JPH0691655B2 - Motion vector detection circuit - Google Patents

Description

The present invention relates to a motion vector detection circuit for image signals,
In particular, the present invention relates to a motion vector detection circuit having a function that can easily cope with an increase in the motion vector detection range.

(Summary of the Invention) The present invention relates to a motion vector detection circuit based on a pattern matching method, and time-parallelly finds a frame difference for each sample vector between a current signal and a previous frame representative point within the vector detection range.

With this, even if the range of motion vectors that can be detected is expanded, the motion vector can be detected only by connecting the hardware subordinately without changing the circuit configuration.

(Prior Art) Conventionally, in a motion vector detection circuit using a pattern matching method, pixels called subsample representative points of an image of a previous frame are moved horizontally or vertically by the number of sample vectors according to the sample vectors. , The absolute value of the frame difference from the image of the current frame is obtained, the absolute values of the frame differences belonging to the same sample vector are added for all the representative points, and the sample vector having the smallest added value is detected as the motion vector. .

(Problems to be Solved by the Invention) In the conventional motion vector detection circuit, since the frame difference point between the previous frame representative point pixel and this and the current signal pixel is obtained, a minimum vertical angle is obtained even with respect to the current signal. Number of direction sample vectors (vertical value of motion vector detection range)
Since the buffer memory for the number of lines corresponding to is needed and the vector detection range is the number of sample vectors as it is, if the vector detection range is expanded further than the correlation detection area, the correlation with the representative point is calculated. Since there are pixels of the current signal that are used more than once, the capacity of the buffer memory for the current signal increases. Also, since the motion vector must be detected in real time,
So far, we have been performing frame difference absolute value addition in parallel for some representative points.However, if the vector detection range is further expanded, the hardware configuration must be changed to perform that detection in real time. There are cases in which they are forced to do so.

Therefore, the present invention solves the above-mentioned drawbacks, and by paying attention to the current signal, by time-parallel finding the frame difference absolute value for each sample vector between the current signal and all previous frame representative points within the vector detection range, Even if the range of motion vectors that can be detected is expanded without the need for a buffer memory for signals, the purpose is simply to connect additional hardware to the hardware structure before expansion without changing the circuit structure. Of the motion vector detection circuit to be achieved.

(Means for Solving the Problem) The vector detection circuit of the present invention for that purpose, that is, in the motion vector detection circuit by the pattern matching method, is a representative that the circuit should calculate the frame difference between the pixel and the pixel of interest of the current signal. A representative point sub-sampling unit that sub-samples point pixels, a representative point buffer memory unit that temporarily stores the sub-sampled representative point pixels, and shifts and outputs the representative point pixels according to the current signal belonging to the correlation detection region. A representative point delay circuit and an output unit for simultaneously outputting a limited number of representative point pixels belonging to the motion vector detection range among the shifted representative point pixels in accordance with the current signal belonging to the correlation detection area; Frame difference between the pixel of interest and the representative point of the previous frame Multiple frame differences that add the total representative points according to the sample vector Addition and memory unit, minimum value detection unit that detects the minimum value of the added value of the frame difference absolute value, and a motion vector based on the memory of the frame difference addition and memory unit that indicates the minimum value and the address of the address generation unit And a vector value conversion unit for converting the value into a value.

(Example) The present invention will be described in detail below with reference to the accompanying drawings.

The positional relationship between the current signal and the representative point to be calculated is shown in FIGS. 2 (a) and 2 (b). White squares (□) and black squares (■) in the figure represent representative points. The interval between the representative points is now 1 pixel in the horizontal direction and m lines in the vertical direction. By detecting the vector and setting the range, the representative point to be calculated for the current signal can be obtained. Conversely, if the representative point to be calculated for the current signal is determined, the range of the detectable vector is uniquely determined. Assuming that the number of representative points to be calculated for the current signal is px in the horizontal direction and py in the horizontal direction, the range of detectable vectors is − {(l × px) / 2−1} ≦ x in the horizontal direction x. ≦ + (l × px) / 2 (1) The vertical direction y is − {(m × py) / 2−1} ≦ y ≦ + (m × py) / 2 (2).

If the number of representative points to be calculated for the current signal is increased,
The detection range of the vector is expanded. In FIGS. 2 (a) and 2 (b), the representative point indicated by the black square mark is the calculation target point for the current signal existing in the shaded area in the figure. In both cases, the number of representative points to be calculated is 3 in the horizontal direction and 2 in the vertical direction.
The total is 6 pieces. At this time, the shaded area in the figure is called the correlation detection area for the representative points of the six black square marks, and the entire area to which the six black square marks belong is called the motion vector detection range. Therefore, the range of the vector that can be detected is Equation (1),
Using the equation (2), the horizontal direction x is-{(3l / 2) -1} ≤x≤ + (3l / 2) The vertical direction y is-{(2m / 2) -1} ≤y≤ + ( 2m / 2).

