JPH0728408B2 - Image motion vector detection device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a motion vector detecting device for a television image.

2. Description of the Related Art A conventional image motion vector detecting device is disclosed in, for example, Japanese Patent Laid-Open No. 61-201581. FIG. 8 is a block diagram of this conventional image motion vector detecting device.

In the figure, a representative point memory 2 stores the signals of pixels at some representative point positions among the image signals of one frame or one field input from the input terminal 1 and the other one from the input terminal 1 When a frame or one field image signal is input and supplied to the correlator 3, the stored signal of the representative point is supplied to the correlator 3. The correlator 3 uses the absolute value of the difference between the signal of the representative point and the signal of the pixel shifted (i, j) in the horizontal and vertical directions from the position of the representative point of the image signal | ΔL | (i, j
j) is taken and supplied to the cumulative addition circuit 4. The cumulative addition circuit 4 cumulatively adds this | ΔL | (i, j) for each shift amount (i, j) to all the representative points in the detection area, and calculates the shift amount (i, j) as Correlation value Σ | ΔL | (i, j) is obtained, and this is supplied to the motion vector generation circuit 5. The motion vector generation circuit 5 detects the shift amount (i, j) at which this correlation value Σ | ΔL | (i, j) is minimized, generates a motion vector, and supplies it to the output terminal 6.

However, in the above configuration, when the S / N ratio of the input image signal is poor, the probability of detecting an erroneous vector is high.

Further, in order to reduce the probability of detecting the erroneous vector, the number of representative points in the detection area and the number of cumulative additions must be set to a certain value or more. Further, even when the S / N ratio of the image signal is good, when the waveform of the image signal changes stepwise as shown in the screen (a) shown in FIG. 9 and the level wave (b) on one line thereof, Since it is unlikely that a change component of the image signal exists on the representative points 7 and 7 ', as shown in the correlation value (c) as shown in FIG. In the equal range 9, since the correlation value Σ | ΔL | (j) takes a constant minimum value, the accuracy of vector detection becomes poor, and furthermore, the motion vector cannot be detected. In order to prevent this, the interval between the representative points must be below a certain value, and therefore the number of representative points in the detection area and the number of cumulative additions must be above a certain value. Therefore, in the conventional device, the number of representative points in the detection area and the number of cumulative additions must be set to a certain value or more, and thus there is a certain limit to the simplification of hardware and the reduction of processing time. Further, when the detection range of the motion vector is set to a certain value or more, and the pixel area in the range deviated within a certain value from each representative point is set as the search area corresponding to the representative point, If the search areas 10 and 10 'do not overlap each other on the screen, and the pixels in each search area and the representative points have a one-to-one correspondence, the apparatus of FIG. When one pixel signal of a frame or field that is input later in time is input, the representative point memory 2
Needs to output a signal of one corresponding representative point, and the correlator 3 and the cumulative adder 4 need to perform only one calculation, respectively, so that the hardware can be greatly simplified. The hardware becomes more complicated as the number of times the search areas 10 and 10 'are overlapped increases. However, in order that the search areas do not overlap, the size 11, 11 ′ cannot be set to more than the interval 8, 8 ′ between the representative points as shown in FIG. Will be limited by.

And in the conventional one, as described above, the interval between each representative point
Since 8,8 'must be kept below a certain level, if we try to simplify the hardware, the range of motion vectors that can be detected will become narrow, and if we try to widen this range,
There was a problem that the hardware became complicated. Third, when the signal waveform level fluctuates as shown in FIG. 11 (a) due to flicker in the image signal or the operation of automatic exposure adjustment during shooting, the difference in the signal from each representative point The absolute value | ΔL | (i) of the above becomes as shown by 12,12 ′ in FIG. 7B, and the correlation value Σ | ΔL | (i) obtained by cumulatively adding it
Is as shown in FIG. 7C, and the minimum point 13 is different from the moving amount of the image. Therefore, an erroneous vector is detected. However, in the conventional example as well, the correlator 3 shown in FIG. 8 does not obtain the absolute value of the difference between the two signals, but the one obtained by multiplying the two signals, and the motion vector generation circuit 5 cumulatively adds them. Deviation (i, j) that gives the maximum value of
The problem can be solved by changing to a motion vector.

