayed until the classification codes for all the pixels in the attention area are obtained.

  Thereafter, when the classification codes for all the pixels in the attention area are stored in the memory 36, the classification codes for all the pixels in the attention area are output to the pattern correction unit 37. The pattern correction unit 37 detects a singular point having a characteristic pattern from the classification code statistics for all the pixels in the region of interest, and corrects the pixel of interest classification code.

  For example, when an edge corner as shown in FIG. 16A exists in the attention area, a pixel for which a classification code indicating directions different from each other by 90 ° is determined in the attention area is detected. . Further, when the end point of a thin line as shown in FIG. 16B exists in the attention area, or when an isolated point as shown in FIG. 16C is included, directions different by 180 ° in the attention area A pixel for which a classification code indicating (opposite directions) is determined is detected. In addition, when an edge having a complicated shape as shown in FIG. 16D exists in the attention area, the classification code of each pixel in the attention area does not continue and changes in a complicated manner.

  Since it is desirable to perform smooth interpolation processing that retains the value of the pixel of interest in such a singular point region, the pattern correction unit 37 performs classification shown in FIG. 8 when a singular point is detected. The code is changed to 8, and the region pattern classification determination is corrected so as to perform an interpolation process that retains the structure of the pixel of interest similar to the texture region interpolation. In this way, the pattern correction unit 37 corrects the classification code of the region of interest, and outputs the corrected classification code to the interpolation coefficient calculation unit 4 and the interpolation calculation unit 5.

  According to the present embodiment, the same effect as the first embodiment described above can be obtained, and the singular point in the image is detected from the statistics of the classification codes for all the pixels in the attention area, and Since the pattern classification is corrected, the accuracy of pattern classification determination can be improved, and deterioration of image quality due to a decrease in contrast and a change in image structure can be more effectively prevented.

(Third embodiment)
17 and 18 are diagrams for explaining an image processing apparatus according to the third embodiment of the present invention. FIG. 17 is a functional block diagram showing a configuration of a main part of the interpolation coefficient calculation unit, and FIG. 18 is an interpolation. It is a functional block diagram which shows the structure of the principal part of a calculating part. In the present embodiment, the same components as those in the first embodiment are denoted by the same reference numerals, description thereof is omitted, and only different points will be mainly described.

  In the present embodiment, after the interpolation coefficient calculation unit 4 calculates three types of filter coefficients from the lookup table, each filter coefficient is weighted to calculate a combined filter coefficient, and the calculated combined filter coefficient is interpolated. It outputs to the part 5, and an interpolation value is calculated by convolution.

  Therefore, as shown in FIG. 17, the interpolation coefficient calculation unit 4 includes a weighting coefficient calculation unit 71 that calculates weights of a plurality of interpolation values, and an interpolation filter coefficient that calculates filter coefficients for calculating a plurality of interpolation values. The calculation unit 72 includes a filter synthesis unit 73 that calculates a synthesis filter coefficient by performing weighted addition of a plurality of filters based on the calculated weighting coefficient. The weighting coefficient calculation unit 71 includes an angle coefficient calculation unit 43, and the interpolation filter coefficient calculation unit 42 includes a horizontal / vertical direction interpolation filter coefficient LUT44, an oblique direction interpolation filter coefficient LUT45, and an intermediate direction interpolation filter coefficient LUT46. The filter composition unit 73 includes a weighted addition unit 74 and a memory 75. Here, the functions of the angle coefficient calculation unit 43, the horizontal / vertical direction interpolation filter coefficient LUT44, the diagonal direction interpolation filter coefficient LUT45, and the intermediate direction interpolation filter coefficient LUT46 are the same as those of the first embodiment shown in FIG.

  The horizontal direction differential value, vertical direction differential value, upper right direction differential value, and upper left direction differential value calculated by the differential operation unit 2 are input to the angle coefficient calculation unit 43, and the angle weighting coefficient calculated here is weighted addition. Input to the unit 74. The classification code calculated by the pattern classification determination unit 3 is input to the horizontal / vertical direction interpolation filter coefficient LUT44, the diagonal direction interpolation filter coefficient LUT45, and the intermediate direction interpolation filter coefficient LUT46 to calculate the respective interpolation filter coefficients. These calculated interpolation filter coefficients are input to the weighted addition unit 74.

  The weighting / adding unit 74, based on the horizontal / vertical edge strength HV and the diagonal edge strength RL calculated by the angle coefficient calculating unit 43 in accordance with the following equation (12), outputs the horizontal / vertical direction interpolation filter output from the lookup table. The combined filter coefficient filterSYN (i, j) is calculated by performing weighted addition processing of the coefficient filterHV (i, j), the diagonal direction interpolation filter coefficient filterRL (i, j), and the intermediate direction interpolation filter coefficient filterMid (i, j). Then, the calculation result is stored in the memory 75 and output to the interpolation calculation unit 5.

