KR20090020134A - Method and apparatus for motion adaptive de-interlacing - Google Patents

Abstract

A motion adaptive de-interlacing method and an apparatus thereof are provided to detect a horizontal boundary to interpolate a pixel by a time axis de-interlacing method, thereby minimizing interpolation error to improve image quality. A motion information detection unit(120) determines a moving area and detects the amount of motion information by using a previous field and a current field. A horizontal boundary detector(140) detects a horizontal boundary by using upper/lower pixels related to a pixel to be interpolated and pixels corresponding to the previous pixel to minimize interpolation error even though the pixel to be interpolated exists in a still area when the pixel to be interpolated exists in the horizontal boundary. The first filter unit(160) calculates a pixel value based on the previous field and current field. The second filter unit(170) calculates a pixel value based on the inputted current field.

Description

Method and apparatus for motion adaptive deinterlacing

The present invention relates to a method and an apparatus for motion adaptive deinterlacing, and more particularly, in the case of performing interpolation of a pixel of a field input by a parallel scan method in performing the motion adaptive deinterlacing. In order to minimize interpolation errors, the present invention relates to a motion adaptive deinterlacing method and apparatus capable of improving image quality by detecting a horizontal boundary and interpolating the pixels.

Recently, devices using progressive scan such as liquid crystal display (PDP), plasma display panel (PDP), and digital video disk (DVD) are widely used, and thus, images of the conventional interlaced scan method are used. The research on the method of converting an image into a sequential scanning method (IPC: Interlace-to-Progressive Conversion) has been actively conducted.

FIG. 1 is a diagram illustrating a parallel scan method. For convenience of description, one frame is illustrated as three scan lines each including five pixels. Referring to FIG. 1, in a parallel scan scheme, a frame is divided into a first field including an even scan line (a second scan line) and a second field including an odd scan line (first and third scan lines), The first and second fields are scanned at a time difference.

De-interlacing is a pixel included between the scan lines included in each field (the first and second fields) (for example, a pixel included in the second scan line in the case of the second field) due to the parallel scanning scheme. (N) is converted to a sequential scanning method by interpolating using adjacent pixels.

Such deinterlacing methods are largely classified into a space axis deinterlacing method and a time axis deinterlacing method.

The spatial axis deinterlacing method uses the fact that pixels adjacent to the scan line to be interpolated have a high correlation with the scan line to be interpolated. Since the interpolation is performed using only the pixels of the current field, a separate space for storing a field other than the current field is required. Not. However, since the scan line to be interpolated has a high correlation with the pixels adjacent to the time axis in addition to the pixels adjacent to the space axis, performance is limited, such as deterioration of image quality.

Such spatial axis deinterlacing methods include line iteration, linear interpolation, and edge based line averaging (ELA).

The line repetition method is a method of copying the scan lines above or below the scan line to be interpolated to the scan line to be interpolated as it is, which causes severe jagged edges and jitter artifacts in the outline portion of the image.

Linear interpolation is a method of interpolating a scan line using an average value of brightness values (hereinafter referred to as pixel values) of each pixel included in the scan lines to be interpolated. The linear interpolation method produces smoother results than the line repetition method. However, like the line repetition method, the image has aliasing and blurring, and the image is blurred overall.

The ELA method, which is the most widely used among the spatial axis deinterlacing methods, uses an average value of pixel values in the most correlated direction in the vertical and diagonal directions among the pixels included in the upper and lower scan lines of the interpolation scan line. However, it is difficult to show good interpolation performance with respect to the diagonal boundary line, but good interpolation performance with respect to the horizontal boundary line, and there is a problem that an image blur occurs with respect to the horizontal boundary of the still area like the other spatial axis filters.

2 is a diagram for explaining a boundary area. The boundary area is an area in which a brightness value of each pixel differs from the boundary when the line is drawn in a horizontal, vertical or diagonal direction in one frame or field. Referring to FIG. 2, the upper and lower contours of the square correspond to the horizontal boundaries, and the left and right contours of the square correspond to the vertical boundaries.

The time axis deinterlacing method is a method of interpolating a scanning line with reference to fields before and / or after a current field in order to consider the correlation on the time axis, which is a problem of the space axis deinterlacing method, and the image quality is improved as the number of the referenced fields increases. However, since all referenced fields must be stored in memory, hardware resources increase.

Such time-base deinterlacing methods include field repetition, non-linear frame interpolation, vertical-time axis median filtering, motion compensation and motion adaptive methods.

The field repetition method is an interpolation method by copying the scan line of the previous field corresponding to the scan line to be interpolated to the position of the scan line to be interpolated.

