KR20030082249A - Motion adaptive spatial-temporal deinterlacing method - Google Patents

Abstract

PURPOSE: A motion adaptive spatial-temporal deinterlacing method is provided to improve the quality of picture by clearly representing a horizontal boundary line in a display device such as a large PDP(Plasma Display Panel). CONSTITUTION: An interlaced scan type image is inputted to a deinterlacing unit(S10). Differences in pixel values are obtained between fields related to pixels around a pixel to be interpolated(S21). The maximum pixel difference is decided as a motion value(S23). It is judged whether the motion value exceeds a threshold(S25). If so, a median filter filters information related to the detected movement through post-processing(S31). If not, the median filter filters the information related to the detected movement through post-processing(S33). A pixel value of the pixel to be interpolated is adaptively interpolated according to the information of the detected movement through temporally axial filtering or spatially axial filtering(S41-S43). A progressive scan type image is output(S50).

Description

Motion Adaptive Spatial-Temporal Deinterlacing Method

The present invention relates to a deinterlacing method for converting an interlaced scan image into a progressive scan image, and more particularly, to improve image quality deterioration of a large screen display device. The present invention relates to a motion adaptive spatial-temporal deinterlacing method.

In general, an NTSC TV system uses an interlaced scan method in which one frame is divided into even and odd fields and scanned at a time difference. A conventional scanning method has been used for a long time because it can reduce image information in half by a simple method. However, since the conventional scanning method sacrifices frequency components in the vertical direction to preserve frequency components in the time direction, serious distortions such as a flicker phenomenon occur frequently in an image having a lot of detailed information in the vertical direction. Recently, display devices that facilitate progressive scan such as LCD (LCD) and Plasma Display Panel (PDP) are widely used, and other display devices such as PCs (PCs) are widely available. Compatibility with is emerging as an important issue. Accordingly, researches on a deinterlacing method for converting a conventional progressive scan image into a sequential scan image have been actively conducted.

The deinterlacing method is largely divided into spatial processing and temporal processing. The spatial axis processing refers to interpolation using only surrounding pixels of a pixel to be interpolated in one field. The spatial axis processing generally eliminates interference in an image by using a function approximating a low pass filter. Functions used here include zero order hold (ZOH) functions, linear interpolation functions, and nonlinear interpolation functions.

The transformation using the ZOH function is a method of using the same pixel value repeatedly, and a serious step phenomenon occurs in a portion where an image boundary appears. The transformation using the linear interpolation function is a method of linear interpolation using a linear function. That is, the missing pixel value between the two pixel values is obtained as the weighted average value of the two pixel values. In this case, when the interpolation ratio of the image is increased, there is a disadvantage that the image quality of the interpolated image may be degraded. The transformation using the ZOH function or the transformation using the linear interpolation function can effectively remove noise because all of them are blurred without considering noise or object contours. However, the interpolated image is blurred overall.

Therefore, a nonlinear filter is designed to remove noise while maintaining the outline of an image, and a representative example thereof is a median filter. The median filter, like a low pass filter, has the effect of smoothing the image, which is useful for noise reduction. However, since the median filter is a non-linear filter, a large amount of calculation is required for signal processing, which places a burden on real time processing.

The spatial axis deinterlacing using the median filter is a deinterlacing method to be used for vertical interpolation by utilizing the contour preservation characteristic of the median filter. This method maintains the contour of the image better than the linear interpolation function, so that a good image can be obtained. However, since this method performs a relatively large amount of calculation, the computational amount is large, and the interpolated image may have irregularity because the pixel value which is the middle value among adjacent pixel values is used as it is.

On the other hand, time axis deinterlacing is a method of filtering on the time axis using motion information between fields. The time axis deinterlacing includes a motion compensated (MC) deinterlacing method for obtaining pixel values of missing scan lines using a motion vector, and a motion non-compensation for interpolation using various types of filters on the time axis without using a motion vector. There is a (non-MC) deinterlacing method.

The MC deinterlacing method is a method of obtaining a motion vector in a field before and after a current field through motion prediction and using the same to interpolate missing pixel values of the current field. The motion prediction can be performed between successive odd and even fields or between fields with the same parity, but better performance when performed between fields with the same parity. The MC de-interlacing method provides an image of good quality, but in this case, the motion prediction operation must be accurate, and the motion prediction is very complicated, which is difficult in real time processing.

