JPH11284958A - Image interpolation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To set an accurate interpolation axis with a few reference pixels in the image method that interpolates other field image with one field image in a video signal of the interlace scanning system. SOLUTION: In the case of evaluating correlation of pixel data in an axial direction obtained by tying point-symmetry pixels around an interpolated pixel, the evaluation of the correlation of the pixel data in the axial direction obtained by tying the point-symmetry pixels is enhanced by referencing not only the data of the point-symmetry pixels but also pixel data of adjacent pixels deviated from the point-symmetry pixels by one pixel in a horizontal direction so as to enhance a probability where an optimum interpolation axis is obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image interpolation method for interpolating one field image in a video signal of an interlaced scanning system with another field image.

[0002]

2. Description of the Related Art In a video signal of an interlaced scanning system, when a signal for one frame is extracted from a moving image signal and displayed or printed as a still image, a field image of a fast moving portion has a comb-like shape. It looks blurred. Therefore, in order to correct the blur, there is known an image interpolation method in which one field signal is generated by calculation from the other field signal to obtain a frame image without blur.

FIG. 3 is an explanatory diagram for explaining an example of a conventional image interpolation method. Here, when the pixel X is to be interpolated from the i-th line and the (i + 1) -th line of the field (odd field) serving as the source of the interpolation calculation to the interpolation line of the even-numbered field between them, first, the i-th line or Attention is paid to an adjacent pixel existing on the ( _{i + 1} ) th line and constituted by a total of 10 pixels (A _{i to} E _i and A _{i + 1 to} E _{i + 1} ) adjacent to the pixel X. as two adjacent pixels (e.g. a _i and E _{i + 1} or B _i and D _{i + 1,} etc.) point-symmetrical pixel forming a husband these points symmetrical pixel s connecting five axial (first to fifth interpolation Evaluate the correlation of the pixel data to the (axis).

[0004] Correlation of pixel data, for example, luminance data of the adjacent pixels that exist in the i-th line and (A _i to E _i) i
The luminance data of the adjacent pixel (A
_{i + 1 to} E _{i + 1} ). When the correlation evaluation indices in the first to fifth interpolation axis directions are Δ1 to Δ5, respectively, Δ1 is | A _i −E _{i + 1} | and Δ2 are | B _i −D
_{i +1} |, Δ3 is | C _i −C _{i + 1} |, Δ4 is | D _i −B
_{i +} 1 |, _Δ5 is | is calculated by | E _i -A _{i + 1.} Then, the values of the correlation evaluation indices Δ1 to Δ5 are compared, and the interpolation axis H is determined by the axis having the smallest value, and the average value of the pixel data of the two point-symmetric pixels on the interpolation axis H is used to determine the value of the pixel X. Interpolated data was obtained.

[0005] FIG. 4 shows the relationship between the i-th line and the i-th line based on the luminance data of the nine pixels on the i-th line and the nine pixels on the (i + 1) -th line using the conventional image interpolation method.
An example is shown in which interpolation data (Vh to Zh) of five pixels (pixels V to Z) in an interpolation line between the (+1) th line is obtained. Here, as shown in FIG. 4A, the luminance data of 9 pixels in the i-th line is 2, 2 from the left,
3, 7, 10, 7, 3, 3, 0, 0, and luminance data of 9 pixels in the (i + 1) th line are 0,
2, 4, 8, 10, 8, 3, 2, and 0.

When the values of the correlation evaluation indices Δ1 to Δ5 of the five pixels (pixels V to Z) in the respective axes in the interpolation line are compared with each other, first, for the pixel V, the correlation evaluation indices Δ1 to Δ5
Since the value of Δ5 is 8, 5, 3, 8, and 7, respectively, the third interpolation axis having the minimum value is determined as the interpolation axis H.
The fifth interpolation axis is set as the pixel W, the fifth interpolation axis is set as the pixel X, the second interpolation axis is set as the pixel Y, and the third interpolation axis is set as the interpolation axis H as the pixel Z.

[0007] Since the pixels V to Z obtain pixel data from the average value of the point symmetric pixels on the interpolation axis H determined respectively, the interpolation data Vh of the pixel V is calculated based on the point symmetric pixels on the third interpolation axis. 4. Average of luminance data 7 and 4
5, and similarly, the interpolation data Wh of the pixel W is 2.
5, the interpolation data Xh of the pixel X is 3.5, the interpolation data Yh of the pixel Y is 5.0, and the interpolation data Zh of the pixel Z is 3.0.