The nonlinear frame interpolation method uses the average value of pixel values included in the scan lines of the before and after fields corresponding to the scan lines to be interpolated. This method shows a good performance with respect to the scan lines in the still area, but the image quality with respect to the scan lines in the motion area. This very cloudy phenomenon occurs.

Vertical-time axis median filtering is a method of using pixels of fields before or after the current field and pixels adjacent to the pixel to be interpolated in the current field, which is excellent for noise removal. The vertical-time axis median filtering is a widely used method due to its excellent noise removal and easy implementation, but has a problem in that a step phenomenon occurs in a moving region.

The motion compensation method is a method using motion estimation and compensation. The algorithm is complicated and is not widely used due to the need for additional hardware such as a motion predictor.

The motion adaptive method distinguishes a still region from a motion region and uses a different deinterlacing method. The motion adaptive method uses a time axis deinterlacing method in a still region and a space axis deinterlacing method in a motion region. The motion adaptive method is divided into two-field, three-field and four-field motion adaptive methods according to the number of fields used for detecting the motion region. In this case, the higher the number of fields used, the more accurately the still / motion area can be detected. However, the image quality can be improved. However, as the number of fields used increases, hardware resources for storing the fields also increase.

Accordingly, in order to detect a motion region using less hardware resources, a two-field motion adaptive method may be used that performs deinterlacing using only two fields of the motion adaptive deinterlacing method.

However, in the conventional two-field motion adaptive method, the motion area is determined by calculating an absolute value of an average value of upper and lower pixel values adjacent to the pixel to be interpolated and an absolute value of a pixel value difference between pixels of the previous field corresponding to the pixel to be interpolated. Therefore, when the pixels to be interpolated exist on the horizontal boundary, an interpolation error that is incorrectly determined to exist in the motion area occurs even though the pixels to be interpolated exist in the still area. Therefore, in the conventional two-field motion adaptive method, when a pixel to be interpolated exists on a horizontal boundary due to the interpolation error, even if the pixel to be interpolated exists in a still region, the interpolation method uses a spatial axis deinterlacing method such as ELA. Therefore, there was a problem that the image quality is lowered.

The present invention has been devised to solve the problems of the prior art as described above, and an object of the present invention is to detect a horizontal boundary and, when a pixel to interpolate exists on the horizontal boundary, a three-point vertical-time axis (VT) median. The present invention provides a motion adaptive deinterlacing method and apparatus capable of improving image quality by interpolating pixels using a time-base deinterlacing method such as a filter.

Further, another object of the present invention is to provide a device having a low hardware resource using a two-field motion adaptive deinterlacing method, even if the pixel to be interpolated exists on a horizontal boundary, even if the pixel to be interpolated exists in the still area. The present invention provides a motion adaptive deinterlacing method and apparatus that can improve image quality by minimizing interpolation errors that are incorrectly determined to exist.

To this end, the motion adaptive deinterlacing method according to the present invention comprises the steps of: a) detecting an amount of motion information with respect to a pixel to be interpolated based on pixels present in a plurality of fields of a parallel scan scheme; b) comparing the amount of motion information with a preset first threshold value and, when the amount of motion information is equal to or greater than a first threshold value, at least one of upper and lower pixels adjacent to the pixel and pixels of a previous field corresponding to the pixel; Detecting a horizontal boundary using one; And c) selecting and interpolating a time axis filter if the pixel is on the horizontal boundary.

In addition, the motion adaptive deinterlacing apparatus according to the present invention comprises: a motion information detection unit for detecting motion information on a pixel to be interpolated based on a plurality of fields of a parallel scan method; A horizontal boundary detector for detecting horizontal boundary information using at least one of upper and lower pixels adjacent to the pixel and pixels of a previous field corresponding to the pixel when the motion information is detected to a predetermined level or more; A first filter unit including a time axis filter interpolating the pixel based on the field; A second filter unit including a spatial axis filter interpolating the pixel based on the field; And a result selector configured to select a result value interpolated through the first filter part or the second filter part using at least one of the motion information and the horizontal boundary information.

In addition, the motion adaptive de-interlacing apparatus according to the present invention, the motion information detection unit for detecting the motion information for the pixel to be interpolated based on a plurality of fields of the parallel scan method; A horizontal boundary detector for detecting horizontal boundary information using at least one of upper and lower pixels adjacent to the pixel and pixels of a previous field corresponding to the pixel when the motion information is detected to a predetermined level or more; A filter selecting unit which selects a filter to interpolate the pixel by using the motion information and the horizontal boundary information; And a deinterlacing unit including a plurality of filters and interpolating the pixel using the selected filter.

Accordingly, the motion adaptive deinterlacing method and apparatus according to the present invention detects a horizontal boundary, and when the pixel to be interpolated exists on the horizontal boundary, the pixel is interpolated by a time-axis deinterlacing method such as a 3-point vertical-time axis (VT) median filter. By interpolating, the interpolation error that is determined to be present in the motion area of the pixel on the horizontal boundary in the still area is minimized, thereby improving image quality.