However, in this case, when the hardware is used, the correlator 3 requires a multiplication circuit. Further, since the number of bits of the digital signal of the correlator 3 becomes large, the amount of calculation in the cumulative addition 4 and the motion vector generation circuit 5 after the correlator 3 becomes large, and the circuit scale also becomes large. Therefore, there are disadvantages that the circuit scale of the entire motion vector detection device is large, and the motion vector detection device becomes complicated and the processing time becomes long.

Also, accurate motion vector detection is difficult when the image signal contains a lot of shading.

Therefore, in the conventional method, the probability of error vector detection is high when the S / N ratio is poor, and in order to reduce this probability, vector detection is performed even when a step signal is input. Therefore, the number of representative points and the number of cumulative additions are set to be very large, which causes a problem that the scale of hardware and the processing time become large. In addition, if the signal waveform level fluctuates due to flicker during shooting or the operation of the automatic exposure adjustment device, false vector detection becomes impossible, and accurate motion vector detection is difficult even when a large amount of shading is included. There was a problem that was.

In view of such a point, the present invention firstly provides a motion vector detection device capable of accurately detecting a motion vector even from an image signal having a poor S / N ratio, and secondly, Motion vectors can be accurately detected even from a bad image signal or an image signal in which the signal waveform changes stepwise, even if the distance between representative points is wide, or the number of representative points and cumulative addition is small. Therefore, it is possible to provide a motion vector detection device capable of widening the range of motion vectors that can be detected without complicating the hardware, and further simplifying the hardware and shortening the processing time. Thirdly, when the level of the signal waveform fluctuates due to flicker or the like, even if a lot of shading is included, a motion vector detecting device capable of accurately detecting the motion vector is provided. An object of the present invention is to provide a road-scale without changing most.

Means for Solving the Problems In order to achieve the above object, the present invention provides a plurality of representative signals which are set with a predetermined horizontal interval LX and vertical interval LY with respect to a signal within a detection region of an input signal of a certain frame. A means for storing a signal of a point and a difference between a stored signal of the plurality of representative points and an input signal around each representative point of the input signals of the frames subsequent to the frame are calculated and totaled to calculate a motion vector detection range. Means for obtaining a correlation value with respect to a shift in the inside, and means for obtaining a minimum point of the correlation value and outputting as a motion vector, at least the predetermined horizontal interval LX, vertical direction with respect to the input image signal. Interval LY
IIR lowpass filter or FIR with transients of
This is a configuration provided with a low-pass filter.

Further, it is configured to have a high-pass filter that suppresses at least a vertical direction DC component or a horizontal direction DC component and a low frequency component of the image signal.

By suppressing the high frequency component of the action image information, the influence of noise can be suppressed, and the motion vector can be detected more accurately. Further, by suppressing the high frequency component of the image, even when a signal whose waveform changes stepwise as shown in FIG. 9 (b) is input, the signal waveform passing through the low pass filter is As shown in Figure (a),
Representative points in which the change component of the image signal is arranged at intervals
It can be extended over a wide range so that it exists on 7, 7 ′, and as a result, the absolute value of the signal difference | ΔL | (i) for each representative point is shown in FIG. Then, the correlation value Σ | ΔL | (i) becomes as shown in FIG. Here, the minimum point 13 of the correlation value is obtained as one point, and the motion vector can be accurately detected even when the interval between the representative points is wide. Furthermore, if the signal waveform level fluctuates due to flicker or the like by suppressing the DC component and low-frequency component of the image, for example, as shown in FIG. The difference between the separated signals is shown in FIG. 5 (a).
The waveform shown in Fig. 3 is obtained, and from this, the correlation value Σ | ΔL |
When (i) is obtained, the correlation value Σ | as shown in FIG.
ΔL | (i) is obtained. Compared with FIG. 9 (c), the motion vector can be detected more accurately than the conventional one.

Practical Example FIG. 1 is a block diagram of an image motion vector detecting apparatus according to a first embodiment of the present invention.