In contrast, the non-MC deinterlacing method is a method of performing deinterlacing using simple filtering on the time axis using motion information for various fields but not using a motion vector. In other words, when deinterlacing, motion information of various fields is used, but the interpolation is performed using only surrounding pixel values without using a motion vector. The non-MC deinterlacing method is similar to the method using a ZOH interpolation function or a linear interpolation function on the spatial axis.

As described above, de-interlacing includes a method using motion information and a method not using motion information. The method of not using the motion information is a spatial axis deinterlacing method. The spatial axis deinterlacing methods include line averaging, weighted median filtering, and edge-based line averaging. A method of using the motion information is a time base deinterlacing method, and a non-MC deinterlacing method is used in a scan converter chip (SDA 9400) of NEC of Japan.

The conventional non-MC deinterlacing method used in the SDA 9400 is a motion adaptive deinterlacing method combining spatial axis filtering and time axis filtering. The conventional non-MC deinterlacing method predicts a motion by using before and after fields of an input image of a conventional scan method and determines the presence or absence of the motion. 3-D motion detection is performed using three fields to extract the information on the motion. In this case, a pixel value difference between lines, a pixel value difference between fields, and a pixel value difference between frames are used. Subsequently, when the motion value for the motion information is determined, the pixel value to be interpolated is interpolated by adaptively applying a spatial axis filter and a time axis filter according to the motion value. That is, since there is more information in the spatial domain when there is motion, it is interpolated using the linear interpolation function as a spatial axis filter. When there is no motion, since there is more information in the time domain, the 3-point median filter ( Interpolate using a 3-point median filter.

More specifically, for example, as shown in FIG. 1, in the k-th field, pixel values of neighboring pixels of the i and i + 1 th lines are a and b, respectively, and i and i + When the pixel value of the line to be interpolated between the first lines is A, if there is the movement, the pixel value A is interpolated by Equation 1 using a linear interpolation function as a spatial axis filter.

A = (a + b) / 2

If there is no motion, the pixel value A is interpolated by Equation 2 using a three-point median filter.

A = med (a, b, Aprev)

Here, Aprev is a pixel value in the same pixel of the interpolated line in the K-1th field.

However, after the motion information is extracted, the wrong motion information due to noise can be corrected by removing the motion separated from each other through the various fields or the motion in the still area. For this purpose, a median filter is commonly used as a post-treatment filter.

However, since the post-processing filter is not used in the conventional non-MC deinterlacing method, the reliability of the motion information is low. In addition, the linear interpolation function used as the spatial axis filter has a problem of blurring the boundary portion of the interpolated image information. As a result, the image quality of the large screen display device is degraded.

Meanwhile, the ELA method is a representative method among methods performed in a spatial domain without using the motion information. The ELA method is a method of using an average value of pixel values in a direction having the highest correlation among vertical and diagonal directions among pixel values in a neighboring line of a lost line. It is used.

The conventional ELA method is a method of determining the direction of an edge in accordance with the correlation between the values of pixels to be interpolated and interpolating by averaging pixel values corresponding to the edge direction. That is, by setting a 3 X 3 window with respect to the pixel to be interpolated, and calculating the correlation between the pixel values placed in the spatially symmetrical state in the window, average the pixel values in the direction of highest correlation The pixel value to be interpolated is estimated.

In more detail, the direction of the boundary line considered in the conventional ELA method is a vertical direction and a diagonal direction. For example, assuming that the pixel to be interpolated is the j-th pixel on the i-th line and that pixel value is x (i, j), as shown in FIG. 2, a 3 X 3 sized window centering on the pixel to be interpolated Set. In the window, j-1, j, and j + 1th pixel values on the i-1th line are x (i-1, j-1), x (i-1, j), and x (i-1, j, respectively. +1) and the j-1, j, j + 1 th pixel values on the i + 1 th line are x (i + 1, j-1), x (i + 1, j), x (i + 1), respectively. , j + 1).

The boundary direction of the pixel to be interpolated is determined by the differences C1, C2, and C3 between three peripheral pixel values. The differences C1, C2, and C3 may be calculated by Equation 3 below.