In addition, the motion adaptive deinterlacing method and apparatus according to the present invention has an effect of minimizing interpolation errors even in a device having less hardware resources by performing deinterlacing using only two fields.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings and preferred embodiments. For reference, in the following description, detailed descriptions of well-known functions and configurations that may unnecessarily obscure the subject matter of the present invention are omitted.

3 is a block diagram of a motion adaptive deinterlacing apparatus according to a first embodiment of the present invention. Referring to the drawings, the motion adaptive deinterlacing apparatus 100 includes a motion information detector 120, a horizontal boundary detector 140, a first filter 160, a second filter 170, and a result selector 180. It includes.

The motion information detector 120 may determine a motion area by using two input fields, that is, a previous field and a current field, and detect the amount of motion information. The amount of motion information may be calculated as an absolute value of a difference between an average value of pixel values of upper and lower pixels of a pixel to be interpolated and a pixel value corresponding to a pixel to be interpolated in a previous field.

For example, if the absolute value is less than the threshold value '16', the motion information detector 120 may determine that the pixel to be interpolated exists in the still area. On the other hand, if the absolute value is greater than or equal to the threshold '16', the pixel to be interpolated exists in the motion region, and the greater the absolute value, the greater the amount of motion information.

However, since the detection of motion information may be inaccurate as noise is mixed when an image is input, the above two pixels and two subsequent pixels existing on the same scan line as the pixels may be used. After the absolute values are obtained, the motion information can be detected using their intermediate values.

4 illustrates pixels referred to for detecting motion information in the pixel X to be interpolated, and the motion information detector 120 determines the amount MDx of the motion information included in the pixel X to be interpolated using the pixels. It is detected through [Equation 1].

MDx = Median (MD1, MD2, MD3, MD4, MD5)

Where MD1 = | (a _{t -2} + b _{t -2} ) / 2-c _{t -2} |

MD2 = | (a _{t -1} + b _{t -1} ) / 2-c _{t -1} |

MD3 = | (a _t + b _t ) / 2-c _t |

MD4 = | (a _{t +1} + b _{t +1} ) / 2-c _{t +1} |

MD5 = | (a _{t +2} + b _{t +2} ) / 2-c _{t +} 2 |

Where a represents a pixel value existing on the upper scan line of the pixel to be interpolated, b represents a brightness value of a pixel existing on the lower scan line of the pixel to be interpolated, and c represents a scan line of the previous field corresponding to the pixel to be interpolated Indicates the pixel value present on the image. On the other hand, t represents the positions of the pixels existing on the same vertical line as the pixel at the X point.

Here, the median filter is a filter that takes an intermediate value of the input pixel values and has an effect of smoothing an image by removing noise while maintaining an outline. Detailed description of the median filter will be omitted since it may obscure the subject matter of the present invention.

When the pixel X to be interpolated is present on the horizontal boundary, the horizontal boundary detection unit 140 detects a large amount of motion information MDx. In order to minimize the interpolation error that is incorrectly determined to exist, when the amount of motion is detected by the motion information detector 120 to a predetermined level or more, the upper and lower pixels and the previous field having a high correlation with the pixel X to be interpolated are high. The horizontal boundary is detected using the corresponding pixels.

5A is a diagram showing that there is a pixel X to be interpolated on a horizontal boundary (hereinafter referred to as a bottom horizontal boundary) including a pixel having a high correlation with a lower pixel, and FIG. 5B is a high correlation with an upper pixel. FIG. 7 is a diagram illustrating that pixels X to be interpolated exist on a horizontal boundary including a pixel having a relationship (hereinafter, referred to as an upper horizontal boundary). As illustrated, the pixels referred to in the horizontal boundary detection may vary depending on which of adjacent pixels has a higher correlation with the pixel X to be interpolated.

FIG. 6 is a diagram illustrating pixels to be referred to when a pixel to be interpolated exists on a lower horizontal boundary. Referring to FIG. 6, when the scan line including the pixel X to be interpolated corresponds to the lower horizontal boundary, the difference AD3 between the upper pixel value and the lower pixel value of the pixel to be interpolated is large. The difference CD1 and CD2 between the pixel value to be interpolated and the lower pixel value thereof and the difference CD3 between the pixel value to be interpolated and the pixel value corresponding to the previous field are small. That is, when the absolute value of AD3 is large and the absolute value of CD1, CD2 and CD3 is small, the pixel X to be interpolated corresponds to the lower horizontal boundary.

