JPH0654976B2 - Motion vector detector - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a motion vector detecting device, and more particularly to a device for correcting a moving part of an image such as system conversion or frequency band compression in a television broadcasting system.

[Prior Art] When the standard system of television broadcasting is different,
When exchanging broadcast programs with each other, a standard method conversion is required, and a method conversion device for that purpose is used.

In the system conversion when the number of scanning lines or the number of fields is different between the standard systems, the resolution of the converted television image deteriorates due to the conversion process, and the motion in the moving image part is unnatural. However, there is a problem that the disturbance such as judder which is not smooth occurs, and therefore the image quality is greatly deteriorated and must be improved.

Conventionally, in this method conversion processing, in order to correct a moving image portion, a motion vector is obtained based on the correlation between the image signal levels of the current frame and the immediately preceding frame, and the motion image is calculated using this motion vector. It was corrected.

As described above, the motion vector is used for the motion correction, and the pattern matching method is used to detect the motion vector.

This is because, for example, when the object A moves at a speed of υ during one frame period as shown in FIG.
Is shifted by v, the image A coincides with the current frame image A ', and as a result, the motion vector v is detected based on the principle that the inter-frame difference in the luminance level of the pixel is the smallest.

This pattern matching method is a determination method based on the fact that there is a correlation between moving images between frames and is based on detecting the difference between pixel signal levels between frames, and detects an optimum motion vector.

Therefore, for each type of motion vector that is possible in the image area, the difference (interframe difference) between the pixel signal level of the current frame and the pixel signal level of the previous frame that are in succession is calculated, and the absolute value The sum of the values is used to evaluate the correlation.

That is, since the motion vector having the smallest sum has the highest correlation, this is set as the correct motion vector for the image area.

On the other hand, when the hardware is used to perform the calculation using all the pixels in the frame, the scale of the device becomes unnecessarily large. Therefore, regarding the pixel of the sub-sampled point called the representative point, Only arithmetic processing.

That is, conventionally, the motion vector is detected by the configuration shown in FIG. 3, and the state of the vector detection region and the representative points are shown in FIG.

As shown in FIG. 4, the representative points are horizontal Δs and vertical Δs.
(As mentioned above, both Δs and Δ can be set larger than one pixel interval because the hardware scale must be reduced.)
Pixels roughly subsampled at intervals of

In FIG. 3, 3 is a frame memory for storing one frame of video signal, 4 is a representative point setting circuit, and there is a possibility that the video signal from the frame memory 3 can be read out and taken for each area called a sample vector. The signal of the level of each representative point corresponding to the reference position of a certain finite number of vectors is output. The subtractor 5 calculates, for each representative point, the difference (frame difference) between the pixel of the current frame corresponding to the position of the representative point offset by the sample vector and the representative point in the immediately preceding frame. After this difference is subjected to absolute value conversion by the absolute value circuit 6, the frame difference addition circuit calculates the frame difference signals calculated at the respective representative points for each sample vector, and the minimum value detection circuit 8 calculates the frame difference addition circuit. The sample vector that gives the minimum value of the added values from 7 is set as the motion vector of this area and is output.

[Problems to be Solved by the Invention] Since the intervals Δs and Δ between the representative points have a large amount of calculation as described above, it is preferable to take them as large as possible in consideration of the scale of hardware.

On the other hand, as shown in FIG. 5, in the case of a thin moving object 9 as indicated by diagonal lines, the motion region does not reach the representative point, so that the motion vector υ is not detected, and the device malfunctions. That is, although the representative points are used to reduce the scale of the hardware, the representative point intervals are small, so the number of representative points that effectively works for vector detection is small in small motion images, and in some cases, There is a possibility that the representative point belonging to the motion area becomes zero, and the motion vector detection accuracy deteriorates.

Therefore, an object of the present invention is to enable accurate motion vector detection without increasing the scale of hardware.

[Means for Solving the Problems] Therefore, the present invention provides an offset means for spatially offsetting pixels used for vector detection over a plurality of frames, and an offset means for each state of the offset over a plurality of frames. The calculation means for calculating the data for calculating the motion vector for each pixel after the offset and the calculation value obtained by the calculation means are convolved over all the offset states, and based on the value obtained by the convolution. And a means for detecting a motion vector.

Further, according to the present invention, an offset means for spatially offsetting pixels used for vector detection over a plurality of frames, and a motion vector calculation for each pixel after the offset obtained by the offset means for each state of the offset over a plurality of frames. And a means for calculating the motion vector based on the value obtained by the convolution of the calculated values obtained by the calculating means over all the offset states. It is characterized by

[Operation] The present invention does not increase the scale of hardware by offsetting pixels used for vector detection over a large number of frames and equivalently increasing the density of pixels used for vector detection within one frame. Stable vector detection.

[Embodiment] The basic structure of the present invention is shown in FIG. 1 and the embodiment is shown in FIG.

In FIG. 1, the offset-type vector detection unit 1 offsets a representative point within a frame and over several frames to obtain a frame difference addition value or a motion vector for detecting a motion vector in each frame. Next, in the convolution circuit 2, this value in each frame is convolved, and in the case of the frame difference addition value, the minimum value is finally detected to obtain the motion vector (the offset type vector detection unit 1 calculates the frame difference. When outputting the added value, the convolution circuit 2 includes the minimum value detection).