[0008]

FIG. 4B shows the interpolation data of the pixels V to Z in the interpolation line obtained in this way, the luminance data of 9 pixels in the i-th line,
It is a graph which each shows the luminance data of 9 pixels in the (i + 1) th line.

As shown in this graph, the luminance data of the i-th line and the luminance data of the (i + 1) -th line are
The correlation is high, and peaks (peaks) of the luminance data are generated. However, similar peaks (peaks) are not generated in the interpolation data of the pixels V to Z, but rather are valleys, which indicates that appropriate interpolation is not performed.

Also, comparing the luminance data of the i-th line with the luminance data of the (i + 1) -th line in FIG. Brightness data are lined up,
The respective interpolation axes H of the pixels V to Z should be originally the second interpolation axes or the third interpolation axes. However, for the pixels W and X, the fifth interpolation axes having completely different directions are interpolated. H.

As described above, in the conventional image interpolation method, the correlation of pixel data is evaluated only in the directions of the first to fifth interpolation axes. Alternatively, even if there is a pixel having a high correlation between adjacent interpolation axes having a relatively large angle between the interpolation axes, such as between the third interpolation axis and the fourth interpolation axis, this is detected. There is a high possibility that the wrong interpolation axis will be set because of the inability to perform the operation, and an undesirable interpolation process is often performed.

[0012]

In order to solve the above-mentioned problems, an image interpolation method according to the present invention comprises: a first and a third image signal sequence linearly arranged adjacent to each other; Second
And extracting a plurality of pairs of point-symmetric pixels in the first and third image signal trains that are point-symmetric about the interpolated pixels in the second image signal train. Calculating the correlation evaluation index indicating the correlation of the pixel data in the axial direction connecting each of the extracted pairs of point-symmetric pixels with respect to each axial direction, as a reference pixel, and Determining an interpolation axis by comparing a correlation evaluation index for each axis direction, and obtaining interpolation data at the interpolation pixel based on the pixel data of the point symmetric pixel on the determined interpolation axis, The correlation evaluation index in each axial direction is a first pixel data correlation index indicating the correlation between pixel data of the pair of extracted point-symmetric pixels, and one of the pair of point-symmetric pixels. A second pixel data correlation index indicating the correlation between pixel data and pixel data between the other pixel of the pair of point-symmetric pixels and an adjacent reference pixel, and the one pixel of the pair of point-symmetric pixels; By comparing a third pixel data correlation index indicating the correlation of pixel data between the adjacent reference pixel and the other pixel of the pair of point-symmetric pixels,
It is determined by the minimum value of the first to third pixel data correlation indices, and further interpolated based on a calculation result obtained by multiplying the correlation evaluation indices in each of the axial directions by a predetermined weighted coefficient. An axis is determined, and further, when the correlation evaluation index for each of the axis directions has the same value, a point at which the distance to the interpolation pixel is the shortest among the axis directions having the same value. An interpolation axis is determined in an axis direction connecting symmetric pixels.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The image interpolation method according to the present invention evaluates the correlation of pixel data in the axial direction connecting point-symmetric pixels. The evaluation accuracy of the correlation of the pixel data in the axial direction connecting the point-symmetric pixels without increasing the number of reference pixels as a whole by also referring to the pixel data of the adjacent pixels shifted by one pixel in the horizontal direction. And an optimum interpolation axis is obtained.

FIG. 1 is an explanatory diagram for explaining an image interpolation method according to an embodiment of the present invention. In the image interpolation method according to the present embodiment, similarly to the conventional image interpolation method, when interpolating the pixel X on the interpolation line existing between the i-th line and the (i + 1) -th line, first, the i-th A total of 1 existing on the line or the (i + 1) th line and adjacent to the pixel X
Attention is paid to an adjacent pixel composed of 0 pixels (A _{i to} E _i and A _{i + 1 to} E _{i + 1} ).

Then, two adjacent pixels (for example, _Ai and _{Ei + 1} or _Bi and _{Di + 1} ) which are point-symmetric about the pixel X.
Is a point symmetric pixel, and 5
The correlation of the pixel data in one axis direction (first to fifth interpolation axes) is evaluated as follows.