The horizontal boundary detector 140 obtains absolute values AD1 to AD5 of the difference between at least two upper and lower pixel values adjacent to the pixel to be interpolated through Equation 2 to detect the lower horizontal boundary, and the median filter. Find the median value ADx using.

ADx = Median (AD1, AD2, AD3, AD4, AD5)

Where AD1 = | (a _{t -2} -b _{t -2} ) |

AD2 = | (a _{t -1} -b _{t -1} ) |

AD3 = | (a _t -b _t ) |

AD4 = | (a _{t +1} -b _{t +1} ) |

AD5 = | (a _{t +2} -b _{t +2} ) |

In addition, the absolute value (BD1, BD2, CD1, CD3, DD1, and DD2) of the difference between the pixel value to be interpolated and the adjacent lower pixel values through [Equation 3] and the correspondence between the lower pixel value of the pixel to be interpolated and the previous field. The absolute values BD3, CD3 and DD3 of the difference between the pixel values are calculated to calculate respective intermediate values LD1, LD2 and LD3, and then the intermediate values LDx are obtained again. 6 illustrates pixels referred to for calculating AD3 and LD2 for convenience of description.

LDx = Median (LD1, LD2, LD3)

Where LD1 = Median (BD1, BD2, BD3)

BD1 = | b _{t -1} -b _{t -2} |, BD2 = | b _{t -1} -b _t |, BD3 = | b _{t -1} -c _{t -1} |

        LD2 = Median (CD1, CD2, CD3)

CD1 = | b _t -b _{t -1} |, CD2 = | b _t -b _{t +1} |, CD3 = | b _t -c _t |

        LD3 = Median (DD1, DD2, DD3)

DD1 = | b _{t +1} -b _t |, DD2 = | b _{t +1} -b _{t +2} |, DD3 = | b _{t +1} -c _{t +1} |

Thereafter, the horizontal boundary detection unit 140 compares ADx and LDx obtained through Equation 2 and Equation 3 with a predetermined threshold value, respectively. Here, the threshold value is obtained through experiments, and may be divided into a third threshold value Th3 and a fourth threshold value Th4 and may be a reference for determining a horizontal boundary. For example, when ADx is larger than the third threshold Th3 and LDx is smaller than the fourth threshold Th4, the horizontal boundary detector 140 may determine that the pixel to be interpolated exists on the lower horizontal boundary. Can be. This is expressed as the following equation.

Horizontal boundary detection condition below: (ADx> Th3) ∧ (LDx <Th4)

Meanwhile, as shown in FIG. 5B, when the pixel X to be interpolated corresponds to the upper horizontal boundary, the difference AD3 between the upper pixel value and the lower pixel value of the pixel to be interpolated is large. The difference between the pixel value to be interpolated and the pixel value above it and the difference between the pixel value to be interpolated and the pixel value corresponding to the previous field are small. That is, the absolute value of the difference between the upper and lower pixel values of the pixel to be interpolated is large, the absolute value of the difference between the left and right pixel values of the upper pixels of the pixel to be interpolated, and the upper pixel value of the pixel to be interpolated and the corresponding value of the previous field. When the absolute value of the difference between pixel values is small, the pixel to be interpolated corresponds to the above horizontal boundary.

Accordingly, in order to detect the horizontal boundary, the horizontal boundary detector 140 obtains ADx through [Equation 2] and the absolute value of the difference between left and right pixel values of the pixels adjacent to the pixel to be interpolated through [Equation 5]. Each intermediate value HD1 using the absolute values ED3, FD3, and GD3 of the difference between (ED1, ED2, FD1, FD2, GD1, GD2) and the upper pixel value of the pixel to be interpolated and the corresponding pixel value of the previous field. , HD2 and HD3), and then again their median value (HDx).

HDx = Median (HD1, HD2, HD3)

Where HD1 = Median (ED1, ED2, ED3)

ED1 = | a _{t -1} -a _{t -2} |, ED2 = | a _{t -1} -a _t |, ED3 = | a _{t -1} -c _{t -1} |

        HD2 = Median (FD1, FD2, FD3)

FD1 = | a _t -a _{t -1} |, FD2 = | a _t -a _{t +1} |, FD3 = | a _t -c _t |

        HD3 = Median (GD1, GD2, GD3)

GD1 = | a _{t +1} -a _t |, GD2 = | a _{t +1} -a _{t +2} |, GD3 = | a _{t +1} -c _{t +1} |

Thereafter, the horizontal boundary detector 140 compares the ADx and the HDx obtained through Equations 2 and 5 with the third threshold Th3 and the fourth threshold Th4, respectively. In this case, when ADx is larger than the third threshold Th3 and HDx is smaller than the fourth threshold Th4, the horizontal boundary detector 140 may determine that the pixel to be interpolated corresponds to the upper horizontal boundary. This is expressed as the following equation.