When the current signal moves from the position shown in FIG. 2 (a) to the block in the right direction as shown in FIG. 2 (b), that is, when the correlation detection region moves, the representative point to be calculated also shifts to the right. That is, the representative points of the six black square marks shown in FIG. 2 (a) are shifted to the representative points of the six black square marks shown in FIG. 2 (b), that is, the motion vector detection range is shifted, and this control is performed. Are controlled by the buffer memory 2 described later. Thus, this movement is sequentially repeated in the horizontal direction and the vertical direction, and the absolute value of the frame difference between the current signal and the representative point to be calculated for the current signal is calculated. The values are added, the minimum value is detected, and the vector is obtained by the so-called pattern matching method.

FIG. 1 shows a block diagram of a configuration of an embodiment of a vector detection circuit according to the present invention. After the video input signal is subsampled by the representative point subsampling unit 1, it is delayed by one frame via the buffer memory 2. This order may be reversed. From the representative point signal of the previous frame, which is the output of the buffer memory 2, the representative point delay circuit and the output unit 3 are used to output the representative point to be calculated for the current signal.
The representative points of the black square marks are output, and the frame differences from the current signal are obtained by the subtractors 4 to 9 in parallel in time.

On the other hand, the address value generated from the address generation unit 16 and the frame difference addition and the memory units 10 to 15 add the frame difference between the current signal and the representative point for each sample vector and store it in the memory.

After the frame difference addition is completed for the video of one frame, the frame difference addition value and the frame difference addition value stored in the memory units 10 to 15 are sequentially read, and the minimum value detection unit 17 detects the minimum value. .

Further, the vector value conversion unit 18 obtains a motion vector value from the frame difference addition that outputs this minimum value and the address value of the memory unit. From the above, the subtractor 4
It will be understood that it is possible to easily widen the detection range of the vector by expanding the number of .about.9 and the frame difference addition and the number of the memory units 10 to 15.

(Effect of the Invention) In the conventional motion vector detection circuit, when the motion vector detection range is expanded beyond the correlation detection region, there are pixels of the current signal that are used more than once in the calculation of the correlation with the representative point. However, the hardware configuration had to be complicated due to the increase of the buffer memory of the current signal and the limitation of the real-time processing, but if the motion detection circuit according to the present invention is used,
A buffer memory for the current signal is not required, and it is easy to detect a wide range of motion vectors simply by adding a part of the circuit to the circuit being used.

[Brief description of drawings]

FIG. 1 shows a block diagram of a configuration of an embodiment according to the present invention, and FIGS. 2 (a) and 2 (b) show two different representative points within a detection range of a vector for a current signal. It is a figure which shows the aspect. 1 ... Representative point sub-sampling section 2 ... Buffer memory 3 ... Representative point delay circuit and output section 4-9 ... Subtractor 10-15 ... Frame difference addition and memory section 16 ... Address generating section 17 ... Minimum value detector 18 ... Vector value converter

 ─────────────────────────────────────────────────── ─── Continuation of the front page (56) Reference JP 61-201581 (JP, A) JP 62-230178 (JP, A) JP 62-230180 (JP, A)

Claims (1)

[Claims] 1. A motion vector detection circuit according to a pattern matching method, which circuit includes a representative point sub-sampling section (1) for sub-sampling a representative point pixel for which a frame difference between the pixel of interest and a current signal is to be calculated. , A representative point buffer memory unit (2) for temporarily storing the sub-sampled representative point pixel and shifting and outputting the representative point pixel according to a current signal belonging to the correlation detection area, and a current signal belonging to the correlation detection area. Accordingly, a representative point delay circuit and an output unit (3) for simultaneously outputting a finite number of representative point pixels belonging to the motion vector detection range among the shifted representative point pixels, a target pixel of the current signal, and a previous frame representative A plurality of frame difference addition and memory units (10 to 15) for adding the absolute value of the frame difference (4 to 9) to the points by all the representative points according to the sample vector; A minimum value detection unit (17) for detecting the minimum value of the added values of the frame difference absolute values, the memory of the frame difference addition and memory unit (10 to 15) indicating the minimum value, an address generation unit (16), and an address. And a vector value conversion unit (18) for converting the motion vector value into a motion vector value.