In the image interpolation method according to the present invention, FIG.
As shown in (a), in order to evaluate the correlation of the pixel data in the direction of the first interpolation axis, the difference | A _i −E between the luminance data of the point symmetric pixels A _i and E _{i + 1} is evaluated. A _i and D _{i + 1} , including not only _{i + 1} |
Is the difference of the luminance data of _{| A i -D i + 1 |} , and B _i
And is the difference of the luminance data of the _{E i + 1 | B i -E} i + 1 | also computed.

Then, | A _i −D _{i + 1} | is Δ1a, | A _i
−E _{i + 1} | is Δ1b and | B _i −E _{i + 1} | is Δ1c.
The correlation evaluation index Δ1p on the first interpolation axis is obtained from the minimum value of Δ1a, Δ1b, and Δ1c. Similarly, FIG.
Δ2a, Δ2b, Δ2c shown in FIG. 1, Δ4 shown in FIG.
a, Δ4b, Δ4c, Δ5a, 5b shown in FIG.
.DELTA.5c are calculated respectively, and the second, fourth,.
Δ2p, Δ4p, Δ which are correlation evaluation indices on the fifth interpolation axis
Obtain a value of 5p.

Regarding the value of Δ3 indicating the correlation of the pixel data in the direction of the third interpolation axis, the calculation result of the point symmetric pixel and the adjacent pixel shifted by one pixel in the horizontal direction is already Δ2.
a, Δ2c, Δ4a, Δ4c as the second interpolation axis or the fourth
Since it is used to calculate each correlation evaluation index on the interpolation axis,
Is the difference of the luminance data of the C _i and _{C i + 1 | C i -C} i + 1 | directly used. Then, the correlation evaluation index Δ1 in each axial direction
The values of p, Δ2p, Δ3, Δ4p, and Δ5p are compared, and the interpolation axis H is determined based on the axis having the smallest value.
The interpolation data at the pixel X is obtained from the average value of the pixel data of the two point symmetric pixels in.

That is, Δ1b, Δ2b, Δ3, Δ4b, Δ1b, Δ2b, Δ3b
In addition to calculating the value of Δ5b, Δna and Δnc from two axial directions approaching these axes but having different angles, respectively.
(N is any one of 1, 2, 4, and 5), and the first
Regarding the value of the correlation evaluation index Δ1p in the direction of the interpolation axis, the minimum value of Δ1a to Δ1c, the value of the correlation evaluation index Δ2p in the direction of the second interpolation axis, the minimum value of Δ2a to Δ2c, the third value
Regarding the value of the correlation evaluation index Δ3 in the direction of the interpolation axis, | C _i −
C i _{+ 1} |, the minimum value of Δ4a~Δ4c regarding the value of the correlation evaluation index Δ4p the fourth interpolation axis, with respect to the value of the correlation evaluation index Δ5p fifth interpolation axis direction and the minimum value of Deruta5a～deruta5c.

FIG. 2 is a diagram showing the relationship between 9 pixels existing on the i-th line and i + 1 using the image interpolation method according to the embodiment of the present invention.
According to the luminance data of 9 pixels existing in the i-th line, i
Interpolation data (Vh-Z) of 5 pixels (pixels V-Z) in the interpolation line between the i-th line and the (i + 1) -th line
h) shows the example obtained.

Here, as shown in FIG. 2A, the luminance data of the nine pixels in the i-th line is the same as the value of the luminance data shown in FIG. ,
10, 7, 3, 3, 0, 0, and luminance data of 9 pixels in the (i + 1) th line are 0, 2, 4,.
8, 10, 8, 3, 2, and 0.

Correlation evaluation indices .DELTA.1p, .DELTA.2p, .DELTA.3, .DELTA. For each axis direction of five pixels (pixels V to Z) in the interpolation line.
In order to obtain the values of 4p and Δ5p, for each pixel V to Z
1a to Δ1c, Δ2a to Δ2c, Δ3, Δ4a to Δ4
c, Δ5a to Δ5c are calculated. First, regarding the pixel V, since the values of Δ1p are 6, 8, and 7, respectively, Δ1a to Δ1c are 6, 8, and 7, which are the minimum values thereof, and the value of Δ2p is Δ2a to Δ2c.
Are 1, 5, and 1, respectively, the minimum values of 1, 3 are the difference between 7 and 4, and the values of Δ4p are Δ4a to Δ4
Since c is 5, 8, and 6, respectively, its minimum value is 5, Δ5.
The value of p is 5, which is the minimum value, because Δ5a to Δ5c are 10, 7, and 5, respectively. And Δ1p, Δ2p, Δ
Since the value of Δ2p is minimum when comparing the values of 3, Δ4p, and Δ5p, the second interpolation axis is determined as the interpolation axis H.

