JPS622515B2 - - Google Patents

Description

Detailed Description of the Invention The present invention calculates the amount of movement of the object per frame from the television signal obtained by imaging a moving object, and calculates the amount of movement of the object per frame, and then calculates the amount of movement of the object per frame by compensating for the amount of movement (following the movement). The present invention relates to an image memory device for obtaining television signals.

The television signal of one frame before is conventionally used, for example, in an interframe coding device that predicts the television signal of the current frame from the television signal of one frame before, and in a noise removal device such as a video tape recorder.

These devices all use information about the movement of the subject in the form of frame time difference signals, so
When the movement of the subject is small, the former has high coding efficiency and the latter has a large noise suppression effect, but when the movement of the subject becomes large, the amount of information to be transmitted increases and the coding efficiency becomes significantly lower. In the latter case, there was a problem that the noise suppression effect deteriorated.

In order to improve this problem, an image memory device has been developed that makes more active use of motion information and outputs a television signal of the previous frame that follows the motion (see Japanese Patent Publication No. 49-4564). The prior art will be briefly explained below, taking as an example an interframe encoding device using the image memory device.

FIG. 1 shows the configuration of an interframe encoding device that can follow the above-mentioned large movements. This device measures the direction and speed of the object's movement, calculates the position of the object one frame before the pixel to be encoded in the current frame, and encodes the difference from this calculated signal as motion information. is configured to do so. That is, the input television signal x ₀ to be encoded for the current frame, the signal y ₀ of the previous frame, and the signal y _o of the previous frame n pixels apart.
Calculate the correlation with a number of correlation circuits, select the signal with the highest correlation (hereinafter referred to as the predicted signal) _yN , and use the difference between this and the signal _x0 of the current frame. encode.

In FIG. 1, the difference between the television signal x ₀ input to the input terminal 1 and the predicted signal y _{N of one frame before following the motion, which will be described later, is calculated by the subtraction circuit 2, and the difference between the television signal x 0} inputted to the input terminal 1 is The signal is encoded via the circuit 4, and the encoding result is output to the output terminal 5. Further, by adding the output of the quantization circuit 3 and the predicted signal y _N in the adder circuit 6, a locally decoded signal x ₀ ' corresponding to the input television signal x ₀ is obtained. This locally decoded signal is stored in a frame memory 8 operated by an address signal W generated by an address counter 7, read out one frame later by an address signal R, and used as a prediction signal for the next frame. The signal read out from the frame memory 8 is input to a delay circuit 9 having six delay elements, and a signal y ₀ exactly one frame period before the input television signal x ₀ and a signal y _{±1 of pixels before and after it are inputted to a delay circuit 9} having six delay elements. , y _±2 , y _±3 , y _±i are obtained, and the switching circuit 10
and a correlation calculation circuit 11 that selects a predicted signal. The correlation calculation circuit 11 consists of a large number of correlation circuits and a selection circuit that selects the correlation circuit that shows the maximum correlation value, and a control signal N that selects the signal (predicted signal) y _N that shows the maximum correlation from the signals y _±i. Output. In other words, the locally decoded signal x _{0 '} corresponding to the input television signal x ₀ and the signal group one frame before
The correlation between y ₀ , y _±1 , y _±2 , and y _±3 is calculated, a control signal N is generated for extracting the signal with the largest correlation, and this control signal N is used to control the switching circuit 10 . is controlled to select the predicted signal _yN .

In the conventional example above, in order to simplify the explanation, a circuit configuration that only follows movement in the horizontal direction is shown.
Generally, it is necessary to follow a subject moving in an arbitrary direction. Movement in the vertical direction can be realized by setting the delay time of the delay element of the delay circuit 9 to one horizontal scanning period. For example, in order to follow an object that moves at m pixels/frame in both the horizontal and vertical directions, the delay circuit 9 includes 2m lines and outputs 2m x 2m signals, and the correlation calculation circuit 11 outputs 2m x 2m signals. It will include a correlation circuit.

With the conventional configuration shown in Figure 1, it is possible to follow a subject that moves horizontally at a speed of 3 pixels or less per frame, and it is possible to obtain highly accurate predicted values.
It is not possible to follow objects that move at even faster speeds.

In order to be configured to correctly follow intense movements in which the amount of movement per frame is several pixels or more, it is necessary to prepare an extremely large number of correlation circuits.

