KR20080032491A - Motion Estimator and Motion Estimation Method - Google Patents

Abstract

The present invention provides a motion vector estimation apparatus for generating an interpolation frame for frame rate conversion based on a current frame and a previous frame, wherein the motion vector estimation device generates an interpolation frame for frame rate conversion based on a current frame and a previous frame. A local search region defining unit comprising: a local search region defining unit for dividing a current frame into a plurality of blocks, and defining a plurality of local search regions for regions where there is a possibility of actual motion by using motion vectors of previous frames for each block; A local candidate vector definition unit defining a motion vector having a minimum motion prediction error value (SAD) for each local search area as a local candidate vector; A global candidate vector definition unit defining a plurality of global candidate vectors each having a minimum motion prediction error value (SAD) in each of the left and right global search regions for each block; For each of a plurality of local candidate vectors and a plurality of global candidate vectors, a sum of weight vector distance between the motion prediction error values (SAD) and the vector consistency reflecting a vector consistency, And a motion vector determiner for determining a final motion vector as a vector having a minimum cumulative value calculated based on the vector size value. Accordingly, in generating the motion vector that is the basis for generating the interpolation frame, a motion vector close to the actual motion can be generated.

Description

Motion Estimator and Motion Estimating Method

1 is a control block diagram of a motion estimation apparatus according to an embodiment of the present invention.

2A is a global search region diagram of a motion estimation apparatus according to an embodiment of the present invention.

2B is a block diagram of calculating candidate vectors for a global search region defined by a motion estimation apparatus according to an embodiment of the present invention.

3 is a local search region diagram of a motion estimation apparatus according to an embodiment of the present invention.

4 is a block diagram of calculating a candidate vector by an average vector method in a motion estimation apparatus according to an embodiment of the present invention.

5 is a block diagram of calculating a candidate vector by a row average vector method in a motion estimation apparatus according to an embodiment of the present invention.

6 is a block diagram of calculating a candidate vector by a global motion vector method in a motion estimation apparatus according to an embodiment of the present invention.

7 is a block diagram for determining a final motion vector with respect to a candidate vector defined in the motion estimation apparatus according to an embodiment of the present invention.

FIG. 8 is a graph for calculating weights for motion vector error values, cumulative distance between vector vectors, and vector size values for candidate vectors defined in the motion estimation apparatus according to an embodiment of the present invention.

9 is a graph for calculating a distance weighted cumulative value between vectors for a specific block and a neighboring block in a motion estimation apparatus according to an embodiment of the present invention.

FIG. 10 is a calculation formula for calculating a weighted cumulative value between vectors in a motion estimation apparatus according to an exemplary embodiment of the present invention.

11 is a flowchart illustrating a control method of a motion estimating apparatus according to an exemplary embodiment of the present invention.

Explanation of symbols on the main parts of the drawings

10: signal receiver 20: signal processor

31: global candidate vector definition unit 33: memory unit

35: local search area definition unit 37: local candidate vector definition unit

39: motion vector determiner 40: frame interpolator

50: display unit 100: motion estimation apparatus

The present invention relates to a motion estimating apparatus and a motion estimating method, and more particularly, to a motion estimating apparatus and a motion estimating method for estimating the motion of each frame to generate an interpolated frame to be inserted between consecutive frames.

In general, frame rate conversion refers to converting a frequency of an input video signal according to an output standard of a display. For example, in order to convert a 50Hz video signal into a 70Hz display output, a new frame is inserted between the input frames. If the input frame is simply repeated or a linear linear interpolation method is used, motion blur (moving) is performed. Blurred) phenomenon, etc. Thus, it is common to use motion estimation and compensation techniques using an improved frame rate conversion method to reduce such motion blur.

The conventional motion estimation method defines a plurality of candidate vectors and determines a final motion vector based on a motion prediction error value (SAD: Sum of Absolute Difference). The interpolation frame to be interpolated is generated based on the selected final motion vector. However, the conventional motion estimation method is a method of estimating motion under the assumption that the candidate vector represents the expected motion with respect to a predetermined reference block provided in the current frame, and the actual motion has a minimum motion prediction error value. .

However, such a conventional motion estimation method may estimate a vector that is not a real motion as a motion vector. When the candidate vector incorrectly predicts a real motion of a reference block, the motion vector may be continuously estimated, and motion may also be estimated. There is a concern that the blur phenomenon may be further increased.