Horizontal boundary detection condition = (ADx> Th3) ∧ (HDx <Th4)

Accordingly, the horizontal boundary detector 140 may detect the horizontal boundary below through Equation 4 and the upper horizontal boundary through Equation 6 to detect the horizontal boundary. A condition that integrates [Equation 4] and [Equation 6], that is, the horizontal boundary detection condition is represented by the following equation.

Horizontal boundary detection condition = (ADx> Th3) ∧ ((LDx <Th4) ∨ (HDx <Th4))

As described above, the horizontal boundary detection unit 140 has five pixels (15 pixels in total of 5 × 5 × 5) respectively present on the upper / lower scanning line of the pixel to be interpolated and the corresponding scanning line of the previous field. The horizontal boundary can be detected using. Therefore, accurate horizontal boundary detection can be performed even with a few pixels.

Meanwhile, the first filter unit 160 calculates each pixel value using a plurality of time axis filters included in the second filter unit 160 based on two input fields, that is, a previous field and a current field. In this case, the first filter unit 160 may use a three-point vertical-time axis (VT) median filter.

First, a method of calculating pixel values based on a three-point vertical-time axis (VT) median filter will be briefly described. The first filter unit 160 is three-point vertical-time base (VT) the median of the pixel values corresponding to the top and bottom of the pixel value and the previous field of the pixel to be interpolated to the pixel to be interpolated on the basis of the median filter (Z _VT To calculate the pixel value. This is expressed as the following equation.

Z _VT = Median (a _t , b _t , c _t )

The second filter unit 170 calculates each pixel value using a plurality of spatial axis filters included in the second filter unit 170 based on the input current field. In this case, it is preferable that the modified ELA filter is used as the second filter unit 170.

7 is a diagram for explaining the calculation of the directionality of the ELA. Hereinafter, a process of calculating pixel values based on the modified ELA filter will be described with reference to the accompanying drawings.

The second filter unit 170 obtains an absolute value of a difference between pixel values in five directions (e, f, g, h, and i) for directional calculation for the pixel to be interpolated through [Equation 9].

nD = Min (e, f, g, h, i)

Where e = | a _{t -2} -b _{t +2} |, f = | a _{t -1} -b _{t +1} |, g = | a _t -b _t |,

h = | a _{t +1} -b _{t -1} | And i = | a _{t +2} -b _{t -2} |

The direction with the smallest difference (nD) is selected as the direction with the largest correlation, and if the 'g' direction has the highest correlation, the average value between two pixels (a _t and b _t ) in the 'g' direction This is used for interpolation. On the other hand, if the 'e' direction has the highest correlation and is larger than the 'h' and 'i' directions, which are opposite to the 'e' direction, by more than a predetermined threshold value Td, the two pixels in the 'e' direction (a _{t − The} mean value between ₂ and b _{t +2} ) is used for interpolation. If e is not greater than the threshold Td greater than h and i in the opposite direction, the average value between the two pixels a _t and b _{t in the} 'g' direction is used for interpolation. Here, the threshold value Td is a value determined through experiment as a value for calculating the direction to be used for interpolation.

As described above, the second filter unit 170 calculates an average value according to the direction in order to prevent discontinuity of the pixels due to incorrect directional calculation of the pixel to be interpolated, and calculates the upper and lower pixel values of the pixel to be interpolated through Equation 10. The pixel value is calculated by calculating the median value Z _ELA of the mean values according to the direction and the direction.

Z _ELA = Median (a _t , b _t , (a _{t -2} + b _{t +2} ) / 2), if (nD = e, | eh |> Td, | ei |> Td)

= Median (a _t , b _t , (a _{t -1} -b _{t +1} ) / 2), if (nD = f, | fh |> Td, | fi |> Td)

= Median (a _t , b _t , (a _{t +1} -b _{t -1} ) / 2), if (nD = h, | he |> Td, | hf |> Td)

= Median (a _t , b _t , (a _{t +2} -b _{t -2} ) / 2), if (nD = i, | ie |> Td, | if |> Td)

= (a _t -b _t ) / 2, else

The result selector 180 uses the amount MDx of the motion information detected by the motion information detector 120 and the horizontal boundary detected by the horizontal boundary detector 140 to generate the first and second filter units 160. The pixel value to be used for interpolation is selected from among the pixel values Z _VT and Z _ELA calculated at 170. In this case, the result selector 180 may select the pixel value using the first threshold value Th1 and the second threshold value Th2 obtained through the experiment.