Similarly, for the pixel W, Δ1p,
The values of Δ2p, Δ3, Δ4p, and Δ5p are 1, 0, 2,
1, 1 so that the second interpolation axis and the pixel X are Δ1p,
The values of Δ2p, Δ3, Δ4p, and Δ5p are 1, 0, 3,
1, 1 so that the second interpolation axis and the pixel Y are Δ1p,
The values of Δ2p, Δ3, Δ4p, Δ5p are 5, 0, 5,
Because of 5, 5, the second interpolation axis, pixel Z, Δ1p,
The values of Δ2p, Δ3, Δ4p, and Δ5p are 3, 0, 0,
Since they are 3 and 8, they are determined as the second and third interpolation axes.

Here, since the value of the correlation evaluation index Δ2p for the second interpolation axis and the value of the correlation evaluation index Δ3 for the third interpolation axis are both 0 and the same value, the interpolation axis H
Was set as both the second interpolation axis and the third interpolation axis.

Then, since the pixels V to Y obtain pixel data based on the minimum value of the point symmetric pixel on the interpolation axis H determined respectively, the interpolation data Vh of the pixel V is calculated based on the point symmetric pixel on the second interpolation axis. 4. Average of luminance data 3 and 8
5, and similarly, the interpolation data Wh of the pixel W becomes 8.
5. The interpolation data Xh of the pixel X is 9.0, and the interpolation data Yh of the pixel Y is 5.0. Also, the interpolation data Z of the pixel Z
h is luminance data of a point symmetric pixel on the second interpolation axis, 3
And 2, and the average of all the values 3 and 3, which are the luminance data of the point symmetric pixels on the third interpolation axis, is 2.75.

FIG. 2B shows the interpolation data of the pixels V to Z on the interpolation line, the luminance data of 9 pixels on the i-th line, and the luminance data of 9 pixels on the (i + 1) -th line. It is a graph shown respectively.

As shown in this graph, the interpolated data of the pixels V to Z has the same correlation as the luminance data of the i-th line or the luminance data of the (i + 1) -th line at which the peak (peak) of the luminance data occurs. The result is high interpolation data, indicating that appropriate interpolation has been performed.

As described above, for the direction of the third interpolation axis, the value of the correlation evaluation index Δ3 is used as it is, and the correlation evaluation index Δnp for each axis direction excluding the third interpolation axis is represented by Δnaｎ.
According to the image interpolation method according to the embodiment of the present invention, which is determined by the minimum value of Δnc (n is any one of 1, 2, 4, and 5), only reference pixel data of 10 pixels is used. However, since the interpolation axis H is determined to be one of the first to fifth interpolation axes after evaluating the correlation of the pixel data for 13 pixel pairs, the accuracy of determining the interpolation axis H is improved. be able to.

That is, according to the image interpolation method according to the embodiment of the present invention, for the pixel W shown in FIG.
Is 1, but since the value of Δ2c is a smaller value of 0, the fifth interpolation axis is not the interpolation axis H, and
The second interpolation axis having the minimum value among the values of p, Δ2p, Δ3, Δ4p, and Δ5p could be set as the interpolation axis H.

In the above description, when the value of the correlation evaluation index in each axis direction becomes the same, a plurality of interpolation axes H are provided, and the average of the luminance data of all the point symmetric pixels on the interpolation axis H is obtained. Although the example in which the interpolation data is obtained by the above is shown, it is also possible to assign a priority to the first to fifth interpolation axes and determine one interpolation axis.

In general, if the correlation between pixel data in the axial direction is the same, generating interpolated data using a pixel closer to the pixel to be interpolated has higher continuity with surrounding pixels. Desirable interpolation without a feeling of strangeness can often be performed. Therefore, when the values of the correlation evaluation indices are the same, for example, the third interpolation axis> the second and fourth interpolation axes, depending on the distance between the pixel to be interpolated and the point symmetric pixel on the first to fifth interpolation axes.
By setting priorities such as interpolation axis> first and fifth interpolation axes, it is possible to increase the interpolation accuracy.

