JPH08107546A - Picture filter and motion vector detector - Google Patents

Abstract

PURPOSE: To set a desired output time base centroid by changing the weighting coefficient of a time base filter corresponding to the judgement signals of moving picture and still picture areas for inputted digital picture signals. CONSTITUTION: A coefficient change means sets the weighting coefficient for the moving picture area in the time base filter 24 when the inputted digital picture signals are the moving picture area and sets the weighting coefficient for the still picture area in the time base filter 24 when the inputted digital picture signals are the still picture area. The time base filter 24 weights and acids plural pixel data in front and behind on the time base of the inputted digital picture signals by the weighting coefficient set corresponding to the moving picture and still picture areas. Then, the digital picture signals for which an output centroid on the time base is at a desired position regardless of the still picture and moving picture areas are outputted from the time base filter 24, filtered in a horizontal direction and in a vertical direction further by a twodimensional low-pass filter 23 and outputted.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image filter and a motion vector detection device, and can be applied to, for example, a television signal format conversion device.

[0002]

2. Description of the Related Art Digital image signals (luminance signals,
Some devices that handle color signals or composite video signals include a motion vector detection device and utilize the detected motion vector for processing.

[0003] For example, in a television signal format conversion apparatus, due to the difference in the number of lines of the television signal before and after the conversion and the number of frames per frame, interpolation and extrapolation using digital image signals having different time axes are performed. Interpolation processing such as insertion is performed, and the interpolation ratio or the like at this time is selected according to the motion vector. Further, in an image coding apparatus that employs predictive coding, one of them is moved according to a motion vector so that the prediction difference between the immediately preceding frame (or immediately preceding field) and the current frame (or current field) becomes small. I am trying to get the prediction difference later.

For the motion vector, for example, digital images that are different in time by an image unit time (field or frame) on the time axis are collated for each region of N × N pixels and one of the digital images is moved, It is often obtained as the movement amount of one digital image whose matching distance is the shortest. In this case, if the change of the pixel data value in the area is large, the matching distance value becomes large even in the case of the optimum matching, and the motion vector detection accuracy is likely to decrease accordingly. Therefore, in the motion vector detection device, before the digital image signal is input to the motion vector detection circuit, the pre-filter smoothes the change of the digital image signal (drops the frequency band; blurs in a popular expression). ) Is being done.

As an image filter used as a pre-filter for a motion vector detecting device, there has conventionally been the one shown in FIG.

In FIG. 2, the first and second field memories 10 and 11 delay the input digital image signal on a field-by-field basis.
Pixel data at almost the same position in the current field, the previous field, and the three fields preceding and following the preceding field can be obtained. Each of the line memories 12, 13 and 14 outputs the pixel data of one line before with respect to the pixel data of the current field, the previous field or the previous field. The six pixel data thus obtained are multiplied by the corresponding coefficient multiplication units 15, ..., 20 with the fixed coefficients output from the coefficient output unit 21, and then the addition unit 2
Given to 2. The adder 22 adds these weighted six pixel data and outputs the result to the two-dimensional low-pass filter 23.

Here, the coefficient from the coefficient output unit 21 is used so that pixel data for smoothing a digital image signal can be obtained by using different pixel data on the time axis, and the afterimage characteristic of the image pickup camera is taken into consideration. It has been selected.

Incidentally, the first and second field memories 1
0 and 11, line memories 12 to 14, coefficient multiplication unit 15
˜20, the coefficient output unit 21 and the addition unit 22 constitute a time axis filter.

The two-dimensional low-pass filter 23 is a horizontal / vertical low-pass filter and extracts the low-frequency component of the digital image signal from the adder 22 and outputs it to a motion vector detection circuit (not shown).

[0010]

The time axis of the digital image signal output from the time axis filter is equal to the time axis of the digital image signal input to the time axis filter, or differs by exactly one field. Is required. However, in the above-described prefilter (image filter), a digital image signal whose center of gravity on the time axis is different from the desired position in the still image area of the input digital image signal is output from the time axis filter. There are challenges.

