JPH0818978A - Motion vector detector - Google Patents

Abstract

PURPOSE:To prevent an unnatural image from being obtained when an image gradient is small. CONSTITUTION:A moving quantity estimating part 100 estimates the moving quantity of a dynamic image based on a gradient method. An estimate output part 200 outputs the estimate X1 of the moving quantity estimating part 100 as the estimate X2 of the moving quantity Vx when the denominator term SIGMADELTAx of an estimation arithmetic equation to estimate the mowing quantity of the dynamic image is larger than a prescribed value (alpha), and outputs zero supplied to an input terminal 330 when the former is smaller than the latter.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION This invention is based on the gradient method.
The present invention relates to a motion vector detection device that detects a motion vector of a moving image by estimating the amount of motion of the moving image.

[0002]

2. Description of the Related Art Generally, when performing interframe coding processing for high efficiency coding or field interpolation processing for field number conversion, a motion vector indicating the direction and magnitude of motion of a moving image is used. To be

Various methods have been conventionally considered as methods for detecting the motion vector. One of them is a method using a gradient method. This method is based on a spatial gradient of an image signal (hereinafter, referred to as “image gradient”) and a difference value in a time axis direction (for example, an inter-field difference value) in a horizontal direction and a vertical direction of a moving image. The motion vector of a moving image is detected by estimating the amount of motion.

For details of this method, see, for example, "Journal of the Television Society, Vol. 45, No. 12, pp. 1534 to 1543 (1.
991) ”, but the explanation is as follows.

Now, it is assumed that the moving image I (x, y) has moved by the motion vector V = (V _x , V _y ) in one field time. In this case, the moving image after the movement is I (x−V _x , y
Represented as -V _y). Here, x indicates a horizontal coordinate and y indicates a vertical coordinate. Further, V _x indicates a horizontal movement amount of the moving image, and V _y indicates a movement amount in the water vertical direction.

At this time, if V is small,

[Equation 2] The first-order approximation formula expressed by However, D (x,
y) is an inter-field difference value.

This equation (1) is applied to m pixels in the horizontal direction and n lines in the vertical direction (m and n are positive integers of 1 or more).
Solving for V _x and V _y under the condition of the minimum squared error for each pixel block consisting of

(Equation 3) Is obtained.

However, in this case,

[Equation 4] Is assumed to hold.

Here, Δ _x indicates the image gradient in the horizontal direction, Δ _y indicates the image gradient in the vertical direction, and

(Equation 5) It is expressed as

Equations (2) and (3) are

(Equation 6) Can be approximated as

In the gradient method, the operations represented by the equations (7) and (8) are executed for each pixel block,
The motion amounts V _x and V _y of the moving image are estimated.

In the gradient method, when the motion amounts V _x and V _y are large, the above-described estimation calculation is repeated several times in order to improve estimation accuracy. This method is usually called the iterative gradient method.

Further, in the actual iterative gradient method, in order to be able to estimate the motion amounts V _x and V _y without being restricted to the estimation range of the motion amounts V _x and V _y , prior to the calculation, A predetermined initial displacement vector is prepared, and the estimation calculation described above is repeated starting from this vector.

[0014]

However, in the conventional motion vector detecting device using the gradient method, when the image gradients Δ _x and Δ _y are small, that is, the denominators of the estimation arithmetic expressions (7) and (8). The terms Σ [Δ _x ] and Σ [Δ _y ] ([Δ
_x ] and [Δ _y ] represent absolute values of Δ _x and Δ _y ), a calculation error occurs and an inappropriate value is obtained as an estimated value of the motion amounts V _x and V _y. . As a result, in the conventional motion vector detecting device, the image gradients Δ _x and Δ _y
When is small, there is a problem that the image is distorted.

Since this problem is caused by a calculation error in the estimation calculation, it is not limited to the gradient method in which the estimation calculation is performed only once, but the gradient method is performed several times, that is, the iterative gradient method or the initial deviation vector. Also occurs in the iterative gradient method using.

[0016]

In order to solve the above-mentioned problems, the present invention provides a means for estimating a motion amount of a moving image based on an image gradient and a difference value in the time axis direction, and a motion of the moving image. As the estimated value of the amount, means for outputting either the estimated value of the motion amount estimating means or a predetermined value based on the magnitude of the image gradient is provided.

[0017]

In the above construction, when the image gradient is larger than the predetermined value, the estimated value of the motion amount estimating means is output as the estimated value of the moving amount of the moving image. On the other hand, when the image gradient is smaller than the predetermined value, a predetermined value is output.
Accordingly, even if the image gradient is small, it is possible to prevent an inappropriate value from being obtained as the estimated value of the motion amount, and thus it is possible to prevent image distortion. .

[0018]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention. In addition, in FIG. 1, among the portions for estimating the horizontal motion amount V _x and the vertical motion amount V _y of the moving image, the portion for estimating the horizontal motion amount V _x is shown as a representative.