For example, when the amount MDx of the motion information is less than the first threshold Th1 (MDx <Th1), the result selector 180 may calculate the pixels calculated by the first and second filter units 160 and 170. Among the values, a pixel value calculated through a three-point vertical-time axis (VT) median filter that takes an intermediate value between upper and lower pixel values of a pixel to be interpolated in the current field and corresponding pixel values of the previous field may be selected. When the amount MDx of the motion information is greater than or equal to the second threshold Th2 (MDx> Th2), the result selector 180 spaces the pixel values calculated by the first and second filter units 160 and 170. The pixel value calculated through the modified ELA filter, which is an on-axis deinterlacing method, can be selected.

Also, the result selector 180 may determine the three-point vertical-time axis when the amount MDx of the motion information is greater than the first threshold Th1 and less than the second threshold Th2 (Th1 ≦ MDx ≦ Th2). VT) A value obtained by mixing the pixel value calculated through the median filter and the pixel value calculated through the modified ELA filter may be selected.

That is, when the amount MDx of the motion information is greater than or equal to the first threshold Th1 and less than the second threshold Th2 and is close to the first threshold Th1, the three-point vertical is performed as shown in [Equation 11]. -When the weighted time (VT) median filter is applied, the weighted value (a) is selected, and when the value approaches the second threshold (Th2), the weighted value (a) is applied to the value applied with the modified ELA filter. have. Here, Z is a value for interpolating the pixel.

Z = (1-a) × Z _VT + a × Z _ELA

Where a = 0, if MDx <Th1,

= 1, if MDx> Th2,

           = (MDx-Th1) / (Th2-Th1), if Th1 ≤ MDx ≤ Th2

However, when selecting a pixel value using only Equation 11, if there is a pixel X to be interpolated on the horizontal boundary, the pixel value calculated by the modified ELA filter is increased as the amount MDx of motion information is large. Will be selected. Therefore, when selecting a pixel value, the result selector 180 selects a 3-point vertical-time axis (VT) median filter when the horizontal boundary detector 140 determines that the pixel X exists on a horizontal boundary. It is desirable to.

Therefore, the motion adaptive deinterlacing apparatus 100 according to the present invention includes the first and second filter units based on the motion information detected by the motion information detector 120 and the horizontal boundary detected by the horizontal boundary detector 140. Deinterlacing may be performed by the result selector 180 selecting or mixing a pixel value suitable for interpolation among the plurality of pixel values calculated at 160 and 170.

8 is a block diagram of a motion adaptive deinterlacing apparatus according to a second embodiment of the present invention. Referring to the drawings, the motion adaptive deinterlacing apparatus 100 ′ includes a motion information detector 120 ′, a horizontal boundary detector 140 ′, a filter selector 160 ′, and a deinterlacing unit 180 ′.

Hereinafter, the functions of the motion information detector 120 'and the horizontal boundary detector 140' are the same as those of the motion information detector 120 and the horizontal boundary detector 140 according to the first embodiment of FIG. Will be omitted.

The filter selector 160 ′ selects a filter to interpolate pixels using the amount MDx of the motion information detected by the motion information detector 120 ′ and the horizontal boundary detected by the horizontal boundary detector 140 ′. do. In this case, the filter selector 160 ′ may select a filter using the first threshold Th1 and the second threshold Th2 obtained through experiments.

For example, when the amount MDx of the motion information is less than the first threshold Th1 (MDx < Th1), the filter selector 160 'may be configured to adjust the upper and lower pixel values of the pixel to be interpolated in the current field. A three-point vertical-time axis (VT) median filter can be selected that takes the median of the corresponding pixel values of the previous field. When the amount MDx of the motion information is greater than or equal to the second threshold Th2 (MDx> Th2), an ELA filter that is a deinterlacing method on the spatial axis may be selected.

Also, when the amount MDx of the motion information is greater than the first threshold Th1 and less than the second threshold Th2 (Th1 ≦ MDx ≦ Th2), the three-point vertical-time axis (VT) median filter and the ELA filter You can choose to mix these results.

In the conventional two-field motion adaptive scheme, when there are pixels to interpolate on the horizontal boundary, the amount of motion information MDx is large. Accordingly, when selecting a filter to interpolate the pixel, the filter selector 160 ′ selects a three-point vertical-time axis (VT) median filter when the pixel exists on the horizontal boundary detected by the horizontal boundary detector 140 ′. It is desirable to choose.

The deinterlacing unit 180 ′ interpolates pixels using the filter selected by the filter selecting unit 160 ′. The deinterlacing unit 180 ′ may use an ELA filter and a three-point vertical-time axis (VT) median filter to interpolate the pixel.

For example, when the filter selector 160 ′ selects an ELA filter based on the first and second thresholds, the deinterlacing unit 180 ′ interpolates pixels using the selected ELA filter, and 3-point vertical − When the time axis (VT) median filter is selected, the deinterlacing unit 180 ′ may interpolate pixels using the selected 3-point vertical-time axis (VT) median filter.