Due to the afterimage characteristic of the image pickup camera, the influence of the pixel data at the same position in the field two years before is included in the pixel data in the current field. If the pixel position is a pixel position in the moving image area, the pixel data in the current field is affected by the pixel data of the same position in the previous field but different contents, while the pixel position is in the still image area. If the pixel position is inside, the influence of the pixel data having the same content at the same position in the field before the previous field will be included in the pixel data in the current field.

Therefore, when the input digital image signal is in the moving image area, the coefficient for weighted addition is set so that the time-axis center-of-gravity position of the output digital image signal from the time-axis filter is a desired position (for example, the current field of the input digital image signal). When the weighting coefficient in the time axis filter is determined so that the still image area of the digital image signal (white circle) is input, as shown in FIG. Is an even field or an odd field, the time-axis center-of-gravity position of the output digital image signal (black circle) from the time-axis filter is deviated from the desired position.

As a result, the digital image signal output from the image filter is different from the original time axis. Since such a digital image signal is input, the motion vector detection circuit also outputs a motion vector having a magnitude even if the input digital image signal to the prefilter is a still image area.

[0014]

According to a first aspect of the present invention, there is provided a time axis filter for weighting and adding a plurality of pixel data which are successively arranged on the time axis of an input digital image signal, and 2.
An image filter that has a two-dimensional low-pass filter and that smoothes changes in the input digital image signal.The time axis filter weighting coefficient is changed according to the determination signal of the moving image / still image area for the input digital image signal. Is provided with a coefficient changing means.

The second aspect of the present invention is a motion vector detecting device comprising a motion vector detecting circuit and a pre-filter provided in the preceding stage thereof for smoothing the change of the input digital image signal. First as a filter
The image filter according to the present invention is applied.

[0016]

In the image filter according to the first aspect of the present invention, the coefficient changing means causes the time axis filter to set the weighting coefficient for the moving image area when the input digital image signal is the moving image area, and the input digital image signal. If is a still image area, a weighting coefficient for the still image area is set in the time axis filter. The time axis filter is used to convert a plurality of pixel data that are consecutive in time on the time axis of the input digital image signal
Weighting addition is performed using a weighting coefficient set according to the still image area. In this way, the time axis filter outputs a digital image signal at a desired position whose output center of gravity on the time axis is in a still image / moving image area. The digital image signal is further filtered in the horizontal and vertical directions by the two-dimensional low-pass filter and output.

In the motion vector detecting device of the second aspect of the present invention, the image filter of the first aspect of the present invention is applied as a pre-filter, and the digital image in which the output center of gravity position on the time axis is at the desired position. Since the signal is input to the motion vector detection circuit, the motion vector detection circuit outputs a highly accurate detection result.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a motion vector detecting device according to the present invention will be described below in detail with reference to the drawings. Here, FIG.
Shows the entire configuration of this embodiment, and the same portions as those in FIG. 2 described above are denoted by the same reference numerals.

In FIG. 1, the motion vector detecting device of this embodiment basically has a motion vector detecting circuit 30.
And the time axis filter 24 and the two-dimensional low-pass filter 2
3 and a prefilter (image filter).

As will be described later, the prefilter is the same as the conventional one except that the weighting coefficient of the time axis filter is switched based on the determination result of the moving image / still image region with respect to the input digital image signal. It is a thing.

That is, the time axis filter 24 of the prefilter is composed of the first and second field memories 10 and 11.
It is composed of three line memories 12 to 14, six coefficient multiplication units 15 to 20, an addition unit 22 and a coefficient output unit 21A.