The illustrated motion vector detecting device comprises a motion amount estimating section 100 and an estimated value output section 200. The motion amount estimation unit 100, based on the above-described estimation calculation formula (7),
A horizontal motion amount V _x of a moving image is estimated. The estimated value output unit 200, when the horizontal image gradient Δ _x , specifically, the denominator Σ [Δ _x ] on the right side of the estimation calculation formula (7) is larger than the predetermined value α, the motion amount estimation unit 100. The estimated value X1 of is output as the estimated value X2 of the motion amount V _x , and when it is small, a predetermined value, for example, 0 is output.

The motion amount estimating section 100 uses the estimation calculation formula (7).
Numerator term calculation unit 110 for obtaining the numerator term ΣD · SGN (Δ _x ) on the right side of and denominator calculation unit 12 for obtaining the denominator term Σ [Δ _x ]
0 and the calculation output of the numerator term calculation unit 110 is calculated as
And a division circuit 130 for division by 20 operation outputs.

The numerator operation unit 110 includes a subtraction circuit 111.
And a sign multiplication circuit 112 and a cumulative addition circuit 113. The denominator term calculation unit 120 includes latch circuits 121 and 12
2, a subtraction circuit 123, an image gradient calculation circuit 124,
And a cumulative addition circuit 125.

The estimated value output unit 200 includes a denominator term calculation unit 12
A determination circuit 210 for determining whether or not the denominator term Σ [Δ _x ] output from 0 is larger than a predetermined value α;
Based on the determination result of, the estimated value X2 of the motion amount V _x is calculated,
The motion amount estimation unit 100 includes a selection circuit 220 that selects either the estimated value X1 or 0.

In FIG. 1, reference numeral 310 is an input terminal to which a current field image signal (hereinafter referred to as "current field signal") A is supplied, and 320 is a previous field image signal (hereinafter referred to as "current field signal"). It is an input terminal to which a "pre-field signal" is supplied. 330 is an input terminal to which 0 is supplied. An input terminal 340 is supplied with a predetermined value α. The predetermined value α can be adjusted as needed. Reference numeral 350 is an output terminal to which the estimated value X2 of the motion amount V _x is supplied.

The operation of the above configuration will be described. First, the motion amount estimating operation of the motion amount estimating unit 100 will be described.
The current field signal A supplied to the input terminal 310 and the previous field signal B supplied to the input terminal 320 are supplied to the subtraction circuit 111 of the numerator operation unit 110. The subtraction circuit 111 outputs, for example, the current field signal A for each pixel.
From the previous field signal B. As a result, the inter-field difference value D is obtained.

This inter-field difference value D is supplied to the sign multiplication circuit 112, and the sign SGN (Δ _x ) of the image gradient Δ _x is supplied.
Is multiplied by. The code SGN (Δ _x ) is obtained by the image gradient calculation circuit 124 described later.

The multiplication output of the sign multiplication circuit 112 is supplied to the cumulative addition circuit 113. The cumulative addition circuit 113 cumulatively adds the multiplication outputs of all the pixels included in each pixel block. As a result, the estimated calculation formula (7)
The numerator term ΣD · SGN (Δ _x ) on the right side of is obtained. The numerator term ΣD · SGN (Δ _x ) is supplied to the division circuit 130.

The previous field signal B supplied to the input terminal 320 is further supplied to the latch circuit 1 of the denominator operation unit 120.
21 and the subtraction circuit 123. Latch circuit 12
The previous field signal B supplied to 1 is delayed by one pixel and then further delayed by one pixel by the latch circuit 122.

The previous field signal B delayed by two pixels by the latch circuits 121 and 122 is supplied to the subtraction circuit 123. The subtraction circuit 123, for each pixel, for example,
The preceding field signal B delayed by two pixels is subtracted from the preceding field signal B not delayed. Thus, the difference value 2.DELTA. _X 2 pixels in a field is obtained.

This difference value 2Δ _x is calculated by the image gradient calculation circuit 1
24. The image gradient calculation circuit 124, based on the difference value 2.DELTA. _X, for each pixel, obtains the image gradient Δx and the sign SGN (Δ _x). The code SGN (Δ _x ) is
It is supplied to the sign multiplication circuit 112 and is multiplied by the inter-field difference value D as described above.

On the other hand, the image gradient Δ _x is supplied to the cumulative addition circuit 125. Cumulative addition circuit 125, for each pixel block, after the absolute value of the image gradients delta _x of all pixels included in the pixel block, cumulatively adds the absolute value. As a result, the denominator term Σ [Δ _x ] of the estimated calculation formula (7) is obtained.

This denominator term Σ [Δ _x ] is used in the division circuit 130.
Is supplied to. The division circuit 130 is a cumulative addition circuit 125.
With the denominator term Σ [Δ _x ] supplied from
The numerator term ΣD · SGN (Δ _x ) supplied from 3 is divided. As a result, the estimated value X1 of the motion amount V _x by the gradient method is obtained.
Is obtained. This estimated value X1 is supplied to the selection circuit 220.

The above is the motion amount estimating operation of the motion amount estimating unit 100. Next, the estimated value output operation of the estimated value output unit 200 will be described.