At this time, since the direction showing the smallest difference is the direction having the highest correlation, x (i, j) is interpolated by the pixel values corresponding to the direction. In the case where C1, C2, and C3 are the minimum, x (i, j) can be calculated by the following equation (4).

That is, when C1 and C3 are minimum, since the boundary line direction is a diagonal direction, the average value of pixel values located in the diagonal direction becomes the pixel value to be interpolated. When C2 is minimum, since the boundary line direction is vertical, the average value of pixel values located in the vertical direction becomes the pixel value to be interpolated.

However, the conventional ELA method works well for a vertical boundary line or a diagonal boundary line, but does not consider the horizontal boundary line and thus cannot interpolate a value corresponding to the horizontal boundary line properly. That is, if there is a horizontal boundary line passing through the pixel to be interpolated, the horizontal boundary line is forcibly considered as a vertical boundary line or a diagonal boundary line, thereby incorrectly interpolating the pixel value x (i, j) to be interpolated. As a result, since the horizontal boundary lines do not appear clearly in the horizontal boundary portions, the image quality deteriorates in the PDP of the large screen.

As described above, the conventional ELA method has a relatively stable performance with respect to various images, but the performance is deteriorated with respect to an image having a large number of horizontal boundaries, thereby degrading image quality. This is because in the conventional ELA method, since there is no consideration of the horizontal boundary line, even the pixel including the horizontal boundary line is treated the same as the vertical boundary line or the diagonal boundary line. To improve this, first, a horizontal boundary line must be detected and the corresponding pixel must be interpolated appropriately. Therefore, if a horizontal boundary line is not detected, it is necessary to interpolate the pixel by a conventional ELA method, and when the horizontal boundary line is detected, the pixel needs to be interpolated in a different manner.

Accordingly, an object of the present invention is to provide a motion adaptive space-time de-interlacing method to clearly display the boundary portion of the image.

Another object of the present invention is to provide a motion-adaptive space-time deinterlacing method for clearly displaying a horizontal boundary portion of an image.

It is still another object of the present invention to provide a motion adaptive space-time deinterlacing method for improving image quality deterioration.

1 is an exemplary diagram for explaining interpolation of pixel values applied to a de-interlacing method on a non-motion compensated (non MC) time axis according to the prior art.

Figure 2 is an exemplary view showing a 3X3 window to which the present invention is applied.

3 is a flowchart showing a motion adaptive space-time deinterlacing method according to the present invention.

4 is an exemplary diagram showing four continuous fields for motion detection to which the present invention is applied.

5 is an exemplary view showing a five-point median filter to which the present invention is applied.

6 is a detailed flowchart illustrating spatial axis filtering of a motion adaptive space-time deinterlacing method according to the present invention.

Motion-adaptive space-time deinterlacing method according to the present invention for achieving the above object

Detecting a motion value by detecting pixel values in a window of a predetermined size including pixels to be interpolated from a plurality of temporally continuous field data of an input image of a parallel scan method, and calculating and comparing pixel value differences between the fields; Determining whether the motion value exceeds a predetermined threshold; Performing spatial axis filtering to interpolate the pixel value of the pixel to be interpolated with weights to spatial axis information if the motion value is considered to be motion exceeding the threshold value; And if it is determined that there is no motion, wherein the motion value is less than or equal to the threshold value, performing time-base filtering to interpolate the pixel value of the pixel to be interpolated with time-axis information.

Preferably, one of 8 and 16 may be selected and used as the threshold.

Preferably, the method may include performing post-processing filtering after determining whether the motion value exceeds the threshold.

Preferably, the space axis filtering step

The pixel values to be interpolated are detected by detecting pixel values in a window of a predetermined size centering on the pixels to be interpolated from predetermined field data of the input image of the interlocking scan method, and obtaining a pixel value difference in horizontal, vertical, and diagonal directions. Determining whether or not it is a horizontal boundary line; If it is determined that the pixel to be interpolated is the horizontal boundary line, horizontal interpolating the pixel to be interpolated; And if it is determined that the pixel to be interpolated is not the horizontal boundary line, interpolating the pixel to be interpolated by ELA method.

Preferably, the step of determining whether or not the pixel to be interpolated is the horizontal boundary line may be performed by determining a difference in pixel values in a vertical, diagonal, and horizontal boundary direction with respect to the pixel to be interpolated in a 3 × 3 size window centering on the pixel to be interpolated. Calculating each; And determining whether the pixel to be interpolated is the horizontal boundary by determining whether the pixel value difference in the horizontal boundary line direction is the minimum among the pixel value differences.

