KR20080040292A - Interpolation method for deinterlacing - Google Patents

Abstract

An interpolation method for deinterlacing is provided to improve picture quality by preventing a corner portion at a right-angle pattern from being broken, and obtain a perfect progressive image from an interlace type image source. Correlations among pixels in a window are obtained by symmetrically moving the window with respect to a certain region of upper and lower line image data of pixels to be interpolated(S51). The respective correlations are analyzed to obtain an interpolation direction(S52). If the interpolation direction indicates a slant line directional edge, pixels adjacent to the corresponding window among the upper and lower line image data are inspected to check the presence of the slant line directional edge again(S53). If the slant line directional edge exists, the slant line direction is determined as an interpolation direction, or otherwise, a vertical direction is determined as the interpolation direction(S54,S56). The interpolation target pixels are interpolated by using the determined interpolation direction(S57).

Description

Interpolation Method for Deinterlacing

1 is an overview of an interlaced video structure;

2 is an overview that is converted to a progressive image,

3 is an example of a method of determining an interpolation direction using a window;

4 illustrates an example in which an interpolation direction is incorrectly determined when a right angle pattern appears;

5 is an embodiment of an interpolation method according to the present invention;

6 is an example of obtaining a correlation according to a window;

7 is another example of obtaining a correlation according to a window;

8 is an embodiment of a process for rechecking an oblique edge;

9 is an example in which interpolation is performed through correction of an interpolation direction;

10 is an example of a state in which accurate interpolation is made in accordance with the present invention.

* Explanation of symbols for main parts of the drawings

21: odd field image 22: even field image

23: Deinterlacer 24,25: Progressive video

z: interpolation target pixels w1, w2: window

l1, l2, r2, r2: window adjacent pixels

The present invention relates to an interpolation method for deinterlacing. In particular, when a pixel to be interpolated has an oblique edge in the process of converting an interlaced input image into a progressive image, the upper and lower line image data is referred to. In fact, it is again necessary to check that the oblique edge is correct, so that the oblique edge can be detected with higher reliability.

Attempts have been made to realize better image quality in various broadcast receivers such as digital TVs. In particular, a deinterlacer that converts interlaced video into progressive video plays a very important role in realizing a clear picture quality.

The interlacing method refers to a method of displaying only one half of a horizontal line in one image frame when displaying one image. That is, as shown in FIG. 1, a screen forming one frame is divided into an odd field and an even field to be implemented as a screen of two fields. Therefore, if the broadcast video is composed of 30 frames per second, 30 odd fields and 30 even fields are included, and the odd fields and even fields are alternately displayed every 1/60 seconds. On the other hand, the progressive method (scan) is a method of displaying the entire frame at a time, it is possible to implement a superior image quality than the interlace method.

Such an interlaced image may be converted into a progressive image through a conversion device such as a deinterlacer.

Referring to FIG. 2, the odd field image 21 and the even field image 22 are converted through the deinterlacer 23 and converted into progressive images 24 and 25, respectively. Pixels marked with 'z' in the converted progressive images 24 and 25 are interpolated pixels to make a progressive image. That is, the deinterlacer 23 artificially generates the scan line information not included in the current field by using the interlaced field image. Generally, the deinterlacer 23 uses the image data of the upper and lower lines. Perform interpolation.

The method of converting interlaced video signals into progressive video signals includes line repetition, inter-field interpolation without motion compensation, and intra-field interpolation. Etc. are suggested in various ways.

Here, the line repetition method is a method of obtaining a frame by simply repeating the line information of the current field. Since each line is repeatedly displayed, the implementation is simple, but since the new frame is formed by simple repetition of the same line, interpolation is performed. There is a disadvantage in that the image quality is sharply lowered. The interfield interpolation method without motion is a method of embedding image data obtained by averaging the previous field and the next field line between the lines of the current field. There is a disadvantage of this deterioration.

In-field interpolation is a method that completes a frame using the average of data between two lines in the area between two lines as the interpolation value. In the case of a motionless image, the step-edge phenomenon is prominent, and the flickering of the screen is aggravated and the image quality deteriorates.

On the other hand, among the methods proposed to solve the disadvantages of the interpolation method and to achieve better image quality, the method of the current interpolation target pixel, such as the method disclosed in the document (published No. 10-2006-104817) There is a method that analyzes the upper and lower lines and performs interpolation in this direction when oblique edges are detected, resulting in better deinterlacing performance.