Accordingly, an object of the present invention is to provide a motion vector estimating apparatus and a motion vector estimating method capable of generating a motion vector close to actual motion in generating a motion vector which is the basis for generating an interpolation frame.

In order to achieve the above object, the present invention provides a motion vector estimator for generating an interpolation frame for frame rate conversion based on the current frame and the previous frame, the current frame is divided into a plurality of blocks, each frame a previous frame A local search area defining unit that defines a plurality of local search areas for areas in which there is a possibility of actual motion by using a motion vector of? A local candidate vector definition unit defining a motion vector having a minimum motion prediction error value (SAD) for each local search area as a local candidate vector; A global candidate vector definition unit defining a plurality of global candidate vectors each having a minimum motion prediction error value (SAD) in each of the left and right global search regions for each block; For each of a plurality of local candidate vectors and a plurality of global candidate vectors, a sum of weight vector distance between the motion prediction error values (SAD) and the vector consistency reflecting a vector consistency, A motion vector estimating unit includes a motion vector determining unit that determines a vector having a minimum calculated cumulative value based on a vector magnitude value as a final motion vector.

Here, for each local search area, the local candidate vector definition unit includes a mean vector, a line mean vector, a zero vector, a global motion vector, An update vector is defined as a local candidate vector.

The global candidate vector definition unit divides the global search area of the previous frame into the left search area and the right search area, and divides the left global search vector and the right global search vector with minimum motion prediction error values for the left and right search areas. It is characterized by defining as a global candidate vector.

The motion vector estimating apparatus according to the present invention further includes a memory unit for storing the motion vector of the previous frame, and the local candidate vector definition unit reads the motion vector of the previous frame from the memory unit.

Here, the motion vector of the previous frame is characterized by including a final motion vector for the previous frame.

According to an aspect of the present invention, there is provided a control method of a motion vector estimator for generating an interpolation frame for frame rate conversion based on a current frame and a previous frame, the method comprising: dividing a current frame into a plurality of blocks; Defining a plurality of local search areas for the area where there is a high possibility of actual motion using the motion vector of the previous frame; Defining a local candidate vector with a motion vector having a minimum motion prediction error value (SAD) in each local search area; Defining a plurality of global candidate vectors each having a minimum motion prediction error value (SAD) in each of the left and right global search regions for each block; Calculating a motion prediction error value for each of the plurality of local candidate vectors and the plurality of global candidate vectors; Calculating a distance weighted cumulative value between vectors based on the correlation between the plurality of local candidate vectors and the plurality of global candidate vectors based on the neighboring vectors; Calculating a vector magnitude value for each of the plurality of local candidate vectors and the plurality of global candidate vectors; Defining as a final motion vector a motion prediction error value, a weighted cumulative distance between vectors, and a vector having a smallest cumulative value calculated for each of the local candidate vectors and the global candidate vectors, respectively; It is also achieved by a control method of a motion vector estimating apparatus comprising a.

Hereinafter, with reference to the accompanying drawings will be described in detail through a preferred embodiment of the present invention.

1 is a control block diagram of a motion estimation apparatus according to an embodiment of the present invention.

As shown in FIG. 1, the motion estimation apparatus according to the present invention generates an interpolation frame based on the current frame and the previous frame. To this end, the motion estimation apparatus according to the present invention includes a global candidate vector defining unit 31, a memory unit 33, a local search area defining unit 35, a local candidate vector defining unit 37, and a motion vector. The decision unit 39 is included. In addition, the motion estimation apparatus according to the present invention may further include a signal receiver 10, a signal processor 20, and a frame interpolator 40. In addition, the motion estimation apparatus according to the present invention may further include a display unit 50 as necessary.

The signal receiver 10 receives a video signal from a predetermined external video source. That is, the signal receiver 10 may include terminals corresponding to various video signal formats so that video signals of various formats can be input. The signal receiver 10 according to the present invention includes a composite input terminal for receiving a composite signal, an S-video input terminal for receiving an S-video signal, a component input terminal for receiving a component signal, a PC signal and It may include at least one of a PC input terminal and a TV input terminal for receiving a TV signal, respectively. Here, the image signal received through the signal receiver 10 is displayed on the display unit 50 for each frame.