An example of the offset of the representative point is shown in FIG. When the o mark is the mth frame, the Δ mark is the m + 1th frame, the x mark is the m + 2th frame, and the □ mark is the representative point of the m + 3th frame,
This is 4 times the density (almost the same scale of hardware) as compared with the conventional representative point interval shown in FIG.
Improvements have been made in terms of points.

(1) The increase in the number of representative points makes it stronger against noise.

(2) The increase in density eliminates the motion vector non-detection state as shown in FIG. 5 (reduction of aliasing distortion).

FIG. 7 shows the hardware configuration when the representative points shown in FIG. 6 are taken.

In FIG. 7, a frame memory 10 and a representative point setting circuit 1
1, frame rank detection circuit 12, subtractor 13, absolute value circuit 14
Further, the frame difference addition circuit 15 is added to the offset type motion vector detection unit 1 of FIG.
19 and the minimum value detection unit 20 correspond to the convolution circuit 2 in FIG.

The frame order detection circuit 12 discriminates the frame based on the vertical synchronizing signal and supplies a signal for offsetting the representative point to the representative point setting circuit 11. The representative point setting circuit 11 sets the representative point for each frame based on the signal from the circuit 12, for example, the sixth point.
Offset as shown. In the subtractor 13, for each offset representative point, the position of this representative point is determined as the pixel of the current frame corresponding to the point offset by the sample vector.
Calculate the difference from the representative point in the frame 1, 2 or 3 frames before. After converting this difference to an absolute value in the absolute value circuit 14,
The frame difference adding circuit 15 adds the frame difference signals calculated at the respective representative points for each sample vector.

The added value signal from the frame difference addition circuit 15 is input to the frame memory 16. The frame memories 16, 17, and 18 have the same capacity as the type of the sample vector, and the shift lock is one frame cycle. The memory contents are moved from 17 to the frame memory 18. That is, from the frame difference addition circuit 15, the frame memory 16, the frame memory 17, and the frame memory 18, respectively, the current frame, 1 frame before, 2 frames before, 3
The frame difference addition value corresponding to each sample vector before the frame is output.

The adder 19 adds the outputs of the frame difference addition circuit 15 and the frame memories 16, 17, and 18. That is, the frame difference addition values of the current frame, one frame before, two frames before, and three frames before are convolved. The minimum value detection unit 20 detects the minimum value of the frame difference addition value corresponding to each sample vector output from the adder 19, and outputs the sample vector giving the minimum value as the motion vector in the area.

With such a configuration, it is possible to improve the S / N ratio of 6 dB for the nozzle compared to the conventional method of FIGS. 3 and 4, and to extend the detection range up to twice the horizontal and vertical banding for folding back. Become.

Here, the frame memories 16, 17, and 18 have a shift clock of 1
Since the frame period is used, the memory capacity only needs to be one of the sample vectors and the memory capacity is very small, and the hardware scale does not increase significantly compared to that in FIG.

On the other hand, in the method of outputting the motion vector for each frame in the offset type vector detecting unit, the minimum value detecting unit in FIG.
Reference numeral 20 is included in the offset type vector detection unit 1, and the capacity of the frame memories 16, 17, 18 is one vector value, which is very small.

In the present embodiment, an example of offsetting the representative points over four frames has been described. However, if the offset is performed within the representative point frame and further between multiple frames, the effect of improvement becomes large accordingly.

[Effects of the Invention] As described above, according to the present invention, the pixel density for detection is increased without complicating the hardware configuration and without significantly enlarging it, and the motion vector is detected stably. be able to.

[Brief description of drawings]

FIG. 1 is a diagram showing a basic configuration of the present invention, FIG. 2 is an explanatory diagram of a pattern matching method, FIG. 3 is a block diagram showing a conventional motion vector detection device, and FIG. 4 is a conventional motion vector detection. FIG. 5 is a diagram showing the representative points used in FIG. 5, FIG. 5 is an explanatory diagram showing that motion vector detection cannot be performed by the conventional method, and FIG. 6 shows an example of the arrangement of offset representative points used in the present invention. FIG. 7 is a block diagram showing an embodiment of the present invention. 10 …… frame memory, 11 …… representative point setting circuit, 12 …… frame order detection circuit, 13 …… subtractor, 14 …… absolute value circuit, 15 …… frame difference addition circuit, 16,17,18 …… Frame memory, 19 ... Adder, 20 ... Minimum value detector.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Yuichi Ninomiya 1-10-11 Kinuta, Setagaya-ku, Tokyo Inside the Broadcasting Technology Laboratory, Japan Broadcasting Corporation (72) Taiji Nishizawa 1-10-10 Kinuta, Setagaya-ku, Tokyo No. 11 Inside the Japan Broadcasting Corporation Broadcasting Technology Laboratory

Claims (2)

[Claims] 1. An offset means for spatially offsetting pixels used for vector detection over a plurality of frames, and a motion vector calculation for each pixel after the offset obtained by the offset means for each state of the offset over a plurality of frames. And a means for convolving the operation values obtained by the operation means over all offset states and detecting a motion vector based on the values obtained by the convolution. A motion vector detection device. 2. An offset means for spatially offsetting pixels used for vector detection over a plurality of frames, and a motion vector calculation for each pixel after the offset obtained by the offset means for each state of the offset over a plurality of frames. And a means for convolving the operation values obtained by the operation means over all offset states and detecting a motion vector based on the values obtained by the convolution. A motion vector detection device characterized by the above.