Preferably, the method may include determining whether the horizontal boundary line is a background region when the pixel value difference in the horizontal boundary line direction is determined to be minimum.

Preferably, when it is determined that the difference in pixel values in the horizontal boundary direction is not the minimum, the pixel to be interpolated may be interpolated by the EL method.

Preferably, if it is determined that the horizontal boundary line is not the background area, it may include determining whether the boundary line is a texture area.

Preferably, when it is determined that the horizontal boundary line is the background area, the pixel to be interpolated may be interpolated by the EL method.

Preferably, when it is determined that the boundary line is not the texture area, the pixel to be interpolated is horizontally interpolated, and when it is determined that the boundary line is the texture area, it may be interpolated by the EL method.

Preferably, the step of horizontal interpolation comprises: determining whether the pixel to be interpolated is a starting point of the horizontal boundary line; Determining the pixel value of the pixel to be interpolated as a line average when it is determined as a starting point of the horizontal boundary line; And if it is determined that the image corresponds to a continuous horizontal boundary line without being determined as a starting point of the horizontal boundary line, interpolating a pixel value of the pixel to be interpolated in consideration of pixels in a horizontal direction.

Hereinafter, a motion adaptive space-time deinterlacing method according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

3 is a flowchart illustrating a motion adaptive space-time deinterlacing method according to an embodiment of the present invention. Referring to FIG. 3, first, when an input image of a caterpillar scanning method is input to a deinterlacing unit (not shown) in step S10, the deinterlacing unit may continuously store various field data of the input image in step S20. Pixel values of specific lines in the liver are used to determine whether there is motion in the video.

In more detail, in order to detect the movement, a buffer for storing previous field data is required. There are cases where a field memory for storing video data of two fields is used for motion detection and a field memory for storing video data of three fields. When the number of field memories is small, motion detection can be easily implemented, but motion detection error is a big disadvantage because it cannot detect fast moving motion. In the present invention, a motion detection method using three field memories, which are known to have the best performance, will be described based on detecting motion.

As shown in FIG. 4, four consecutive fields for motion detection, e.g., the i, j, k, l th fields are Fi, Fj, Fk, Fl, the current field is Fk and the previous field is Fi , Fj and the next field is Fl. In the fields Fi and Fk, pixel values of the j-th pixel in the i-th line are ai and ak, respectively, and pixel values of the j-th pixel in the i + 1th line are bi and bk. The pixel values of the j th pixel in the line to be interpolated between the i and i + 1 th lines in the fields Fj and Fl are cj and cl. In the field Fk, the pixel value of the pixel between the i-th, i-th line, j-th pixel, that is, the pixel to be interpolated is zk.

In step S21, the pixel value difference between fields for pixels around the pixel to be interpolated is obtained. That is, the pixel value difference D1 between the pixel values ai and ak, the pixel value difference D2 between the pixel values bi and bk, and the pixel value difference D3 between the pixel values cj and cl are calculated by Equation 5. .

After the calculation of the pixel value differences D1, D2, and D3 is completed, in step S23, the pixel value differences D1, D2, and D3 are compared with each other to determine the maximum pixel value difference among them as a motion value.

After the motion value is determined, it is determined in step S25 whether the motion value exceeds a threshold, e.g., 8 (or 16) determined by the experimental value. At this time, if the motion value exceeds the threshold, it is detected that there is the motion. If the movement does not exceed the threshold, it is detected that there is no movement.

After the motion value is compared with the threshold value, in step S30, the information on the detected motion is post-processed filtered by, for example, a median filter. In other words, if the motion value exceeds the threshold, in step S31, the information on the detected motion is post-processed filtered by, for example, a median filter. If the motion value does not exceed the threshold value, in step S33, the information on the detected motion is post-processed filtered by, for example, a median filter. This is to compensate for the movement in the stationary region caused by the momentary noise, and the lost movement in the movement region. In image processing applications, especially deinterlacing, median filters are known to be most suitable for improving edges and removing noise. In binary motion information, the median filter shows excellent performance in eliminating separate motion information or filling in lost motion information. The median filter used in the present invention has a 5-point cross shape, as shown in FIG. 5, and shows better performance than erosion and dilation. On the other hand, the post-processing filtering process is not necessarily necessary but optional.