The operation will be described below. First, an image signal of at least two fields is input to the input terminal 1. The input image signal is supplied to the low-pass filter 14, the high-frequency component of either or both of the horizontal direction and the vertical direction is suppressed, and the image signal is supplied to the representative point memory 2 and the correlator 3. The representative point memory 2 stores the signals of the pixels at the positions of some representative points among these signals. When the image signal of another field or frame is supplied to the correlator 3 through the low-pass filter 14, at least one field or one frame of the representative point signal from the representative point memory 2 is supplied to the correlator 3. It The correlator 3 receives the signal of the representative point,
The absolute value | ΔL | of the difference from the signal of the pixel shifted by (i, j) from the position of the representative point of another image signal is taken, and this is supplied to the cumulative addition circuit 4. The cumulative addition circuit 4 cumulatively adds the absolute values | ΔL | for all the representative points in the detection area for each shift amount (i, j) to obtain Σ | ΔL | (i, j). This Σ |
ΔL | (i, j) is used as a correlation value for the shift amount (i, j) and is supplied to the motion vector generation circuit 5. The motion vector generation circuit 5 uses the deviation (i, i, j) that minimizes the correlation value Σ | ΔL | (i, j).
j) is detected, and this is supplied to the output terminal 6 as a motion vector of the image of the detection area.

This effect can suppress the influence of noise by suppressing the high frequency components of the image information, and the motion vector can be detected more accurately. Even if a signal whose waveform changes stepwise as shown in FIG. 9 (b) by suppressing the high frequency component of the image, the signal waveform passed through the low pass filter is shown in FIG. As shown in a), the effect of the step response is longer than the interval between the representative points 7 and 7'because of the low-pass filtering process, the absolute value of the signal difference at each representative point | ΔL | (i) Is the correlation value Σ | ΔL |
(I) becomes as shown in FIG. Here, the minimum point 13 of the correlation value is found to be 1, and even if the interval between the representative points is wide, it is possible to accurately detect the motion vector.

FIG. 6 shows the first effect of the first embodiment of the present invention. The figure shows the presence or absence of a low-pass filter,
6 is a graph showing the relationship between the S / N ratio of an image and the average value of the detected vector error magnitudes. Here, as the above-mentioned low-pass filter, an IIR type low-pass filter represented by the following equation was used. In the formula, z and ω represent delays of one pixel in the horizontal direction and one line in the vertical direction, respectively.

Α = 3/4, β = 1/2; α = 15/16, β = 7/8 Further, the number of representative points and the cumulative number of additions in the detection area at this time are 16 respectively. Conventionally, about 200 times or more is required, but in this embodiment, it is extremely small. As shown in FIG. 6, in comparison with the conventional example which does not use the low-pass filter, the present embodiment has a clearly smaller error in the detection vector, enables the motion vector to be detected more accurately, and has a large effect. Although the IIR type low-pass filter is used here, other low-pass filters may be used.

FIG. 2 is a block diagram of an image motion vector detecting device according to the second embodiment of the present invention. Second
In the figure, 1 is an image signal input terminal, 15 is a high-pass filter, and others are the same as in FIG.

The operation of the image motion vector detection apparatus of this embodiment further configured as described above will be described below. First, the image signal input from the input terminal is supplied to the high pass filter 15. The high-pass filter 15 suppresses the DC component and low frequency component of the image signal and supplies the output to the representative point memory 2 and the correlator 3. The following operation is the same as that of the first embodiment. The result will be described later.

FIG. 3 is a block diagram of an image motion vector detecting device according to a third embodiment, which is a combination of the first and second embodiments of the present invention. In FIG. 3, reference numeral 16 denotes a bandpass filter. The bandpass filter 16 suppresses high-frequency components of either or both of the horizontal direction and the vertical direction of the image, and at the same time suppresses DC components and low-frequency components.

The operation is the same as in the first and second embodiments.

The effects of the second and third embodiments are that when the DC component and the low frequency component of the image are suppressed and the signal waveform level changes due to flicker, for example, as shown in FIG. In this case, when this is passed through a high-pass filter, for example, when the difference between the signals separated by a fixed pixel in the horizontal direction is taken, a waveform as shown in FIG. 5 (a) is obtained. Absolute value for each representative point | ΔL |
When (i) is obtained, the correlation value Σ | ΔL | (i) obtained by cumulative addition is obtained as shown in FIG. Compared with FIG. 9 (c), a motion vector that is clearly more accurate than the conventional one can be detected.