The signal processor 20 processes the display unit 50 to output an image signal input through the signal receiver 10. Here, the signal processing unit 20 processes an image corresponding to the video signal to be displayed on the display unit 50 under the control of a processing controller not shown.

In FIG. 1, the signal processing unit 20 is shown at the input terminal side of the global candidate vector defining unit 31 and the local candidate vector defining unit 37, but the signal processing unit 20 is the global candidate vector defining unit 31 and the local unit. Of course, it can be provided on the output terminal side of the candidate vector defining unit 37.

Here, the global candidate vector defining unit 31 divides the current frame into a plurality of blocks and defines a plurality of global candidate vectors each having a minimum motion prediction error value SAD in each of the left and right global search areas.

Also, the global candidate vector defining unit 31 divides the global search area of the previous frame into the left search area and the right search area, and the left global search vector and the right with the minimum motion prediction error values for the left and right search areas. It defines global search vector as global candidate vector.

The memory unit 33 stores the final motion vector calculated for each frame. In this case, the memory unit 33 may store the final motion vector for the plurality of frames as the previous motion vector, but may store only the final motion vector for the frame immediately before the current frame as the previous motion vector.

The local search area definer 35 divides the current frame into a plurality of blocks, and uses the motion vector of the previous frame of the memory unit 33 as the local search area for each block. define.

The local candidate vector defining unit 37 defines a motion vector having a minimum motion prediction error value SAD for each local search area defined by the local search area defining unit 35 as a local candidate vector.

The method of defining the global candidate vector and the local candidate vector will be described later.

,

,

,

,

,

,

}가 있으며,

는 전역 탐색에 의해 산출된 벡터이고,

는 전역 탐색 벡터가 위치한 영역의 반대쪽(왼쪽 또는 오른쪽) 영역에서 전역 탐색에 의해 산출된 벡터이고,

는 현재블록과 그 주변블록의 이전 프레임 벡터를 이용하여 정의한 영역에서 국부영역 탐색에 의해 산출된 벡터이고,

는 현재블록과 같은 수직 위치에 있는 블록의 이전 프레임 벡터를 이용하여 정의한 영역에서 국부영역 탐색에 의해 산출된 벡터,

는 영 벡터를 중심으로 정의한 영역에서 국부영역 탐색에 의해 산출된 벡터이다. 그리고

는 전역 움직임 벡터(global motion vector)를 중심으로 정의한 영역에서 국부영역 탐색에 의해 산출된 벡터이고,

는 평균벡터에 일정 크기의 수평, 수직 성분을 더한 벡터, 현재블록의 이전 프레임 벡터, 그리고 현재블록에서 수평방향으로 3번째 이전블록과 3번째 이후블록의 이전 프레임 벡터 중 최소 SAD값을 갖는 벡터이다.The motion vector determiner 39 calculates the distance weighted cumulative value and the vector size value between the local candidate vectors and the global candidate vectors, which reflect the correlation between the motion prediction error value (SAD) and the surrounding vector (Vector Consistency). The final calculation is based on a vector having a minimum cumulative value calculated as a motion vector. The vector used in the motion vector determination process is {

,

,

,

,

,

,

},

Is the vector produced by the global search,

Is the vector produced by the global search in the region opposite (left or right) of the region where the global search vector is located,

Is a vector computed by local region search in the region defined by the previous frame vector of the current block and its neighbors,

Is a vector computed by local region search in the region defined using the previous frame vector of the block at the same vertical position as the current block,

Is a vector calculated by searching for a local region in a region defined around a zero vector. And

Is a vector calculated by searching for a local region in a region defined around a global motion vector,

Is a vector having a minimum SAD value of a mean vector plus a horizontal and vertical component of a predetermined size, a previous frame vector of the current block, and a previous frame vector of the third previous block and the third subsequent block in the horizontal direction from the current block. .

A method of determining the motion vector by the motion vector determiner 39 will be described later.