Referring to FIG. 3, the concept of windows 31 and 32 symmetrically moving in a predetermined area of the upper line Line i-1 and the lower line Line i + 1 of the line Interpolation line i which is currently interpolated is introduced. Block matching in the window unit is performed to determine whether there is an edge in a specific direction. Here, an example is shown in which the window is set to a size including five pixels, and the window moving area is set to eight pixels left and right of the pixel z that is currently interpolated. The direction of the edge is determined by examining the magnitude of the block matching value. If there is no oblique edge, vertical average interpolation is performed, and if there is an oblique edge, interpolation is performed in that direction, thereby greatly improving the quality of a progressive image produced by deinterlacing.

Assuming that the window 31 of the upper line and the window 32 of the lower line represent the interpolation direction with respect to the pixel z to be interpolated at present, one example of determining the interpolation value is the J + 1 th pixel value of the upper line ( After the middle pixel value of the window 31) and the J-1th pixel value (the middle pixel value of the window 32) of the lower line are added, the value divided by two can be interpolated to the value of the pixel z.

However, as shown in FIG. 4A, when the original image has a right edge (black portion in the lower line), when the above method is used, the current interpolation pixel c is diagonally edged by the window 31 and the window 32. Will be judged to have That is, the pixel c, which is currently the interpolation target, actually has a right angled edge, and is incorrectly determined as an oblique edge, so that the interpolation in the diagonal direction is performed as shown in FIG. 4B. Assuming that each pixel of the upper line and the lower line shown in FIG. 4B has white and black values as shown, the value of the interpolation target pixel c is interpolated to white.

When the interpolation is performed, the corners of the rectangular edges deteriorate significantly unlike the originals. Such deterioration is prominent in images containing characters or patterns with many rectangular edges, and causes a significant deterioration in image quality.

Accordingly, the present invention has been made to solve the above problems, and when the interlaced input image is converted into a progressive image, the existence of the oblique edge is rechecked by using the original image data. It is an object of the present invention to provide an interpolation method for deinterlacing that can detect a fragrance edge with higher reliability.

In order to achieve the above object, an interpolation method for deinterlacing according to the present invention includes: obtaining correlation between pixels in a window while symmetrically moving a window with respect to a predetermined area of upper and lower line image data of an interpolation target pixel; Analyzing the correlation to obtain an interpolation direction; If the interpolation direction is an oblique direction, re-checking whether the diagonal edge is correct by irradiating pixels around the window; Determining the diagonal direction as an interpolation direction if the diagonal direction is correct, and otherwise determining the interpolation direction as a vertical direction; And performing interpolation of the interpolation target pixel using the determined interpolation direction.

The determining of the interpolation direction may be configured to irradiate a pixel adjacent to a corresponding window among the upper and lower line image data to determine whether the diagonal edge is correct.

In this case, when the difference between the left pixel values adjacent to the window is within a predetermined range and the difference between the right pixel values respectively adjacent to the window is within a predetermined range, the diagonal edge may be determined to be correct.

The correlation may include a value obtained by subtracting pixel values of corresponding pixels in the mirrored window pair, and adding up the absolute values of the subtracted values. At this time, it may be determined that the correlation is larger as the calculated value is smaller.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Hereinafter, the shape (eg circle, rectangle) or color (eg black, white, gray) of the pixels shown in each drawing does not have any special meaning unless specifically mentioned, and merely helps the intuitive understanding of the description. For the sake of clarity. However, when the shape of the pixel has a special meaning such as a pixel value, it will be described separately.

An embodiment of an interpolation method for deinterlacing according to the present invention will be described with reference to FIG. 5.

First, a correlation between pixels in a window is obtained while symmetrically moving a window with respect to predetermined areas of upper and lower line image data of an interpolation target pixel (S51).

The predetermined area of the upper and lower line image data may be arbitrarily set as needed. Hereinafter, an example in which eight pixels of left and right are targeted based on the pixel position J as the current interpolation target will be described. Let's do it. In addition, the window is a concept applied to the upper line and the lower line, respectively, and means a set number of pixels. The window is moved symmetrically in the upper line and the lower line, and the number of pixels included in the window can be arbitrarily set as necessary, but will be described below as an example of a window including five pixels.

The correlation may be set to a value obtained by subtracting pixel values of corresponding pixels in the mirrored window pairs and adding up the absolute values of the subtracted values.