In addition, when the amount MDx of the motion information is greater than or equal to the first threshold Th1 and less than the second threshold Th2 and is close to the first threshold Th1, as shown in Equation 11, 3-point When the weight-a is applied to the value to which the vertical-time axis (VT) median filter is applied and approaches the second threshold Th2, the pixel may be interpolated by applying the weight (a) to the value to which the ELA filter is applied.

Therefore, the deinterlacing unit 180 'may select a filter selected by the filter selecting unit 160' based on the motion information detected by the motion information detector 120 'and the horizontal boundary detected by the horizontal boundary detector 140'. De-interlacing may be performed by interpolating pixels using the interpolation.

As such, the motion adaptive deinterlacing apparatus according to the present invention applies a three-point vertical-time axis (VT) median filter instead of a spatial axis filter such as an ELA filter with respect to the horizontal boundary. The interpolation error on the horizontal boundary can be minimized, and the vibration can be reduced in case of the horizontal boundary in the still area. In addition, since the interpolation error can be minimized even by a device having low hardware resources by detecting the motion information and the horizontal boundary using only two fields, the recording and playback must be performed at the same time as the DVR (Digital Video Recorder). This is more efficient on devices with limited storage space.

9 is a flowchart illustrating a motion adaptive deinterlacing method according to the present invention.

The motion information detector 120 of the motion adaptive deinterlacing apparatus 100 detects the motion information by using the input current field and the previous field (S910). At this time, if the amount MDx of the motion information detected at the X point is less than the first threshold Th1 (S915), the result selector 180 is a 3-point vertical-transmitted by the first filter unit 160. A pixel value calculated through a time axis filter such as a time axis median filter is selected (S920).

However, if the amount MDx of the motion information detected at the X point is greater than or equal to the first threshold Th1, the horizontal boundary detector 140 has a high correlation with the pixel to be interpolated (pixel at the X point). The horizontal boundary is detected using the pixels and the corresponding pixels of the previous field (S925). The result selector 180 selects a pixel value calculated through a three-point vertical-time axis (VT) median filter when the pixel exists on a horizontal boundary (S920), and when the pixel does not exist on a horizontal boundary, It is determined whether the motion information MDx in the pixel is greater than or equal to the second threshold Th2 (S930). If the motion information MDx is greater than or equal to the second threshold Th2, the pixel is present in the motion area. As a result, the result selector 180 performs a spatial axis filter such as an ELA filter modified by the second filter 170. The pixel value calculated through the selection is selected (S935). If the motion information MDx is less than the second threshold Th2, the result selector 180 selects a pixel value calculated through the vertical-time axis VT median filter and the ELA filter, and 3- The weight is applied to the value of the point vertical-time axis (VT) median filter or the weight is applied to the value of the ELA filter (S940). Thereafter, the pixel may be interpolated by using the selected result value (S945).

Accordingly, the motion adaptive deinterlacing method detects a horizontal boundary in the motion region and interpolates pixels existing on the horizontal boundary by using a three-point vertical-time axis (VT) median filter, so that the pixel to be interpolated is obtained. When present on the horizontal boundary, the image quality may be improved by minimizing an interpolation error that is incorrectly determined to exist in the moving area even though the pixel to be interpolated exists in the still area. In addition, since interpolation is possible using only two fields, interpolation errors can be minimized even in a device having low hardware resources using a two-field motion adaptive deinterlacing method.

Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art to which the present invention pertains can implement the present invention in other specific forms without changing the technical spirit or essential features, The examples are to be understood in all respects as illustrative and not restrictive.

In addition, the scope of the present invention is specified by the appended claims rather than the detailed description, and all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts are included in the scope of the present invention. Should be interpreted as

1 is a diagram illustrating a parallel scan method.

2 is a diagram for explaining a boundary area.

3 is a block diagram of a motion adaptive deinterlacing apparatus according to a first embodiment of the present invention.

FIG. 4 is a diagram illustrating pixels referred to for detecting motion information in a pixel at an X point.

FIG. 5A illustrates that pixels to be interpolated exist on a lower horizontal boundary, and FIG. 5B illustrates that pixels to interpolate exist on an upper horizontal boundary.

6 is a diagram illustrating pixels to be referred to when a pixel to be interpolated exists on a lower horizontal boundary.

7 is a diagram for explaining the calculation of the directionality of the ELA.

8 is a block diagram of a motion adaptive deinterlacing apparatus according to a second embodiment of the present invention.

9 is a flowchart illustrating a motion adaptive deinterlacing method according to the present invention.