Instead of directly comparing the correlation evaluation indices obtained by the calculation, the interpolation axis may be determined by a value obtained by multiplying each correlation evaluation index by a coefficient weighted based on the above-mentioned priority. . That is, 1 for the correlation evaluation index of the third interpolation axis, 2 for the correlation evaluation index of the second and fourth interpolation axes, and 3 for the correlation evaluation index of the first and fifth interpolation axes. The interpolation axis may be determined by comparing the multiplied values.

Further, in the above embodiment, an example was shown in which the interpolation data of the pixel to be interpolated was obtained by the average value of the pixel data of the two point symmetric pixels on the interpolation axis H. The pixel data of the point symmetrical pixel is 0.
7 and the pixel data of the point symmetric pixel on the (i + 1) -th line may be multiplied by a value of 0.3, and the interpolation data may be obtained by adding the respective values.

[0034]

As described above, according to the image interpolation method according to the present invention, not only the pixel data of the point symmetric pixel but also the point symmetric By referring to the pixel data of the adjacent pixel which is shifted by one pixel with respect to the pixel, the evaluation of the correlation of the pixel data in the axial direction connecting the point symmetric pixels can be improved, and the number of reference pixels as a whole can be reduced. This has the effect of increasing the probability of obtaining an appropriate interpolation axis without increasing it.

When the correlation evaluation indices in the respective axial directions have the same value, a point symmetric pixel having the shortest distance between the interpolation pixel and the point symmetric pixel among the axis directions having the same value is connected. By determining the interpolation axis in the axial direction, the probability of obtaining an appropriate interpolation axis can be further increased.

Further, by determining the interpolation axis by a value obtained by multiplying the correlation evaluation index in each axis direction by a predetermined weighted coefficient, the probability of obtaining an appropriate interpolation axis can be further increased.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram for explaining an image interpolation method according to an embodiment of the present invention.

FIG. 2 is a diagram for explaining interpolation data obtained by performing an interpolation operation using the image interpolation method according to the embodiment of the present invention.

FIG. 3 is an explanatory diagram for explaining a conventional image interpolation method.

FIG. 4 is a diagram for explaining interpolation data obtained by performing an interpolation operation using a conventional image interpolation method.

[Explanation of symbols]

Δ1, Δ1p: correlation evaluation index of pixel data in the first interpolation axis direction Δ2, Δ2p: correlation evaluation index of pixel data in the second interpolation axis direction Δ3: correlation evaluation index of pixel data in the third interpolation axis direction Δ4 , Δ4p... Correlation evaluation index of pixel data in the fourth interpolation axis direction Δ5, Δ5p... Correlation evaluation index of pixel data in the fifth interpolation axis direction

Claims (3)

[Claims] 1. An image interpolation method for generating a second image signal sequence between first and third image signal sequences linearly arranged adjacent to each other, wherein the second image signal sequence A plurality of pairs of point symmetric pixels in the first and third image signal strings that are point symmetric with respect to the interpolation pixel in the above are extracted as reference pixels, and the extracted point symmetric pixels of each pair are connected to each other. Calculating a correlation evaluation index indicating the correlation of pixel data in the axial direction for each axis direction, and determining an interpolation axis by comparing the calculated correlation evaluation index for each axis direction; Obtaining interpolation data at the interpolation pixel based on the pixel data of the point symmetric pixel on the determined interpolation axis, wherein the correlation evaluation index in each axis direction is between the pair of extracted point symmetric pixels. Painting A first pixel data correlation index indicating the correlation of the raw data, and a pixel data of one of the pair of point-symmetric pixels and another of the pair of point-symmetric pixels and an adjacent reference pixel. A second pixel data correlation index indicating a correlation, and a correlation between pixel data of a reference pixel adjacent to the one pixel of the pair of point symmetric pixels and the other pixel of the pair of point symmetric pixels. A third pixel data correlation index indicating a characteristic of the image, and determining the minimum value of the first to third pixel data correlation indexes. 2. The image interpolation apparatus according to claim 1, wherein an interpolation axis is determined based on a calculation result obtained by multiplying the correlation evaluation index in each of the axial directions by a predetermined weighted coefficient. Method. 3. When the correlation evaluation indices with respect to the respective axial directions have the same value, among the axial directions having the same value, the axial direction connecting the point symmetric pixels having the shortest distance from the interpolation pixel. 3. An interpolation axis is determined.
The described image interpolation method.