After the post-processing filtering is completed, in step S40, the pixel value of the pixel to be currently interpolated is adaptively interpolated according to the detected motion information. In more detail, when the pixel value of the pixel to be interpolated currently is x (i, j) in a window of a specific field t as shown in FIG. 2, the motion value of x (i, j) is an arbitrary threshold. If less than the value, in step S41, the pixel value x (i, j) is time-base filtered by a median filter, for example a 7-point weighted median filter. In the weighted median filter, time-domain input samples are used repeatedly. This increases the probability that a time domain sample is output as a result. The weighted median filter is an extension of the standard median filter. The weighted median filter may be represented by Equation 6 below.

y = med [x (i-1, j-1, t), x (i-1, j, t), x (i-1, j + 1, t), x (i + 1, j-1 , t), x (i + 1, j, t), x (i + 1, j + 1, t), 3 X x (i, j, t-1)]

If the motion value is larger than the threshold value, that is, if there is the motion, spatial axis filtering is performed to weight the spatial axis information in step S43. That is, as shown in FIG. 6, first, in step S430, it is determined whether the pixel to be interpolated of the input image of the catenary scanning method is a horizontal boundary line. In more detail, in step S431, the pixel value difference C4, in order to obtain a correlation in the horizontal direction with respect to the pixel value x (i, j) to be interpolated in the 3X3 window as shown in FIG. C5) is calculated by the following equation. As already explained, C1, C2, and C3 are also calculated.

Here, C4 is a difference in pixel values of horizontal left and right pixels with respect to the pixel to be interpolated on the line above the i-th line where the pixel to be interpolated is located, that is, the i-1 th line. C5 is a pixel value difference of horizontal left and right pixels with respect to the pixel to be interpolated in the line below the i-th line where the pixel to be interpolated is located, i.

After the C1, C2, C3, C4 and C5 are calculated, the minimum value of the C1, C2, C3, C4 and C5 is selected by comparing the C1, C2, C3, C4 and C5 in step S433. At this time, the boundary line direction having the minimum value becomes the boundary line direction having the highest correlation. For example, if C4 or C5 is the minimum value among the C1, C2, C3, C4, and C5, the boundary line direction having the highest correlation becomes the horizontal boundary line direction, and thus the pixel to be interpolated is considered to be on the horizontal boundary line. If C4 or C5 is not the minimum among the C1, C2, C3, C4, and C5, the pixel to be interpolated is not considered to be on the horizontal boundary because the direction having the highest correlation is not the horizontal boundary.

However, in the case of determining whether or not a horizontal boundary line is used in this manner, the background area also has a high possibility of an error to be determined as a horizontal boundary line. Therefore, in order to prevent an error in which the background area is determined as a horizontal boundary line, it is necessary to determine whether the horizontal boundary line is the background area.

If the pixel to be interpolated is considered to be on the horizontal boundary line, it is determined in step S435 whether the horizontal boundary line is the background area or not. That is, it is determined whether the difference in values between adjacent upper and lower lines of the pixel to be interpolated is equal to or greater than a threshold value, for example, 5 to 8, which is preset in the vertical direction.

In this case, when the pixel value difference between the upper and lower lines is greater than or equal to the threshold value, the boundary line is not a background area, so the boundary line is regarded as a horizontal boundary line. When the pixel value difference between the upper and lower lines is smaller than the threshold value, the boundary line is regarded as a background area.

After the boundary line is determined as the background area, it is determined in step S437 whether the area having the boundary line is a texture area or not. In other words, the difference d1, d2, and d3 of the pixel value is obtained by the following expression (8). Although the boundary line is regarded as a horizontal boundary line, in a texture-type image in which the pixel value changes rapidly, the detected boundary line is not easily detected as a horizontal boundary line even though the boundary line is not a horizontal boundary line. For sake.

Then, it is determined whether the signs of the differences d1, d2, and d3 are all the same positive (or negative). At this time, if the sign of the difference (d1), (d2), (d3) are all the same (or negative), the boundary line can be seen as a continuous boundary line in the 3X3 window of Figure 2, only when the boundary line is true It can be considered as a horizontal boundary line.