  As shown in FIG. 18, the interpolation calculation unit 5 includes an outline region interpolation calculation unit 51, a texture region interpolation calculation unit 52, a flat region interpolation calculation unit 53, and an interpolation output selection unit 54. Constitute. The input image is input to the contour region interpolation calculation unit 51, the texture region interpolation calculation unit 52, and the flat region interpolation calculation unit 53 via the region extraction unit 1, and is calculated by the interpolation coefficient calculation unit 4 shown in FIG. The contour interpolation filter coefficient is input to the contour region interpolation calculation unit 51. The classification code determined by the pattern classification determination unit 3 is input to the interpolation output selection unit 54.

  Based on the input classification code, the interpolation output selection unit 54 selects the processing of the contour region interpolation calculation unit 51 when the attention region is a contour region, and the texture region interpolation calculation unit when it is a texture region. 52, if the region is a flat region, the processing of the flat region interpolation calculation unit 53 is selected, the processing of the selected interpolation calculation unit is turned on, and the processing of the non-selected interpolation calculation unit is controlled OFF. The output of the selected interpolation processing unit is output as the final output image.

  Here, the interpolation output selection unit 54 may be installed before the contour region interpolation calculation unit 51, the texture region interpolation calculation unit 52, and the flat region interpolation calculation unit 53. In this case, the interpolation output selection unit 54 uses the outline region in the case where the attention region is the outline region based on the classification code determined by the pattern classification determination unit 3 as in the above-described form shown in FIG. The contour area interpolation value output from the interpolation calculation unit 51 is selected. When the texture area is a texture area, the texture area interpolation value output from the texture area interpolation calculation unit 52 is selected. The flat area interpolation value output from the calculation unit 53 is selected and output as the final output value.

  Here, the contour area interpolation calculation unit 51, when the interpolation output selection unit 54 determines that the region of interest is the contour line region, the combined filter coefficient filterSYN (calculated by the interpolation coefficient calculation unit 4 shown in FIG. i, j) is used to calculate an interpolated value by performing a convolution operation with the input pixel according to the following equation (13), and output this interpolated value as a final output value.

  A texture region interpolation calculation unit 52, when the interpolation output selection unit 54 determines that the region of interest is a texture region, saves a fine texture structure using a predefined interpolation filter, as in FIG. A texture region interpolation value is calculated by performing an interpolation process using an interpolation method (for example, bicubic interpolation), and the calculated texture region interpolation value is output as a final output value.

  Similarly, when the interpolation output selection unit 54 determines that the region of interest is a flat region, the flat region interpolation calculation unit 53 performs a simple and high-speed flat region using a predefined interpolation filter, as in FIG. The flat area interpolation value is calculated by a known interpolation method (for example, bilinear interpolation) that is not easily affected by the noise, and the calculated flat area interpolation value is output as the final output value.

  According to the present embodiment, the same effects as those of the first embodiment described above can be obtained, and in the processing of the interpolation coefficient calculation unit 4, the memory necessary for storing the filter coefficients can be unified. It is possible to reduce the scale in hardware. Further, since the contour line area interpolation processing unit 51 can calculate an interpolation value by a single convolution operation, the process can be simplified.

  In each of the above-described embodiments, the functions of the region extraction unit 1, the differential calculation unit 2, the pattern classification determination unit 3, the interpolation coefficient calculation unit 4, and the interpolation calculation unit 5 may be realized by a program using a computer. it can.