Referring to FIG. 6, the window w1 of the current upper line Line i-1 includes the J-1 th pixel to the J + 3 th pixel, and the window w2 of the lower line Line i + 1 is the J-3 th time. Pixel to J + 1 th pixel. The correlation at this time may be obtained by adding up the absolute values of the corresponding pixel value differences in the window w1 and the window w2, respectively. If this is expressed as an equation, Equation 1 below.

Here, P (x, y) means the pixel value of the pixel Pixel located in the y-th column of the x-th line. That is, the absolute value of the value obtained by subtracting the (J-3) th pixel value of the line i + 1 from the (J-1) th pixel value of the line i-1, and the (J-1 + 1) th pixel of the line i-1. Absolute value of the value minus the (J-3 + 1) th pixel of line i + 1, the (J-3 + of the line i + 1 the (J-1 + 2) th pixel value of line i-1 2) The absolute value of the value minus the first pixel value, the absolute value of the value minus the (J-3 + 3) th pixel value of line i + 1 from the (J-1 + 3) th pixel value of line i-1, Correlation 1 is the sum of the absolute values of the (J-1 + 4) th pixel value of the line i-1 minus the (J-3 + 4) th pixel value of the line i + 1.

Meanwhile, the window w1 may move one pixel to the right, and thus the window w2 may move one pixel to the left. In this way, as shown in Fig. 7, the window w1 of the upper line includes the J + 4th pixel and the J + 8th pixel, and the window w2 of the lower line has the J-8th pixel to J-4. The second pixel. In this case, the correlation is expressed by the following equation.

That is, the absolute value of the value obtained by subtracting the (J-8) th pixel value of the line i + 1 from the (J + 4) th pixel value of the line i-1, and the (J + 4 + 1) th pixel of the line i-1. Absolute value of the value minus the (J-8 + 1) th pixel of line i + 1, the (J-8 + of the line i + 1 of the (J + 4 + 2) th pixel value of line i-1 2) absolute value of the value minus the th pixel value, absolute value of the (J + 4 + 3) th pixel value of the line i + 1 minus the (J-8 + 3) th pixel value of the line i-1, The correlation 6 is the sum of the absolute values of the (J + 4 + 4) th pixel value of the line i-1 minus the (J-8 + 4) th pixel value of the line i + 1.

In this way, a total of six correlations can be obtained when the window w1 moves to the right and the window w2 moves to the left.

In addition, there are times when both the window w1 and the window w2 include the J-2nd pixel to the J + 2th pixel. The correlation at this time may be expressed as Equation 3 below.

In addition, since the window w1 can be moved to the left six times (in this case, the window w2 is moved to the right), six correlations can be obtained as described above.

In this way, a total of 13 correlations can be obtained, and the correlations obtained by moving the correlations by one pixel from the window position shown in FIG. 6 to the window position shown in FIG. The correlations obtained by moving the window position of the upper line by one pixel from the vertical position will be referred to as correlations 8 to 13, respectively.

For example, correlation 8 and correlation 13 may be expressed as Equations 4 and 5, respectively.

Then, each correlation obtained in step S51 is analyzed to obtain an interpolation direction (S52). For example, the interpolation direction may be obtained by determining which correlation has the smallest value among a total of 13 correlations. Since the correlation is calculated by block matching (e.g. subtracting each pixel value and then summing the absolute values) for the window of the upper line and the window of the lower line, the value for a similar set of pixels Because it will come out the smallest.

Here, the smallest correlation among correlations 1 through 6 is referred to as MIN (1 through 6), and the smallest correlation among correlations 8 through 13 is referred to as MIN (8 through 13). 6) and MIN (8 ~ 13) are same, it can be judged as vertical correlation.

That is, the interpolation direction of the pixel currently being interpolated is first revealed through step S52. For example, if the vertical correlation has the smallest value, the vertical interpolation direction has the smallest value, and if the correlation has the smallest value, the interpolation direction has the diagonal interpolation direction. For example, when correlation 1, which is the correlation at the window position shown in FIG. 6, is the smallest, it has a diagonal interpolation direction from window w1 to window w2.

Now, when the interpolation direction obtained first in step S52 indicates an oblique edge, the pixel adjacent to the window is irradiated among the upper and lower line image data to check whether the diagonal edge is correct (S53).

Step S53 is a process of investigating the validity of the diagonal edge determination obtained in step S52. The reason for this is that when the right edge appears in the original image as described above, this diagonal edge determination may give an incorrect result.

Referring to FIG. 8, an embodiment of determining the validity of the diagonal edge in step S53 will be described.