On the other hand, if the signs of the difference (d1), (d2), (d3) are not all the same (or negative), the boundary line is not seen as a continuous boundary line in the 3X3 window of FIG. The boundary at this time is regarded as a texture region.

Although information on the horizontal boundary line has been found by the process up to now, it is difficult to interpolate values close to the pixel values to be interpolated of the horizontal boundary line only with this information. This is because x (i, j-1) is known around the pixel value x (i, j) to be interpolated, but x (i, j + 1) is not known.

When interpolating a pixel value x (i, j) which is a horizontal boundary line without knowing the value of x (i, j + 1), x (i, j-1) becomes important information. This is because x (i, j-1) is closest to the pixel value x (i, j) of the horizontal boundary among pixel values currently known in the 3X3 window of FIG. 2. However, there is a point to be considered when reflecting the value of x (i, j-1). If x (i, j) is the starting point of the horizontal boundary line, x (i, j-1) would have been interpolated by the algorithm of the conventional ELA method, and the pixel is likely not the horizontal boundary line. If the pixel value x (i, j-1) that is not the horizontal boundary line affects the interpolation of x (i, j), the adverse effect continues on the horizontal boundary line area.

Considering this point, if the pixel to be interpolated is on the horizontal boundary line, the pixel value to be interpolated is processed by the horizontal interpolation of the present invention in step S530, and the pixel to be interpolated is not on the horizontal boundary line. In step S630, the pixel values to be interpolated are processed by interpolation of the conventional ELA method.

Referring to the process of step S530 in more detail, it is determined in step S531 whether the pixel to be interpolated is a starting point of a horizontal boundary line. At this time, if the pixel to be interpolated is the starting point of the horizontal boundary line, it is best to give a value at the position closest to the pixel to be interpolated irrespective of the horizontal boundary line or the diagonal boundary line. Therefore, in step S533, the pixel value x (i, j) corresponding to the starting point of the boundary line is uniformly equal to the average value of the spatially closest pixel values x (i-1, j) and x (i + 1, j). Decide on That is, if the pixel is the starting point of the horizontal boundary line, the pixel value x (i, j) is determined by the following equation (9).

x (i, j) = {x (i-1, j) + x (i + 1, j)} / 2

On the other hand, if the pixel is not at the starting point of the boundary line but at a position corresponding to a continuous horizontal boundary line, in step S535, the value x (i, j) of the pixel is converted to the algorithm of the conventional ELA method x (i, j− Interpolate to reflect 1). This takes into account the correlation between the pixels and the continuity of the horizontal boundary. That is, in the case of a pixel corresponding to a continuous horizontal boundary line, the pixel value x (i, j-1) adjacent to each other in the horizontal direction may be used, and the value x (i, j) is determined by Equation 10 below.

Subsequently, after the interpolation in the step S530 or the step S630 is completed, the interpolated image may be sequentially output in step 50 of FIG. 3.

Therefore, the present invention can adaptively interpolate according to the present invention. In addition, the present invention can obtain a PSTN (pixel signal to noise) value superior to the conventional ELA method by considering the starting point of the horizontal boundary line, and can more clearly represent the horizontal boundary line portion. As a result, image quality deterioration can be improved in a display device such as a large LCD or a PDP.

Meanwhile, in the present invention, interpolation for one pixel has been mainly described. In fact, once interpolation of the pixel is completed, interpolation is successively performed for the next neighboring pixel. Descriptions thereof will be omitted for convenience of description in order to avoid duplication of description.

As described in detail above, the motion-adaptive space-time deinterlacing method according to the present invention detects pixel values in a window having a predetermined size including pixels to be interpolated from several temporally continuous field data of an input image of a conventional scanning method. The motion value is detected by comparing and comparing the pixel value difference between the fields, and it is determined whether the motion value exceeds a certain threshold. If the motion value is less than or equal to the threshold value, it is assumed that there is no motion and time axis filtering is performed. If the motion value exceeds the threshold, the motion is assumed to be present and spatial axis filtering is performed. That is, in a specific field of the input image of the conventional scanning method, current interpolation is performed by detecting pixel values in a window of a predetermined size including a pixel to be interpolated currently, and calculating and comparing the difference in the vertical, horizontal, and diagonal directions of the pixel values. It is determined whether or not the pixel to be made is a horizontal boundary line. At this time, if it is determined that it is not a horizontal boundary line, the pixel to be interpolated is interpolated by the existing ELA method. If it is determined that the image is a horizontal boundary line, the pixel to be interpolated is horizontally interpolated in consideration of the horizontal boundary line. That is, when it is determined that the pixel to be interpolated is a starting point of a horizontal boundary line, the pixel value to be interpolated is determined as an average value of pixel values of pixels of the upper and lower lines closest to the pixel to be interpolated. If it is determined that the pixel to be interpolated is not the starting point of the horizontal boundary but corresponds to a continuous horizontal boundary, the pixel value of the pixel before the pixel to be interpolated is interpolated by the ELA method.