1 is a functional block diagram illustrating a main configuration of an entire image processing apparatus according to a first embodiment of the present invention. It is a flowchart which shows schematic operation | movement of 1st Embodiment. It is a functional block diagram which shows the principal part structure of the differential calculating part shown in FIG. It is a figure which shows an example of the filter coefficient used with the differential calculating part shown in FIG. It is a figure which shows the other example of the filter coefficient used with the differential calculating part shown in FIG. It is a functional block diagram which shows the principal part structure of the pattern classification determination part shown in FIG. It is a conceptual diagram which shows the direction of the vector which the differential value according to direction by the differential calculating part shown in FIG. 5 represents. It is a conceptual diagram which shows the classification | category of the outline direction pattern determined by the pattern determination part shown in FIG. It is a functional block diagram which shows the structure of the principal part of the interpolation coefficient calculation part shown in FIG. It is a graph which shows the change of the magnitude | size with respect to the angle of the outline of the differential value according to direction by the differential calculating part shown in FIG. It is a conceptual diagram which shows the position of the original pixel and interpolation pixel in an attention area. It is a conceptual diagram which shows the shape table of the interpolation filter coefficient in 1st Embodiment. It is a functional block diagram which shows the principal part structure of the interpolation calculating part shown in FIG. It is a functional block diagram which shows the principal part structure of the outline area | region interpolation calculation part shown in FIG. It is a functional block diagram which shows the principal part structure of the pattern classification determination part in the image processing apparatus which concerns on 2nd Embodiment of this invention. It is a conceptual diagram which shows the pattern of the pixel in the singular point explaining the operation | movement of the pattern classification determination part shown in FIG. It is a functional block diagram which shows the principal part structure of the interpolation coefficient calculation part in the image processing apparatus which concerns on 3rd Embodiment of this invention. It is a functional block diagram which shows the principal part structure of the interpolation calculating part in the image processing apparatus which concerns on 3rd Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Area extraction part 2 Differential calculation part 3 Pattern classification determination part 4 Interpolation coefficient calculation part 5 Interpolation calculation part 21 Horizontal differential value calculation part 22 Vertical differential value calculation part 23 Upper right differential value calculation part 24 Upper left differential value calculation part 25 Differential absolute value Sum calculation unit 31, 32, 33, 34 Positive / negative determination unit 35 Pattern determination unit 36 Memory 37 Pattern correction unit 41 Weighting coefficient calculation unit 42 Interpolation filter coefficient calculation unit 43 Angle coefficient calculation unit 44 Horizontal / vertical direction interpolation filter coefficient LUT
45 Diagonal interpolation filter coefficient LUT
46 Intermediate direction interpolation filter coefficient LUT
47, 48, 49 Memory 51 Outline region interpolation calculation unit 52 Texture region interpolation calculation unit 53 Flat region interpolation calculation unit 54 Interpolation output selection unit 61, 62, 63 Convolution calculation unit 64 Weighting addition unit 71 Weighting coefficient calculation unit 72 Interpolation filter Coefficient calculation unit 73 Filter synthesis unit 74 Weighted addition unit 75 Memory

Claims (8)

In an image processing apparatus for enlarging an input image,
A region extraction unit that extracts a region of interest including a pixel of interest from the input image;
A differential operation unit for calculating a differential value of a pixel in the region of interest;
A pattern classification determination unit that determines a region pattern in the region of interest based on the differential value;
An interpolation coefficient calculation unit that calculates an interpolation coefficient for calculating a pixel value of an interpolation pixel in the region of interest based on the region pattern and the differential value;
An interpolation calculation unit that calculates a pixel value of the interpolation pixel based on the interpolation coefficient;
An image processing apparatus comprising: The differential operation unit includes a plurality of differential filters that calculate differential values in at least four directions, and a differential absolute value that calculates a differential absolute value sum that is an absolute value sum of the plurality of differential values calculated by the plurality of differential filters. A sum calculator,
The image processing apparatus according to claim 1, wherein the pattern classification determination unit determines a region pattern in the region of interest based on the plurality of differential values and the differential absolute value sum.   The image processing apparatus according to claim 2, wherein the differential calculation unit calculates the plurality of differential values and the sum of differential absolute values for the target pixel. The differential calculation unit calculates the plurality of differential values and the differential absolute value sum for all pixels in the region of interest,
The pattern classification determination unit determines a region pattern in the region of interest based on statistics of the region pattern determined based on the plurality of corresponding differential values and the sum of differential absolute values for all the pixels. The image processing apparatus according to claim 2.   The interpolation coefficient calculating unit calculates a plurality of directional interpolation filter coefficients based on the classification code as the interpolation coefficient, and calculates an angle weighting coefficient for each of a plurality of directions based on the plurality of differential values. The image processing apparatus according to claim 2, wherein the image processing apparatus is characterized.   The interpolation calculation unit calculates a plurality of direction interpolation values in the region of interest based on the direction interpolation filter coefficient, weights the calculated plurality of direction interpolation values based on the angle weighting coefficient, and The image processing apparatus according to claim 5, further comprising an outline area interpolation calculation unit that calculates a pixel value of a target pixel as an outline area interpolation value. In enlarging the input image,
Extracting a region of interest including a pixel of interest from the input image;
Calculating a differential value of a pixel within the extracted region of interest;
Determining a region pattern in the region of interest based on the calculated differential value;
Calculating an interpolation coefficient for calculating a pixel value of an interpolation pixel in the region of interest based on the determined region pattern and the calculated differential value;
Calculating a pixel value of the interpolation pixel based on the calculated interpolation coefficient;
An image processing method comprising: On the computer,
A function for extracting a region of interest including a pixel of interest from an input image;
A function for calculating the differential value of the pixel within the extracted region of interest;
A function of determining a region pattern in the region of interest based on the calculated differential value;
A function of calculating an interpolation coefficient for calculating a pixel value of an interpolation pixel in the region of interest based on the determined region pattern and the calculated differential value;
A function of calculating a pixel value of the interpolation pixel based on the calculated interpolation coefficient and enlarging the input image;
An image processing program characterized by realizing the above.

Images