)의 탐색영역(S

)을 좌측 탐색영역(S

)과 우측 탐색영역(S

)로 나누고, 도 2b에서와 같이 현재 블록과 가장 유사한 블록을 블록 정합방법(Block Matching Algorithm)으로 찾아 두 블록 사이의 공간상 위치 이동을 각각 좌측 전역탐색 벡터(

)와 우측 전역탐색 벡터(

)로 정의한다. 블록간 유사성을 측정하기 위한 척도로서 일반적으로 SAD(Sum of Absolute Difference)값이 사용되며, 다음과 같이 계산한다.The global candidate vector defining unit 31 is composed of the previous frame f as shown in FIG. 2A.

Navigation area (S)

) To the left navigation area (S

) And right navigation area (S

), And as shown in FIG. 2B, the block most similar to the current block is found by the block matching algorithm, and the spatial positional movement between the two blocks is calculated using the left global search vector (

) And the right global search vector (

To be defined). In general, a SAD (Sum of Absolute Difference) value is used as a measure for measuring similarity between blocks, and is calculated as follows.

[Equation 1]

는 현재블록 B에 속한 화소의 좌표를 나타내며,

는 탐색영역 내에서 현재블록으로부터 떨어진 상대적 위치를 나타낸다. 그리고 탐색영역 S내에서 SAD값을 최소화하는 벡터를 다음과 같이 움직임 벡터로 결정한다.here,

Represents the coordinates of the pixel belonging to the current block B,

Denotes a relative position away from the current block in the search area. A vector that minimizes the SAD value in the search area S is determined as a motion vector as follows.

[Equation 2]

)과 우측 전역탐색 벡터의 SAD값(

)을 비교하여 다음 수식에 의해 전역탐색 벡터(

)와 반대방향의 전역탐색 벡터(

)를 결정한다.The SAD value of the left global search vector defined in this way (

) And the SAD value of the right global search vector (

) And compare the global search vector (

Global search vector in the opposite direction to

Is determined.

[Equation 3]

은 국부탐색영역이고,

는 각 국부탐색영역에 대한 좌표값을 나타낸다.As illustrated in FIG. 3, the local search area definer 35 defines an area having a high probability of actual motion as a local search area using the motion vector of the previous frame. In Figure 3

Is the local search area,

Denotes a coordinate value for each local search area.

The local candidate vector definition unit 37 selects a motion vector having a minimum SAD value as a candidate vector for the local search region defined by the local search region definition unit 35. The candidate vectors calculated by the local candidate vector definition unit 37 include a mean vector, a line mean vector, a zero vector, a global vector, and an updater vector. Vector), respectively.

~ S

)은 현재블록(4)과 주변 8개 블록(0,1,2,3,5,6,7,8)으로부터 일부를 선택하여 정의된다. 각 서브그룹별로 해당 블록의 이전 프레임의 움직임 벡터로부터 평균 벡터(

:

,

,

,

,

)를 산출하고 평균 벡터와 서브 그룹에 속한 블록의 벡터간 거리를 평균하여 벡터간 거리 평균값(

:

,

,

,

,

)을 계산한다. 벡터간 거리 평균값이 가장 작은 서브 그룹의 평균 벡터를 현재블록의 평균 벡터(

)로 다음과 같이 정의한다.In the average vector method, the local candidate vector definition unit 37 reads the motion vector of the previous frame from the memory unit 33, divides the current frame into a plurality of blocks, and sets one of the divided blocks as the current block. The local candidate vector defining unit 37 of the average vector method has five subgroups (S) as shown in FIG.

~ S

) Is defined by selecting a part from the current block 4 and the surrounding eight blocks (0,1,2,3,5,6,7,8). For each subgroup, the average vector (from the motion vector of the previous frame of the block)

:

,

,

,

,

) And average the distance between the vectors of the blocks belonging to the subgroup with the mean vector (

:

,

,

,

,

Calculate The average vector of the subgroup having the smallest average value of the distance between vectors is used as the average vector of the current block (

) Is defined as:

[Equation 4]

)를 중심으로 한 국부 탐색영역에서 최소 SAD값을 갖는 벡터를 다음과 같이 후보벡터로 결정한다.This mean vector (

In the local search area centered on), a vector having a minimum SAD value is determined as a candidate vector as follows.

[Equation 5]

)보다 크고 주변벡터와의 거리 평균값이 임계값(

)보다 작은 벡터만을 평균하여 평균벡터(

)를 다음과 같이 계산한다.As shown in FIG. 5, the local candidate vector defining unit 37 of the row average vector method has a threshold value of the motion size among the previous frame vectors of the block at the same vertical position as the current block.