The denominator term Σ [Δ _x ] obtained by the cumulative addition circuit 125 and the predetermined value α supplied to the input terminal 340.
Are supplied to the determination circuit 210. The determination circuit 210 is
Both input signals are compared to determine whether the denominator term Σ [Δ _x ] is larger than a predetermined value α.

The result of this determination is supplied to the selection circuit 220. In the selection circuit 220, the denominator term Σ [Δ _x ] has a predetermined value α.
If it is larger, the estimated value X1 of the motion amount estimation unit 100 is selected as the estimated value X1 of the motion amount V _x , and the output terminal 350 is selected.
Supply to. On the other hand, the denominator Σ [Δ _x ] is the predetermined value α
If it is smaller, 0 is selected as the estimated value X2 of the motion amount V _x , and is supplied to the output terminal 350.

As a result, when the denominator Σ [Δ _x ] is small, it is possible to prevent an inappropriate estimated value X2 of the motion amount V _{x from} being obtained. As a result, it is possible to prevent the image from being distorted when the molecular term ΣΔ _x is small.

Although a detailed description is omitted, the part for estimating the vertical motion amount V _y of the moving image has the same structure as the part for estimating the horizontal motion amount V _x .

Further, this configuration is not limited to the gradient method for estimating the motion amounts V _x and V _y by one estimation calculation, but also the gradient method for estimating the movement amounts V _x and V _y by several calculations. That is, it can be applied to the iterative gradient method (including the one using the initial deviation vector).

According to this embodiment described in detail above, the following effects can be obtained. (1) First, according to this embodiment, the image gradients Δ _x , Δ
_y , that is, the denominator Σ of the estimated arithmetic expressions (7) and (8)
If [Δ _x ] and Σ [Δ _y ] are larger than the predetermined value α, the estimated value X1 of the motion amount estimation unit 100 is set to the estimated value X2 of the motion amount V _x.
Output as if small, since the output a 0, image gradient delta _x, if delta _y is small, the motion amount V _x, as an estimate X2 of V _y, inappropriate value obtained It can be prevented. Thus, according to this embodiment, when the image gradients Δ _x and Δ _y are small,
It is possible to prevent an unnatural image from being obtained.

Further (2), according to this embodiment, the image gradient delta _x, if delta _y is small, the motion amount V _x, as an estimate X2 of V _y, since so as to output 0, the As the estimated value X2, positive movement amounts V _x and V _y and negative movement amount V
_x, it is possible to output an intermediate value of V _y. As a result, the average error between the estimated value X2 and the actual motion amounts V _x and V _y can be minimized.

Although one embodiment of the present invention has been described in detail above, the present invention is not limited to the above embodiment.

(1) First, in the above embodiment, the estimated values X of the motion amounts V _x and V _y when the image gradients Δ _x and Δ _y are small.
The case where 0 is output as 2 has been described. However, the present invention may output a value other than 0.

(2) Further, in the above-mentioned embodiment, the case where the present invention is applied to the motion vector detecting device using the equations (7) and (8) as the estimation arithmetic expression has been described. However, theoretically, the estimation of the motion amounts V _x and V _y is performed by the equation (2),
It is also possible to use (3). Therefore, the present invention can also be applied to the motion vector detection device that uses the equations (2) and (3) as the estimation calculation formula.

(3) In addition to this, it is needless to say that the present invention can be variously modified without departing from the scope of the invention.

[0044]

As described above in detail, according to the present invention, when the image gradient is larger than a predetermined value, the result of the estimation calculation is output as an estimated value of the motion amount, and when it is smaller, a predetermined value is set. Therefore, even if the image gradient is small, it is possible to prevent the estimation value of the inappropriate motion amount from being obtained. As a result, according to the present invention, it is possible to prevent an unnatural image from being obtained when the image gradient is small.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention.

[Explanation of symbols]

 100 ... Motion amount estimation unit 110 ... Denominator term operation unit 120 ... Numerator term operation unit 130 ... Division circuit 111 ... Subtraction circuit 112 ... Sign multiplication circuit 113 ... Cumulative addition circuit 121, 122 ... Latch circuit 123 ... Subtraction circuit 124 ... Image gradient Operation circuit 125 ... Cumulative addition circuit 200 ... Estimated value output unit 210 ... Judgment circuit 220 ... Selection circuit

Claims (2)

[Claims] 1. An image is divided into blocks each having a predetermined number of pixels and a predetermined number of lines, and for each block, based on a spatial gradient of an image signal and a difference value in the time axis direction,
A motion amount estimating means for estimating a motion amount of a moving image; and if the gradient is larger than a predetermined value, the estimated value of the motion amount estimating means is output as the motion amount estimated value. A motion vector detection device, comprising: an estimated value output means for outputting a determined value. 2. The motion amount estimating means is configured to estimate the motion amount of the moving image based on the following equation: The estimated value output means outputs the estimated value of the motion amount estimating means as the estimated value of the motion amount when the denominator term on the left side of the above equation is larger than a predetermined value, and outputs 0 when it is smaller. It is constituted so that it may be constituted.
The motion vector detection device described.