An example of the effects of the second and third embodiments will be described with reference to FIG. FIG. 7 is a graph of the flicker of the image signal and the average value of the error magnitude of the vector when the motion vector is detected from this signal. The respective results are also shown for the case of an image having a relatively high contrast and the case of an image having a low contrast. In the figure, what is shown without HPF is (α, β) = (15
/ 16, 7/18) of a motion vector detection device equipped with a low-pass filter. Also, the one with HPF is provided with a high-pass filter H '(z) = 1-z- ¹⁶ as shown in the following formula in combination with the low-pass filter, and therefore the band-pass filter is composed of a low-pass filter and a high-pass filter. The motion vector detecting apparatus according to the third embodiment.

Comparing the two, a motion vector detection device without a high-pass filter cannot accurately or even detect a motion vector even from an image signal with a flicker of 1 to 4%. The vector detection device can detect a motion vector accurately even from an image signal with a flicker of 16 to 50%, and its effect is very large. Further, here, as the high-pass filter, one that takes the difference between the signals at 16 pixels apart in the horizontal direction as shown in the above expression is used, but other high-pass filters may be used.

Further, regarding shading, since it is considered that this is a fixed pattern of low frequency, its influence can be suppressed by a high-pass filter.

Here, an example in which one motion vector is detected from the correlation value between two frames or fields has been shown, but the screen is divided into several detection areas, the motion vector is calculated for each detection area, and these are judged and averaged. The present invention can also be applied to a motion vector detection device that sets a motion vector for the entire screen by a plurality of motion vectors and a motion vector detection device that reduces the vector detection error by weighted averaging a plurality of motion vectors obtained over several fields. Can be applied and its effect is great.

Further, the two frames or fields mentioned here do not necessarily have to be temporally continuous, and one or several frames or one or several fields may be provided between them.

EFFECTS OF THE INVENTION As described above, according to the present invention, first, a motion vector can be accurately detected even from an image signal containing noise. Secondly, a motion vector can be accurately detected even from an image signal containing noise or an image signal whose waveform changes in steps, even with a very small number of representative points and cumulative addition. Therefore, for example, if the present invention enables 256 representative points to be used for motion vector detection with 16 representative points, the capacity of the representative point memory and the number of calculations can be reduced to 1/16.
It's done. Also, since the representative point interval can be made significantly wider than the conventional one, it is possible to make the size of the search area large without overlapping the search areas with each other or by reducing the number of times of overlapping. Therefore, a wider range of motion vectors can be detected without making the hardware more complicated than the conventional one. Third, even if the signal waveform level fluctuates due to flicker,
Further, even when a lot of shading is included, it is possible to accurately detect the motion vector.

Therefore, the performance of the image motion vector detecting device can be greatly improved, and at the same time, the cost thereof can be made very low, and the practical effect of the present invention is great.

[Brief description of drawings]

1, FIG. 2 and FIG. 3 are block diagrams of the motion vector detecting device for image according to another embodiment of the present invention, and FIGS. 4 and 5 are operation waveform diagrams of the device, and FIG. And FIG. 7 is a relational diagram comparing the same embodiment and the conventional example,
FIG. 8 is a block diagram of an image motion vector detecting device of a conventional example, FIG. 9 is a schematic diagram and an operation waveform diagram for explaining the operation of the conventional example, and FIG. 10 is a diagram showing each search area and representative points. FIG. 11 is a diagram schematically showing the relationship, and FIG. 11 is an operation waveform diagram of a conventional example. 2 ... Representative point memory, 3 ... Correlator, 4 ... Cumulative adder, 5 ... Motion vector generation circuit, 14 ... Low pass filter, 15 ... High pass filter, 16 ... Hand pass filter.

Claims (2)

[Claims] 1. A predetermined horizontal interval LX and a vertical interval LY for signals in a detection area of an input signal of a frame.
Means for storing signals of a plurality of representative points set by opening the input signal of the frame after the frame,
A means for obtaining the difference between the stored signal of the plurality of representative points and the surrounding input signal including each representative point, calculating the correlation value for the deviation within the motion vector detection range, and the minimum point of the correlation value An IIR low-pass filter or a FIR low-pass filter having a transient that is at least the predetermined horizontal interval LX and vertical interval LY with respect to the input image signal is provided. Characterized image motion vector detection device. 2. A motion vector detecting apparatus for an image according to claim 1, further comprising a high-pass filter for suppressing a DC component and a low frequency component of at least a vertical direction or a horizontal direction of the image signal.