Therefore, the spatial axis filtering according to the present invention improves the blurring of the horizontal boundary lines generated when the EL method is interpolated, so that the horizontal boundary lines can be clearly displayed in a display device such as a large screen PDP. As a result, image quality deterioration can be improved.

On the other hand, the present invention is not limited to the contents described in the drawings and detailed description, it is obvious to those skilled in the art that various modifications can be made without departing from the spirit of the invention. .

Claims (11)

Detecting a motion value by detecting pixel values in a window of a predetermined size including pixels to be interpolated from a plurality of temporally continuous field data of an input image of a parallel scan method, and calculating and comparing pixel value differences between the fields; Determining whether the motion value exceeds a predetermined threshold; Performing spatial axis filtering to interpolate the pixel value of the pixel to be interpolated with weights to spatial axis information if the motion value is considered to be motion exceeding the threshold value; And And performing time-base filtering to interpolate the pixel value of the pixel to be interpolated with a weight on time-axis information, when the motion value is considered to be no motion, which is less than or equal to the threshold value. 2. The method of claim 1, wherein one of 8 and 16 is selected and used as the threshold value. 2. The method of claim 1, comprising determining whether the motion value exceeds the threshold and then performing post-processing filtering. The method of claim 1, wherein the spatial axis filtering step The pixel values to be interpolated are detected by detecting pixel values in a window of a predetermined size centering on the pixels to be interpolated from predetermined field data of the input image of the interlocking scan method, and obtaining a pixel value difference in horizontal, vertical, and diagonal directions. Determining whether or not it is a horizontal boundary line; If it is determined that the pixel to be interpolated is the horizontal boundary line, horizontal interpolating the pixel to be interpolated; And And if the pixels to be interpolated are not the horizontal boundary lines, interpolating the pixels to be interpolated by ELA. The method of claim 4, wherein the determining whether the pixel to be interpolated is the horizontal boundary line is as follows. Calculating pixel value differences in the vertical, diagonal, and horizontal boundary directions with respect to the pixel to be interpolated in a 3 × 3 window centering the pixels to be interpolated; And And determining whether or not the pixel to be interpolated is the horizontal boundary line by determining whether the pixel value difference in the horizontal boundary line direction is the minimum among the pixel value differences. The motion adaptive space-time deinterlacing method of claim 5, further comprising: determining whether the horizontal boundary line is a background region when the pixel value difference in the horizontal boundary direction is determined to be minimum. The motion adaptive space-time deinterlacing method according to claim 5, wherein when it is determined that the difference in pixel values in the horizontal boundary direction is not the minimum, the pixel to be interpolated is interpolated by the EL method. The method of claim 6, further comprising: determining whether the boundary line is a texture area or not when the horizontal boundary line is not the background area. The motion adaptive space-time deinterlacing method according to claim 6, wherein when the horizontal boundary line is determined to be the background area, the pixel to be interpolated is interpolated by the EL method. 10. The method of claim 8, wherein if it is determined that the boundary line is not the texture area, horizontal interpolation of the pixel to be interpolated is performed. Motion adaptive space-time deinterlacing method. 11. The method of claim 4 or 10, wherein the step of horizontal interpolation Determining whether the pixel to be interpolated is a starting point of the horizontal boundary line; Determining the pixel value of the pixel to be interpolated as a line average when it is determined as a starting point of the horizontal boundary line; And And determining that the pixel value of the pixel to be interpolated is interpolated in consideration of pixels in a horizontal direction, if it is determined that the image corresponds to a continuous horizontal boundary line without being determined as a starting point of the horizontal boundary line. .