Greater than) and the average distance from

Average only vector less than

) Is calculated as

[Equation 6]

은 이전 행 움직임벡터이고,

은 이전 행 움직임벡터의 주변블록에 대응하는 이전 움직임벡터이고, 현재블록 및 주변블록의 개수가 N개인 경우이다. L은 주변블록의 개수이다.here,

Is the previous row motion vector,

Is the previous motion vector corresponding to the neighboring block of the previous row motion vector, and the number of the current block and the neighboring block is N. L is the number of neighboring blocks.

)를 중심으로 한 국부탐색영역에서 최소 SAD값을 갖는 벡터를 다음과 같이 후보벡터

로 결정한다.Calculated mean vector (

In the local search area centered on

Decide on

[Equation 7]

)를 중심으로 한 국부 탐색영역에서 SAD값을 최소화하는 벡터를 다음과 같이 후보벡터

로 결정한다.The local candidate vector definition unit 37 of the zero vector method is for estimating the vector of the stationary region, and the zero vector (

In the local search area centered on

Decide on

[Equation 8]

The local candidate vector defining unit 37 of the global motion vector system is a vector representing the motion of the entire screen by camera movement (camera panning, tilting, zooming, etc.). The global motion vector is defined by the following equation, and is calculated by a modeling technique by sampling some blocks of the previous frame vector as shown in FIG.

[Equation 9]

,

는 움직임 벡터의 x성분 값과 y성분 값이며, n, m는 수평방향과 수직 방향의 블록 index를 나타낸다. 그리고 a_x, a_y는 각각 수평, 수직방향의 zooming factor, b_{x ,} b_y는 수평, 수직방향의 panning factor를 나타낸다. 산출된 전역 움직임 벡터를 중심으로 한 국부 탐색영역에서 최소 SAD값을 갖는 벡터를 다음과 같이 후보벡터로 결정한다.

,

Are the x and y component values of the motion vector, and n and m represent block indexes in the horizontal and vertical directions. And a _x and a _y denote horizontal and vertical zooming factors, and b _{x and} b _y denote horizontal and vertical panning factors, respectively. A vector having a minimum SAD value in a local search region centered on the calculated global motion vector is determined as a candidate vector as follows.

[Equation 10]

)에 국부 탐색영역 범위(X_L, Y_L)보다 큰 수평성분(mx), 또는 수직성분(my)을 각각 더하거나 뺀 벡터(

,

,

,

), 현재블록(m,n)의 이전 프레임 벡터(

), 그리고 현재블록에서 수평방향으로 3번째 이전블록(m,n-3) 및 3번째 이후블록(m,n+3)의 이전 프레임 벡터(

,

) 중 최소 SAD값을 갖는 벡터를 다음과 같이 후보벡터로 정의한다.The local candidate vector definition unit 37 of the updater vector method uses a vector for fast vector convergence as a candidate vector. The local candidate vector defining unit 37 is an average vector calculated by Equation (4).

) Plus or minus the horizontal component (mx) or vertical component (my), respectively, greater than the local search range (X _L , Y _L ).

,

,

,

), The previous frame vector of the current block (m, n)

) And the previous frame vector of the third previous block (m, n-3) and the third subsequent block (m, n + 3) in the horizontal direction from the current block (

,

), The vector having the smallest SAD value is defined as a candidate vector as follows.

[Equation 11]

,

,

,

,

,

,

}에 대해 각각 움직임 추정 비용함수를 계산하여 이를 최소화하는 벡터를 최종 움직임 벡터로 결정한다.The motion vector determiner 39 includes seven candidate vectors computed by the global candidate vector definer 31 and the local candidate vector definer 37.

,

,

,

,

,

,

} And the motion estimation cost function for each is determined as the final motion vector.

