JPS61107886A - Smoothins circuit of motion vector - Google Patents

Abstract

PURPOSE:To decrease discontinuity in the time base direction of a motion vector and to improve a visual characteristic when the motion is corrected by using a strong correlation in the time base direction of the motion vector and executing the smoothing processing. CONSTITUTION:A digital TV signal from the terminal 21 is supplied to a representative point extracting circuit 22 and a block data extracting circuit 24. An absolute value of the frame difference data from a converting circuit 26 is supplied to an integrating circuit 27, the integrating value of the frame difference data is supplied to a switch circuit 28 and distributed to graph memories 29-32 for an area each by the circuit 28. Frame difference integrating data stored in the memories 29-32 respectively are supplied respectively to minimum value detecting circuits 33-36. The motion vector from the circuits 33-36 respectively are supplied to a selecting circuit 37. The motion vector for one field unit from the circuit 37 is supplied to a smoothing circuit 37, the noise is removed, the discontinuity in the time base direction is decreased and removed to an output terminal 39.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention provides a motion vector smoothing circuit that smoothes a motion vector indicating the movement of a television camera when the entire shooting screen moves due to panning or the like of the television camera. Regarding.

[Conventional technology]

As one of the transmission methods that narrows the band of high-definition television signals, when the entire screen moves in the same direction, such as when panning a television camera, this image is treated as a still image, and 4 frames are combined. The amount of data is reduced to 1/4 by shifting the sub-sampling position during the field period, and a motion vector indicating the movement of the TV camera is transmitted, and the receiving side uses 4 fields of data to create one image. In addition to reproducing the still image, there is a method that performs motion correction by shifting the coordinate axes according to the motion vector and reading out the still image from memory.

[Problem that the invention seeks to solve]

However, when an incorrect motion vector is determined due to noise or the like, if the above-described motion correction is performed using this incorrect motion vector, discontinuity in the time axis direction becomes noticeable.

Therefore, an object of the present invention is to perform smoothing processing using the strong correlation of motion vectors in the time axis direction.
It is an object of the present invention to provide a motion vector smoothing circuit that can reduce undesired motion vectors in the time axis direction and improve visual characteristics when motion compensation is performed.

Another object of the present invention is to provide a motion vector smoothing circuit that can quickly follow changes in motion vectors and can obtain a sufficient noise removal effect in a steady state.

[Means for solving problems]

The present invention includes a weight generator 14 that generates a weight coefficient that changes sequentially to become smaller than a reset state and is held at a predetermined value, and a motion vector obtained for each field that is inputted and outputted from the weight generator 14. According to the weighting coefficient of Yk=WkXX, + (1-W) Yk-1 (where X
3 is the kth input sample, Yk is the kth output sample, and Wk is the second weighting coefficient. ) Means for performing a weighted average expressed as 4. 5. 6. 8. This is a motion vector smoothing circuit characterized by comprising:

(Operation) Discontinuities between detected motion vectors are eliminated by performing smoothing processing using means 4.5 and 6.8 using strong correlation in the time axis direction. By sequentially decreasing the weighting coefficients from the 7) state and holding them at predetermined values, it is possible to closely follow changes in the motion vector and obtain a sufficient noise removal effect in a steady state.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings. In FIG. 1, 1 indicates an input child to which data of a motion vector determined for each field is supplied.

This motion vector is generated by the configuration shown in FIG. 2, for example. In FIG. 2, reference numeral 21 indicates an input terminal for a digital television signal photographed by a television camera and digitized. This digital television signal is transmitted to the representative point extraction circuit 22 and the block data extraction circuit 2.
The pixel data of the representative point obtained from the output of the representative point extracting circuit 22 is supplied to the representative point memory 23.

The pixel data of the representative point stored in the representative point memory 23 is that of the previous frame.

The block data extraction circuit 24 extracts pixel data for each block included in the current frame and supplies it to the subtraction circuit 25 . This subtraction circuit 25 calculates the difference between each pixel in the block and the representative point of the previous frame, that is, frame difference data. This frame difference data is transferred to the conversion circuit 26.
and converted to absolute value.

The absolute value of the frame difference data from the conversion circuit 26 is supplied to the integration circuit 27, and the integrated value of the frame difference data is supplied to the switch circuit 28. As shown in FIG. 3A, the frame difference data is obtained for each of the areas 41A, 41B, 41C, and 41D that are formed by dividing the screen 40 of one field into four. The integrated value of the frame difference data from the integrating circuit 27 is stored in the graph memory 29.30 for each area by the switch circuit 28.
It is divided into ゜31.32. The graph memory 29 is
This memory has a capacity of 1 block and stores the integrated value of frame difference data (frame difference integral data) of each block included in the area 41A. Graph memory 30, 31.3
2 similarly stores frame difference integral data of areas 41B, 41C, and 41D.

The frame difference integral data stored in each of these graph memories 29 to 32 is sent to minimum value detection circuits 33 and 34.
．． 35 and 36 respectively.

The minimum value detection circuits 33 to 36 detect the position of the minimum value in the frame difference integral data of the corresponding graph memories 29 to 32, ie, the motion vector of each of the regions 41A to 41D.

Motion vectors from each of these minimum value detection circuits 33 to 36 are supplied to a selection circuit 37. The selection circuit 37 is
FIG. 3B obtained for four regions 41A to 41D
A motion vector for one field is generated from motion vectors as shown in FIG. This selection circuit 37
has, for example, a majority logic circuit configuration or a configuration that calculates a multiple average.

