Device and method for discriminating movement in a video signal.

The invention relates to a motion detector and a method for determining dynamic picture components of a video signal.

To determine dynamic image proportions are motion detectors known in which the vectorial direction of motion and the speed of an image point can be determined. Given a matrix of example is set 3x3 pixels around the pixel. This 3x3 matrix is ​​to a matching matrix of the previous image. To determine a motion all the pixels of an example are 15x15 matrix in the vicinity of the pixel tested for their direction of movement and speed to determine the area in which the correlation matrix has a maximum. In this manner, a direction of movement of the image point is determined and the speed at which the image motion occurs. the direction of movement of blocks So it is determined. The values ​​obtained are used as additional information to broadcast an HD / TV signal. In the transmitter provided by the moving directions and speeds are selected as 16 vectors per image or field. However, only four moving directions and speeds are allowed in the receiver per block. This means that, only four moving directions and speeds are allowed in a block having as 32x32 pixels. This is compared to the very large number of moving directions and speeds a very strong restriction, which would for example, result in a push Up, in which each pixel has a different direction and speed to unacceptable playback. In such a case, the additional information is switched off by the receiver. It is switched to an interpolation, for example, by averaging two lines to form an intermediate can exist. For such a well-known motion detector, a high circuit complexity and rapid signal processing is required. In US-A-4,667,233 a motion detection is performed for two-dimensional pixel blocks. For an extensive hardware and high speed processing is required. Since only whole blocks for motion is detected, the motion of small objects can not be detected.

The object of the invention is to provide a motion detector which detects with a minimum amount of circuitry and a low signal processing speed reliable pixel by pixel movement, the hardware and the control thereof can be exploited for a dependent on the result of motion detection vertical filtering. This object is achieved by the features specified in claim 4 features. Advantageous developments of the invention are described in the dependent claims. The corresponding method is defined in claim 1st

The motion detector according to the invention are processed in parallel, in each case shifted by a clock, the values ​​of vertically superposed pixels. There are required according to the number of used for the motion detection pixels motion detectors. For the motion detection of a pixel, the absolute values ​​of differences of vertically adjacent weighted pixels of two temporally following frames are determined. These absolute values ​​are summed and the sum of all values ​​is weighted. If it is found that the sum exceeds a predetermined threshold, this pixel is detected as a moving pixel. For improved weighting of this information in relation to the neighboring settlements in the horizontal direction pixels, the results of a clock shifted and latched motion detectors outputs are examined and in the event that a number of motion detectors motion has detected, the pixel as moving is treated pixel. Advantageously, it follows that in the inventive determination of the motion of a pixel noise, as can occur, for example in large parts of the image, is also not detected as motion if one of the pixels from the motion detector, an extremely high value is determined, while all other pixels below the motion threshold. Because for each pixel a logical value is determined, which indicates a statement whether motion has been detected for this pixel.

A further advantage lies in the signal processing speed. Since there are according to the number of used for the motion detection pixels motion detectors, a motion detector output is determined for one pixel per clock processing. Thus, no increase in processing speed is required. The circuitry is also low.

Hereinafter, an embodiment will be explained with reference to the drawings:<dl id="dl0001"><dt>Fig. 1</dt><dd>Motion detector</dd><dt>FIG. 2</dt><dd>Averaging motion detector signals.</dd></dl>

Fig. 1 shows a motion detector according to the invention. At the input of a delay chain which is on 1-4 to be examined signal. This signal can be both a luminance and a chrominance signal. This signal represents the image contents of the previous image n-1. At the entrance of a second delay chain 5-8 is a signal of the current frame n. Each of the input signals of the delay chain 1-4, 5-8, and the outputs of these delay lines 1-4, 5-8 a subtractor 9-13 are supplied. The outputs of subtractors 9-13 are absolute values ​​formers supplied 14-18, wherein the respective successive outputs of the absolute value be 14-18 adders 19-22 supplied. At the output of the adder 22 is then the sum of all absolute values. This output 22 leads to a switch 38, which summarizes the outputs of lying parallel motion detectors the same structure and depending on the current Berechungsstand switched to a comparator 24th This comparator 24 provides a signal with respect to the motion of the picture element when the sum of the absolute values ​​from the output of the adder 22 exceeds a threshold value.

Fig. 2 shows a circuit for combining motion detector signals. The information of the image content of the current image and the previous image are applied to a motion detector 30 is supplied, the construction of which is described in FIG. 1. There are a total of five motion detectors parallel exists. The information about the current picture is the same vertical filter 40 is supplied, for example, 5x are also present in parallel. The output of the motion detector 30 leads to a delay chain 31-34. Each input of the delay elements 31-34 and the output of the delay element 34 is connected to an input of the adder and comparator 35th The output of the adder and comparator 35 leads to an input of the changeover switch 42. The vertical filter 40 are connected via a delay circuit 41 to inputs of the change-over switch 42nd At the output of the changeover switch 42, the filtered signal of the current frame n can be tapped. Depending on the result, whether the adder and comparator 35 has detected a static or dynamic information for the current pixel, the pixels are switched to the output of the switch 42, through the filtering for static or filtering for the dynamic case have.

With this type of combination of several motion detector signals is avoided that the environment of the to be examined pixel by eg 5x5 pixels must be examined separately, ie, it prevents 25 pixels subtracted formed the amount, and the totals are added. The same result is achieved in that 5 image points which lie vertically one above the other, be examined for their movement towards the determination of the movement of the center pixel, and that the final results of five adjacent motion detector signals are analyzed for motion. So it will be processed and the result of five motion detector signals stored only five pixels. The storage takes place in that four delay lines are used 31-34, so that the same is the result of five motion detectors available.

Since not the average value has to be determined, but only an information whether there is motion or not, it is sufficient to binary information available movement / no movement to save. This binary information is summed in the adder and comparator 35, and in the case that the sum ≦ 2, is used for the current picture element, which rests in the output of the delay element 41, "no motion" is detected. If the sum is&gt; 2, this means that it is a moving pixel in this pixel. Advantageously, due to the fact that, for example, would be detected at a threshold of 30, on the movement of an image point, not the absolute values, but the binary information is evaluated. If for example four pixels have a value below the threshold, however, a point lying by trouble far above the threshold, majoresiert this value is not the remaining four pixels. It represents only one equivalent information. In this case, the pixel would be detected as a static pixel.