), 주변벡터와의 상관성을 나타내는 벡터간 거리 가중누적값(

), 그리고 벡터 크기 (l(

)) 측정값으로 구성된다. 세 측정값을 하나의 비용함수로 묶기 위해, 각 측정값의 신뢰도로부터 가중치를 산출한다. 이러한 가중치는 매 블록마다 계산되며, 한 블록의 모든 후보벡터에 대해서는 동일한 값으로 적용된다. SAD값의 가중치(

)는 현재블록의 gradient값(

)으로부터 계산되며, gradient값이 작은 블록은 SAD값의 변별력이 상대적으로 떨어지므로, SAD값의 변별력을 보상하기 위해서 큰 가중치가 할당되도록 수학식 12와 같이 조정된다. 이 가중치를 구하는 함수는 제 8도의 첫번 째 그래프와 같이 단조 감소 함수의 형태를 띠게 된다.As shown in FIG. 7, the cost function is a SAD value representing the degree of matching between blocks (

), The weighted cumulative value between vectors representing the correlation with the surrounding vector (

), And the vector size (l (

)) Consists of measured values. To combine the three measurements into one cost function, the weights are calculated from the reliability of each measurement. This weight is calculated for each block, and the same value is applied to all candidate vectors of one block. Weight of the SAD value (

) Is the gradient value of the current block (

Since the discrimination power of the SAD value is relatively low, the block having the small gradient value is adjusted as shown in Equation 12 so that a large weight is assigned to compensate for the discrimination power of the SAD value. This weighting function takes the form of a monotonic reduction function as shown in the first graph of FIG.

[Equation 12]

)로 SAD값을 나주어 줌으로써 블록정합 오류값을 정규화한다.The standard deviation of the block calculated by

Normalize the block matching error value by giving the SAD value.

[Equation 13]

)은 도 9, 도 10과 같이 현재블록과 주변블록의 이전 프레임 벡터(

~

)와 후보벡터(

)간의 거리를 미리 정의된 블록간 거리별 가중치(w0 ~ w8)로 곱하여 수학식 14와 같이 계산한다. 블록간 거리별 가중치는 정의된 블록 크기에 따라 블록 중심간의 거리에 비례하도록 정의된다. 이러한 벡터간 거리 가중누적값은 인접한 블록의 움직임 벡터는 서로 유사하다는 영상 특성을 이용한 측정값으로, 실제 움직임 벡터를 추정할 때 유용한 지표이다.Cumulative cumulative distance between vectors

) Is the previous frame vector ((b) of the current block and the neighboring blocks as shown in FIGS.

To

) And candidate vectors (

) Is multiplied by a predefined weight of each inter-block distance (w0 to w8) and calculated as shown in Equation 14. The inter-block weights are defined to be proportional to the distance between block centers according to the defined block size. The distance weight cumulative value between the vectors is a measurement value using the image characteristic that the motion vectors of adjacent blocks are similar to each other, and are useful indicators when estimating the actual motion vectors.

[Equation 14]

) 대신에 영벡터(

)를 대입하여 수학식 15와 같이 계산한 참조벡터(

~

)의 크기 가중누적값(

)을 나누어줌으로써 벡터간 거리 가중누적값을 정규화한다. 이러한 정규화 과정을 통해 움직임이 빠른 영역이나 정지에 가까운 움직임이 느린 영역에서 주변벡터와의 상관성 정도를 동일하게 표시할 수 있다.In addition, in Equation 14, the candidate vector (

Zero vector instead of

) And the reference vector (

To

) Weighted cumulative value (

) To normalize the weighted cumulative distance between vectors. Through this normalization process, the degree of correlation with the surrounding vector can be expressed in the same region in the fast motion region or in the slow motion region.

[Equation 15]

)는 벡터간 거리 가중누적값을 계산할 때 이용되는 참조벡터의 상관성 정도(

)에 의해 수학식 16과 같이 계산된다. 참조벡터의 상관성 정도는 참조벡터의 평균벡터와 참조벡터간의 거리 누적값(

)을 참조벡터 크기의 누적값(

)으로 나누어 계산된다. 즉, 벡터의 크기에 비해 벡터간 상관성이 떨어지는 블록은 전경과 배경이 교차하여 인접블록간 벡터가 상이한 벡터필드의 불연속 경계에 해당한다. 이러한 블록에서는 참조벡터가 서로 일관되지 못하므로, 벡터의 상관성을 반영하는 측정값인 벡터간 거리 가중누적값은 더 이상 벡터 결정에 유효하지 않다. 따라서 이러한 블록에 대해서는 벡터간 거리 가중누적값이 비용함수에 반영되지 못하도록 가중치를 줄여 벡터간 거리 가중누적값이 감소하도록 도 8의 두번째 그래프와 같이 단조 감소함수의 형태를 갖게 된다.Weights of distance-weighted cumulative values between vectors (