Claims (13)

In the motion adaptive deinterlacing method, a) detecting an amount of motion information with respect to a pixel to be interpolated based on pixels present in a plurality of fields of a parallel scan scheme; b) comparing the amount of motion information with a preset first threshold value and, when the amount of motion information is equal to or greater than a first threshold value, at least one of upper and lower pixels adjacent to the pixel and pixels of a previous field corresponding to the pixel; Detecting a horizontal boundary using one; And c) selecting and interpolating a time-base filter if the pixel is on the horizontal boundary. The method of claim 1, wherein step b) b-1) calculating a first condition for horizontal boundary detection by using brightness values of upper and lower pixels adjacent to the pixel, and calculating brightness values of upper or lower pixels adjacent to the pixel and the previous field corresponding to the pixel. Calculating a second condition for horizontal boundary detection using brightness values of the pixels; b-2) detecting the horizontal boundary by comparing the first condition, the second condition, and a threshold value for determining a horizontal boundary, respectively. The method of claim 2, wherein the step b-1), The first condition is calculated by using a first intermediate value of at least one upper pixel adjacent to the pixel and an absolute value of brightness values of lower pixels corresponding to the upper pixels, respectively. De-interlacing method. The method of claim 2, wherein the step b-1), With respect to absolute values of the difference between the brightness values of each of at least one up or down pixel adjacent to the pixel, the absolute value of the brightness value difference of the pixels above or below the pixel and the brightness value difference of the corresponding pixels of the previous field And calculating the second condition using the third intermediate value, which is the intermediate value of the calculated second intermediate values, after calculating the second intermediate values, respectively. The method of claim 1, wherein after step c) d) if the pixel is not on a horizontal boundary, comparing the amount of motion information with a second threshold greater than the first threshold; And  e) selecting a spatial axis filter when the amount of motion information exceeds the second threshold value; selecting a result of mixing a spatial axis filter and a time axis filter when the amount of motion information is less than or equal to a second threshold value; And interpolating the motion adaptive deinterlacing method. In a motion adaptive deinterlacing device, A motion information detector for detecting motion information on a pixel to be interpolated based on a plurality of fields of a parallel scan method; A horizontal boundary detector for detecting horizontal boundary information using at least one of upper and lower pixels adjacent to the pixel and pixels of a previous field corresponding to the pixel when the motion information is detected to a predetermined level or more;  A first filter unit including a time axis filter interpolating the pixel based on the field; A second filter unit including a spatial axis filter interpolating the pixel based on the field; And And a result selector configured to select a result value interpolated through the first filter part or the second filter part using at least one of the motion information and the horizontal boundary information. In a motion adaptive deinterlacing device, A motion information detector for detecting motion information on a pixel to be interpolated based on a plurality of fields of a parallel scan method; A horizontal boundary detector for detecting horizontal boundary information using at least one of upper and lower pixels adjacent to the pixel and pixels of a previous field corresponding to the pixel when the motion information is detected to a predetermined level or more; A filter selecting unit which selects a filter to interpolate the pixel by using the motion information and the horizontal boundary information; And And a deinterlacing unit including a plurality of filters and interpolating the pixel using the selected filter. The method of claim 6 or 7, wherein the horizontal boundary detection unit, A first condition for horizontal boundary detection is calculated by using brightness values of upper and lower pixels adjacent to the pixel, and brightness values of upper or lower pixels adjacent to the pixel and brightness values of pixels of a previous field corresponding to the pixel. Calculates a second condition for detecting a horizontal boundary using a method, and detects the horizontal boundary information by comparing the first condition, the second condition, and a threshold value for determining a horizontal boundary, respectively. De-interlacing device. The method of claim 8, wherein the horizontal boundary detection unit, The first condition is calculated using five upper pixels adjacent to the pixel and a first intermediate value of absolute values of brightness values of lower pixels corresponding to each of the upper pixels, and the five upper pixels adjacent to the pixel are calculated. Or second intermediate values are respectively calculated with respect to absolute values of brightness values of respective lower pixels, absolute values of brightness values of upper or lower pixels of the pixels, and brightness values of corresponding pixels of a previous field, respectively. And the horizontal condition information is detected by calculating the second condition using a third intermediate value which is a median of the calculated second intermediate values. The method of claim 6, wherein the result selector, And if a pixel to be interpolated is positioned on the horizontal boundary, selecting a result value interpolated through the time axis filter included in the first filter unit. The method of claim 6, wherein the second filter unit, And a modified edge based line averaging (ELA) filter that interpolates the pixel based on the current field as a spatial axis filter. The method of claim 7, wherein the filter selector, And selecting a time axis filter when the pixel to be interpolated is located on the horizontal boundary. The method of claim 7, wherein the deinterlacing unit, And a modified edge based line averaging (ELA) filter for interpolating the pixel based on the current field as a spatial axis filter.

Images