The motion vector in units of one field from the selection circuit 37 is supplied to the smoothing circuit 38, noise is removed, and the motion vector is sent to the output terminal 3.
It is taken out at 9. As shown in Figure 3B, the case where the four motion vectors are the same is an ideal case; in reality, noise and image patterns (differences between low-frequency and high-frequency patterns ) or due to the influence of moving objects, the selection circuit 3
The field-by-field motion vectors from 7 may include motion vectors with low accuracy. A smoothing circuit 38 is provided to reduce the discontinuity in the time axis direction that occurs due to this.

FIG. 1 shows the configuration of an embodiment of the present invention applied as the smoothing circuit 38 described above. Since the motion vector is position data, it has an X component and a Y component on coordinates, and the configuration shown in FIG. 1 is applied to each component.

In FIG. 1, the motion vector in field units from the input terminal indicated by 1 is subtracted by subtracting circuit 4, subtracting circuit 7 and AND.
The output of the subtraction circuit 4 is supplied to the gate 11 and the output of the subtraction circuit 4 is supplied to the multiplication circuit 5.
supplied to This multiplication circuit 5 is supplied with weighting coefficients from a weight generation section 14 . The output of the multiplier circuit 5 is supplied to an adder circuit 6, and the output of the adder circuit 6 is supplied to a subtracter circuit 7. A register 8 and an AND gate 2 which cause a delay of one field.

The subtraction circuit 4, the multiplication circuit 5, the addition circuit 6, and the register 8 constitute a smoothing filter.

That is, if the second value of the input motion vector is X, and the weighting coefficient from the weight generator 14 at that time is W, then the output data YK output from the adder circuit 6 is expressed by the following equation. Become.

Yk=(Xk-Ytsu-+)Wk+Y*-r=WkxX
k1(1-Wk)Yk-8 As is clear from the above equation, the motion vector Yk processed by the smoothing filter is the motion vector Y of the previous field multiplied by the coefficient of (1-W). , the sum (ie, weighted average value) of the motion vector X of the current field multiplied by the weighting coefficient Wk.

Although not shown, the weight generation unit 14 includes a counter and a ROM that stores a plurality of weighting coefficients.
A field clock from terminal 2 is supplied. This field clock causes the value of the weighting coefficient to change using the field as the minimum cycle. At the same time, the weight generating section 14 is supplied with a scene change signal from the terminal 3 as its reset signal. This scene change signal is used to reset the weight generator 14 and return the weight coefficients to their initial values when the television camera changes scenes.

If the count value of the counter of the weight generator 14 is set to O at the time of reset, the value of the weighting coefficient is gradually decreased from 1 as shown in the table below every time the field clock advances, and is held at the value of 1/8. be done.

The reason why the weighting coefficient is changed in this way is that when a scene change occurs, the weighting coefficient is set to 1 to quickly follow the change in the motion vector, and after a steady state is reached, a small weighting coefficient of 1/8 is used. This is to obtain an effective noise removal effect. Further, by setting the weighting coefficient to a power of 2, the multiplication circuit 5 can be configured as a shift circuit.

The weighting coefficient may be changed by changes other than those in the table above, for example (1, 1/2.
1/3.1/4.115.1/6.1/7°1/8).

A motion vector from an input terminal 1 is supplied to an AND gate 11, and a motion vector passed through a smoothing filter is supplied to an AND gate 12.

The outputs of these AND gates 11 and 12 are OR gate 1.
3 to the output terminal 15.

The AND gates 11 and 12 constitute a selector that is switched by the output of the comparison circuit 10.

The comparison circuit 10 receives the absolute value of the difference between the motion vector from the input terminal 1 obtained by the subtraction circuit 7 and the conversion circuit 9 and the motion vector output from the smoothing filter, and the threshold data from the terminal 16. is supplied.

When the absolute value of this difference is smaller than the threshold data, it is determined that the influence of noise is small, and the comparison circuit 10 outputs an AND
One comparison output supplied to the gate 11 is at a high level, and the other comparison output supplied from the comparison circuit 10 to an AND gate 12 is at a low level. Therefore, the motion vector from input terminal 1 is selected by AND gate 11,
It is taken out to the output terminal 15 via the OR gate 13.

On the other hand, when the absolute value of the difference between the motion vector from the input terminal 1 and the motion vector output from the smoothing filter is larger than the threshold data, it is determined that the influence of noise is large, and the comparison circuit 10 outputs the AND gate 11. When one comparison output supplied to the AND gate 12 is at a low level, the other comparison output supplied from the comparator circuit 10 to an AND gate 12 is at a high level. Therefore, the motion vector from the smoothing filter is selected by AND gate 12 and taken out via OR gate 13 to output terminal 15.

Note that such selection may not be made, and motion vectors that have been passed through a smoothing filter may always be output.

[Effects of the Invention] According to this invention, since the motion vector obtained for each field is smoothed in the time axis direction,
Discontinuities caused by errors in motion detection due to noise or the like can be removed, and image quality can be improved when motion correction is performed.

Furthermore, by making the weighting coefficients variable, the present invention has good followability to changes in motion vectors, and can also increase the noise removal effect in a steady state.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention, FIG. 2 is a block diagram of an example of a motion vector detection circuit, and FIG. 3 is a route diagram used to explain motion vector detection. 1: Input terminal, 2: Field clock input terminal, 3
: input terminal of shift chain 2 signal, 4°7.25: subtraction circuit, 15: output terminal, 38: smoothing circuit.

Claims (1)

[Scope of Claims] A weight generation unit that generates a weighting coefficient that changes sequentially to become smaller than a reset state and is held at a predetermined value;
The motion vector determined for each field is input, and according to the weighting coefficient from the weight generation section, Y_k=W_k×X_k+(1-W_k)Y_k_-_
1 (where X_k is the k-th input sample, Y_k is the k-th output sample, and W_k is the k-th weighting coefficient.) Vector smoothing circuit.