) Is the degree of correlation of the reference vector used when calculating the weighted cumulative value

Is calculated as shown in Equation (16). The degree of correlation of the reference vector is the cumulative value of the distance between the reference vector's average vector and the reference vector (

) Is the cumulative value of the reference vector size (

It is calculated by dividing by. That is, a block having low correlation between vectors compared to the size of a vector corresponds to a discontinuous boundary of a vector field in which the foreground and background cross each other and the vectors between adjacent blocks differ. Since the reference vectors are not consistent with each other in such a block, the distance weight accumulation value between vectors, which is a measure reflecting the correlation of the vector, is no longer valid for vector determination. Therefore, for such a block, the weighted cumulative cumulative value between the vectors is reduced so that the weighted cumulative cumulative value between the vectors is reduced so as not to be reflected in the cost function.

[Equation 16]

))는 후보벡터의 수평, 수직성분의 절대값을 합하여 계산한 값으로 다음과 같이 계산된다.And, vector size (l (

)) Is calculated by summing the absolute values of the horizontal and vertical components of the candidate vector.

[Equation 17]

)는 영 벡터의 SAD값(

)과 전역탐색의 최소 SAD값(

)의 차이로부터 계산되며, 두 SAD값의 차이가 작을수록 벡터 크기의 가중치가 증가하도록 수학식 18과 같이 계산한다. 이는 영 벡터의 SAD값과 최소 SAD값의 차이 정도에 따라 영 벡터로부터 떨어진 거리, 즉 벡터 크기를 가중함으로써 영 벡터의 SAD값이 최소 SAD값과 유사한 경우, 상대적으로 벡터 크기가 작은 벡터가 선택되도록 정지 영역의 선택비중이 가중되게 한다.Weight of vector size (

) Is the SAD value of the zero vector (

) And the minimum SAD value of global search (

The difference between the two SAD values is calculated as shown in Equation 18 so that the weight of the vector size increases. This is done by weighting the distance away from the zero vector, that is, the vector size according to the difference between the SAD value of the zero vector and the minimum SAD value, so that when the SAD value of the zero vector is similar to the minimum SAD value, a relatively small vector size is selected. This increases the selection weight of the still area.

Equation 18

,

,

)를 후보벡터별로 계산한 세 측정값(

(

)/

, d

(

)/d

(

), l(

))에 곱하여 합산한 비용함수를 최소화하는 벡터를 수학식 19와 같이 움직임 벡터로 최종 결정한다.Equivalent weights calculated in blocks (

,

,

), Three measures (for each candidate vector)

(

) /

, d

(

) / d

(

), l (

A vector that minimizes the cost function multiplied by)) is finally determined as a motion vector as shown in Equation 19.

[Equation 19]

If the reliability of a particular measurement in the motion estimation cost function falls, the weight of the measurement is reduced by weighting. Therefore, by adjusting the weight by the reliability of each measurement value in the vector determination by the motion estimation cost function, the wrong measurement value is excluded, which enables accurate motion vector estimation.

The final motion vector determined by the motion vector determiner 39 in the above-described manner is the basis for generating the interpolation frame. That is, the final motion vector determined by the motion vector determiner 39 is used by the frame interpolator 40 to generate a new interpolated frame. That is, the frame interpolator 40 interpolates a new interpolation frame inserted between the current frame and the previous frame based on the final motion vector.

The display unit 50 receives an image signal received through the signal receiver 10 and displays an image on the screen for each frame. The display unit 50 includes a display module (not shown) in which an image is displayed, and a module driver (not shown) for processing an image signal to display an image on the display module.

The display module according to the present invention is applicable to various types of display modules such as a cathode ray tube (CRT), a digital light processing (DLP), a liquid crystal display (LCD), a plasma display panel (PDP), and the like. Here, when the display module is a DLP, the module driver includes an optical engine. When the display module is an LCD, the module driver includes a printed circuit board for converting an image signal input from a color converter into a data signal and a gate signal. do. Similarly, each display unit 50 may have a configuration of a corresponding module driver according to the type of each display module.

11 is a flowchart illustrating a control method of a motion estimating apparatus according to an exemplary embodiment of the present invention.