First, the left pixel value and the right pixel value adjacent to the upper and lower windows having the smallest correlation are checked (S81).

Referring to FIG. 9 (assuming that the white pixel display represents the pixel value 'white' and the black pixel display represents the pixel value 'black'), the window w1 of the upper line and the window w2 of the lower line are respectively When it is determined that there is a minimum correlation, the pixel values of the left adjacent pixels l1 and l2 of the window w1 and the window w2 and the pixel values of the right adjacent pixels r1 and r2 are checked.

As a result of the check, the pixel value of l1 and the pixel value of l2 are equal to or different from each other within a predetermined value (S82), and the pixel value of r1 and the pixel value of r2 are equal to or different from each other within a predetermined predetermined value. If there is (S83), it is determined that the diagonal edge is correct (S84), otherwise, it is determined that it is not the diagonal edge (S85).

In the example shown in FIG. 9, the pixels l1 and l2 are the same because they are all white pixels, but the pixel values of the pixel r1 are white and the pixel values of the pixel r2 are black. At this time, if the difference between the white pixel value and the black pixel value is not within a preset range (this range can be arbitrarily set, but can be set to a very small value for accurate edge determination), It is determined that the interpolation direction is not the diagonal direction that was originally determined.

Referring to FIG. 5 again, if the diagonal direction is correct as a result of checking in step S53, the diagonal direction is determined as the interpolation direction (S54, S55), otherwise, the vertical interpolation direction is determined (S54, S56). In the example shown in FIG. 9, since the interpolation direction is determined not to be an oblique direction through irradiation of adjacent pixels as described above, the interpolation direction is corrected to the vertical direction.

The interpolation target pixel is interpolated using the determined interpolation direction (S57).

The method of performing interpolation in step S57 can be configured in various ways. For example, the middle position pixel values of two windows indicating the determined interpolation direction are summed and divided by two.

Referring back to FIG. 9, if the interpolation direction is determined in the oblique direction determined as the primary through step S52, the pixel value of the interpolation target pixel z is obtained by adding pixel values of a1 and a2 which are intermediate positions of the window w1 and the window w2. The value is divided by two. However, as described above, since the interpolation target pixel is modified in the vertical direction, the interpolation direction is not interpolated using the pixels a1 and a2 as in the prior art, but interpolation is performed using b1 and b2. The pixel value is grayed out.

Referring to FIG. 10 (assuming that the white pixel display represents the pixel value 'white' and the black pixel display represents the pixel value 'black'), as shown in FIG. 9, the original image is black. Even when a right edge consisting of pixels appears, it can be seen that an error determined as an oblique edge can be prevented and the interpolation direction can be corrected in the vertical direction. That is, the pixel value of the interpolation target pixel z is interpolated to gray instead of white as in the prior art. As a result, deterioration of corners can be prevented when a right edge appears in the original image.

As described above, the diagonal edge can be detected with higher reliability. Accordingly, it is possible to prevent the corner portion from being broken at the edge of the right angle pattern, thereby improving image quality, and to obtain a more perfect progressive image from an interlaced image source.

Claims (7)

Obtaining a correlation between the pixels in the window while symmetrically moving the window with respect to a predetermined region of the upper and lower line image data of the interpolation target pixel; Analyzing the correlation to obtain an interpolation direction; If the interpolation direction is an oblique direction, re-checking whether the diagonal edge is correct by irradiating pixels around the window; Determining the diagonal direction as an interpolation direction if the diagonal direction is correct, and otherwise determining the interpolation direction as a vertical direction; And Performing interpolation of the interpolation target pixel using the determined interpolation direction. The method of claim 1, The step of confirming the interpolation direction is configured to irradiate a pixel adjacent to a corresponding window among the upper and lower line image data to determine whether the diagonal edge is correct. The method of claim 2, Deinterlacing, characterized in that it is configured to determine that the diagonal edge is correct when the difference between the left pixel values adjacent to the window is within a predetermined range and the difference between the right pixel values respectively adjacent to the window is within a predetermined range. Interpolation method. The method of claim 1, And the correlation is a value obtained by subtracting pixel values of corresponding pixels in the mirrored window pair, and adding up the absolute values of the subtracted values. The method of claim 4, wherein And determining that the correlation is larger as the calculated value is smaller. The method according to any one of claims 1 to 5, And a predetermined region of upper and lower line image data of the interpolation target pixel is configured as eight pixels, respectively, left and right. The method of claim 6, And the window is set to a size including five pixels.

Images