Here again, the first and second field memories 10 and 11 delay the input digital image signal on a field-by-field basis, so that the pixel data A and E of the current field, the previous field and the field before the previous field are delayed. , I
The line memories 12, 13, 14 respectively take out pixel data C, G, K delayed by one line with respect to the pixel data A, E, I of the current field, the previous field or the field before the previous field, and a total of 6 Pixel data A of
Coefficient multiplication unit 1 corresponding to C, E, G, I, and K, respectively
5, ..., 20 are given. Each coefficient multiplication unit 15, ..., 2
0 to the coefficient output unit 2 via a coefficient switching unit 31 described later.
1A outputs a set of weighting factors M0, N0, O
0, P0, Q0 and R0 corresponding elements or another set of corresponding weighting factors M1, N1, O1, P1, Q1 and R1 corresponding elements are given, and the pixel data and the weighting coefficient are multiplied, Pixel data B, D, F after weighting processing
It is given to the H, J, or L addition unit 22. The adder 22 adds the pixel data B, D, F, H, J, and L and outputs the result as a digital image signal filtered in the time axis direction to the two-dimensional low-pass filter 23.

The two-dimensional low-pass filter unit 23 is a horizontal and vertical low-pass filter, and the addition unit 22
The low frequency component of the digital image signal is extracted and the extracted digital image signal T is supplied to the motion vector detection circuit 30 as the output of the prefilter.

The motion vector detection circuit 30 detects a motion vector based on a digital image signal whose change is smoothed (frequency band is dropped) by such a prefilter. Any detection method may be used.

In the case of this embodiment, as described above, the weighting coefficient of the time axis filter 24 is switched based on the determination result of the moving image / still image area in the input digital image signal, which is different from the conventional case. And is a feature of the embodiment.

The digital image signal inputted to the time axis filter 24 is also given to the moving picture / still picture judging section 32.
Since the moving image / still image determination unit 32 determines in advance which weighting coefficient set to use and outputs the coefficient switching control signal S, the moving image / still image determination unit 32 outputs the coefficient switching control signal S from the digital image signal input to the time axis filter 24. The one before the time required for the processing is given to the moving image / still image determination unit 32. The moving image / still image determination unit 32 is mainly composed of, for example, a motion detection circuit that detects a motion from digital image signals of two consecutive frames. The moving image / still image determination unit 32
For example, N × at the same position of two frames of digital image signal
The sum of the absolute difference values (or the squared difference) of the pixel data of M pixels is compared with a threshold value, and when it is large, it is determined as a moving image area, when it is small, it is determined as a still image area, and is synchronized with the processing of the time axis filter The coefficient switching control signal S is given to the coefficient switching unit 31 at the timing.

The coefficient switching unit 31 is composed of six 2-input 1-output switches. Both input terminals a and b of each switch are connected to the coefficient output unit 21A, and the output terminals correspond to each other. , 20 are connected to the coefficient multiplying units 15. The coefficient switching unit 31 selects the input terminal a or b of the internal switch according to the coefficient switching control signal S, that is, the determination result signal S of the moving image area / still image area of the input digital image signal. The coefficient output unit 21A gives a set of weighting coefficients M0, N0, O0, P0, Q0 and R0 for the moving image area to the input terminal a of each internal switch of the coefficient switching unit 31, and also to the input terminal b. Gives a set of weighting factors M1, N1, O1, P1, Q1 and R1 for the still image area.

Therefore, when the coefficient switching control signal S indicates a moving image area, each coefficient multiplying unit 15, ...
, The weighting factors M0, N0, O0, P for the moving image area
0, Q0, R0 are given and weighted, while if the coefficient switching control signal S indicates a still image area,
, 20 are given weighting factors M1, N1, O1, P1, Q1, and R1 for the still image area.

Here, a set of weighting factors M0, N0, O0, P0, Q0 and R0 for the moving image area is the center of gravity of the output even if the moving image area of the input digital image signal is filtered on the time axis. The position is preselected so that the position becomes a desired position (for example, the position of the current field), and a set of weighting factors M1, N1, O for the still image area is set.
1, P1, Q1 and R1 are preselected so that the center of gravity of the output will be at a desired position (for example, the position of the current field) even if the still image area of the input digital image signal is filtered on the time axis. It was done.