As shown in FIG. 11, the motion vector estimation method according to the present invention is a control method of a motion vector estimating apparatus for generating an interpolation frame for frame rate conversion based on a current frame and a previous frame, wherein the current frame is divided into a plurality of blocks. Divided (S10).

Using the motion vector of the previous frame, a plurality of local search areas are defined for an area where there is a high possibility of actual motion (S20). A motion vector having a minimum motion prediction error value (SAD) in each local search area is defined as a local candidate vector (S30). Each block defines a plurality of global candidate vectors each having a minimum motion prediction error value (SAD) in the left and right global search regions (S40). A motion prediction error value is calculated for each of the plurality of local candidate vectors and the plurality of global candidate vectors (S50). For each local candidate vector and each global candidate vector, a weighted cumulative value between the vectors based on the correlation between the neighboring vectors is calculated (S60). A vector size value is calculated for each local candidate vector and each global candidate vector (S70). For each local candidate vector and each global candidate vector, a vector having a minimum cumulative value calculated based on a motion prediction error value, a distance accumulation vector between the vectors, and a vector size value is defined as a final motion vector (S80).

Although some embodiments of the invention have been shown and described, it will be apparent to those skilled in the art that modifications may be made to the embodiment without departing from the spirit or spirit of the invention. will be. It is intended that the scope of the invention be defined by the claims appended hereto and their equivalents.

As described above, according to the present invention, there is provided a motion vector estimating apparatus and a motion vector estimating method capable of generating a motion vector close to actual motion in generating a motion vector based on generating an interpolation frame.

Claims (6)

A motion vector estimator for generating an interpolation frame for frame rate conversion based on a current frame and a previous frame, A local search area definition unit for dividing the current frame into a plurality of blocks and defining a plurality of local search areas for areas where there is a possibility of actual motion by using the motion vector of the previous frame for each block; A local candidate vector definition unit defining a motion vector having a minimum motion prediction error value (SAD) for each local search area as a local candidate vector; A global candidate vector definition unit defining a plurality of global candidate vectors each having a minimum motion prediction error value (SAD) in each of the left and right global search regions for each block; A sum of weight vector distance between each of the plurality of local candidate vectors and the plurality of global candidate vectors, reflecting a vector prediction of a motion prediction error value (SAD) and a surrounding vector. And a motion vector determiner for determining a final motion vector as a final vector, the cumulative value being calculated based on the vector magnitude value. The method of claim 1, For each local search region, the local candidate vector definition unit includes a mean vector, a line mean vector, a zero vector, a global vector, and an updater. A motion vector estimator, wherein a vector is defined as a local candidate vector. The method of claim 1, The global candidate vector defining unit divides the global search region of the previous frame into a left search region and a right search region, and the left global search vector and the right global region having the minimum motion prediction error values for the left search region and the right search region. A motion vector estimator, wherein a search vector is defined as a global candidate vector. The method according to any one of claims 1 to 3, Further comprising a memory unit for storing the motion vector of the previous frame, And the local candidate vector definition unit reads a motion vector of a previous frame from the memory unit. The method of claim 4, wherein And a motion vector of the previous frame includes a final motion vector for the previous frame. In the control method of the motion vector estimator for generating an interpolation frame for frame rate conversion based on the current frame and the previous frame, Dividing the current frame into a plurality of blocks; Defining a plurality of local search areas for a region in which there is a high possibility of actual motion using the motion vector of the previous frame; Defining a local candidate vector with a motion vector having a minimum motion prediction error value (SAD) in each local search area; Defining a plurality of global candidate vectors each having a minimum motion prediction error value (SAD) in each of the left and right global search regions for each block; Calculating a motion prediction error value for each of the plurality of local candidate vectors and the plurality of global candidate vectors; Calculating a weighted cumulative value between vectors based on a correlation between a plurality of local candidate vectors and each of the plurality of global candidate vectors based on a peripheral vector; Calculating a vector magnitude value for each of the plurality of local candidate vectors and the plurality of global candidate vectors; The final motion of the local candidate vectors and the global candidate vectors, respectively, is a final motion vector having a minimum cumulative value calculated based on the motion prediction error value, the weight accumulation value between the vectors, and the vector size value. The control method of the motion vector estimating apparatus comprising the step of defining as a vector

Images