The coefficient switching unit 31 responds to a determination signal (coefficient switching control signal) S of a moving image / still image determination unit 32 which determines a moving image / still image area using digital image signals separated by at least one frame. To select the weighting factor,
Since the position of the center of gravity of the time axis filter is switched, it is possible to reduce the occurrence of a motion vector error when detecting a motion vector between signals separated by at least one field.

As described above, the set of pixel data from the coefficient multiplying units 15 to 20 weighted and multiplied by the weighting factor appropriately selected according to the area is added by the adding unit 22 as described above, and is added to the time axis. The output data of the filter is input to the two-dimensional low-pass filter unit 23.

Therefore, according to the above-described embodiment, the time axis filter in the pre-filter is adapted to switch and use the weighting coefficient based on the judgment of the still picture area and the moving picture area. Can output a digital image signal at a desired position regardless of whether the output center of gravity on the time axis is a still image / moving image area. Since the motion vector detection circuit 30 obtains a motion vector for a digital image signal at a desired position on such a time axis, it is possible to improve the detection accuracy of the obtained motion vector as compared with the conventional case.

In the above embodiment, the weighting process is switched at the stage of giving to the coefficient multiplying units 15 to 20, but the coefficient multiplying unit also separately prepares for the still image area and the moving image area and performs the multiplication. The switching may be performed at the stage after the processing, or even the adding unit may be separately prepared for the still image area and the moving image area and switched at the stage after the addition processing.

In the above embodiment, the image filter, which is a feature of the present invention, is applied to the pre-filter of the motion vector detecting apparatus. However, such an image filter smooths the change of the digital image signal. The present invention can also be applied to other devices for which it is required as a part of the processing to (reduce the frequency band). For example, some devices that process digital image signals with a coarse pixel pitch include
Although there is a device that applies an image filter that drops the frequency band so that the roughness is not noticeable, the above-described image filter can be applied to such a device.

[0035]

As described above, according to the first aspect of the present invention,
In an image filter for smoothing the change of an input digital image signal, a coefficient changing unit for changing a weighting coefficient of a time axis filter is provided according to a determination signal of a moving image / still image region for the input digital image signal. Therefore, regardless of whether the input digital image signal is in the moving image area or the still image area, the output time axis center of gravity can be set to a desired value.

Further, according to the second aspect of the present invention, since the image filter according to the first aspect of the present invention is applied as the pre-filter of the motion vector detecting device, the accuracy of detecting the motion vector can be improved. it can.

[Brief description of drawings]

FIG. 1 is a block diagram showing a motion vector detection device of an embodiment.

FIG. 2 is a block diagram showing a prefilter of a conventional motion vector detection device.

FIG. 3 is an explanatory diagram of a conventional defect.

[Explanation of symbols]

10, 11; Field memory, 12, 13, 14; Line memory, 15-20; Coefficient multiplication unit, 21A; Coefficient output unit, 22; Adder unit, 23; Two-dimensional low-pass filter,
24: time axis filter, 30: motion vector detection circuit,
31: coefficient switching unit, 32: moving image / still image determination unit.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI technical display location // H04N 7/32

Claims (3)

[Claims] 1. A time axis filter for weighting and adding a plurality of pixel data that are successively arranged on a time axis of an input digital image signal, and a two-dimensional low-pass filter are provided to smooth changes in the input digital image signal. The image filter according to claim 1, further comprising a coefficient changing unit that changes a weighting coefficient of the time axis filter according to a determination signal of a moving image / still image region regarding an input digital image signal. 2. The image filter according to claim 1, wherein the time-axis filter delays the input digital image signal in units of fields to extract a plurality of pixel data that are consecutive in time. An image filter having means for weighting and adding a plurality of different pixel data in field units. 3. A motion vector detection device comprising a motion vector detection circuit and a pre-filter that is provided in the preceding stage thereof to smooth the change of the input digital image signal, wherein the pre-filter is used as the pre-filter. Alternatively, a motion vector detection device characterized by applying the image filter described in 2.