That is, the scale of the delay circuit 9 and the correlation calculation circuit 11 is approximately proportional to the product of the horizontal and vertical speeds per frame period that can be tracked, and there is a problem in that when the trackable speed is increased, the circuit scale becomes extremely large. Although not shown, in another example of interframe coding that follows motion, a large number of prediction circuits that use y ₀ , y _o , etc. as predicted values are prepared instead of the above-mentioned large number of correlation circuits, and one of them is Some devices are configured to select the one with the smallest prediction error, but in this case, an extremely large number of prediction circuits are required to follow intense movements, and in any case, the scale of the device is large. Become.

SUMMARY OF THE INVENTION It is an object of the present invention to solve the above-mentioned problems and provide an image memory device having a motion compensation function that can accurately follow a subject even in intense motion without increasing the circuit scale.

In order to achieve the above object, the motion compensation means of the present invention combines the input television signal (pixel of the current frame) x ₀ with the motion-compensated television signal of the previous frame (predicted signal) obtained from the frame memory. Detect motion compensation error information from _yN ,
By correcting the amount of movement per frame using this error information, a motion-compensated television signal of the previous frame is obtained.

The present invention will be explained in detail below using Examples. FIG. 2 shows a principle block diagram of an image memory device implementing the motion compensation means of the present invention. In the same figure, a television signal input to an input terminal 12 is outputted from an address counter 13 operating in synchronization with the television signal, that is, a delay circuit is inserted into a predetermined storage location of a frame memory 14 designated by a first address signal W. 15 and are sequentially stored. The contents of the frame memory 14 are read from the location specified by the second address signal R, which is independent of the first address signal W, and the output terminal 16 receives a motion compensated signal, i.e.
A signal at the location of the same subject one frame before the television signal input to the input terminal 12 is obtained. The second address signal R for reading a signal that compensates for the amount of movement due to movement is the write address signal W.
This can be obtained by adding the contents of the register 17 that stores the amount of movement per frame period in the adder circuit 18. Then, the input television signal and the motion-compensated frame memory output are input to the error detection circuit 19, and by correlating both signals, it is possible to determine whether the amount of movement stored in the register 17 is correct or slightly different. If they are different, the contents of the register 17 are changed in a direction that makes the correlation stronger. The feature of the motion compensation means of the present invention is that the motion compensation means is configured to minimize motion compensation errors using the feedback loop. Note that the delay circuit 15 is
This is to prevent the signal of the previous frame from disappearing when the subject moves downwards on the screen, and the storage capacity is sufficient as long as it can store the signal corresponding to the maximum amount of movement per frame. Alternatively, it is configured using buffer memory or the like.

FIG. 3 shows a configuration example in which the image memory device having a motion compensation function of the present invention is applied to an interframe coding device.

Similar to FIG. 1, this figure only shows movement in the horizontal direction to simplify the explanation. In the figures, the same reference numerals indicate the same or equivalent parts, including a subtraction circuit 2, a quantization circuit 3, an encoding circuit 4,
Since the adder circuit 6 performs the same operation as the conventional device shown in FIG. 1, a description thereof will be omitted. The locally decoded signal x ₀ ' obtained from the adder circuit 6 is sent to the delay circuit 15.
The signal is input to the frame memory 14 via the frame memory 14. The delay circuit 15, frame memory 14, address counter 13, register 17, and addition circuit 18 are connected in the same way as in FIG. 2 and perform the same operations. Now, the motion obtained from the frame memory 14 is compensated for 1
The signal before the frame is input to a delay circuit 20 consisting of two delay elements, and a signal y _N that is compensated for the movement of one frame before corresponding to the input signal x ₀ and signals y _N±1 before and after it are obtained. There is. The output signal y _N is used as a predicted value of the input signal x ₀ . At the same time, the output signals y _N , y _N±1 are input to the correlation calculation circuit 21 . The correlation calculation circuit 21 has three correlation circuits and a circuit for selecting the signal showing the maximum correlation, and selects the signal showing the maximum correlation with the local decoded signal x ₀ ' (y _N , y _N+1 , y _{N- 1} ) and generates a motion compensation error signal (0, +1, -1) corresponding to the selected signal.

The contents stored in the register 17 are corrected using this motion compensation error signal to prepare for prediction of the next pixel signal.

Note that the delay circuit 20 and the correlation calculation circuit 21
is a portion corresponding to the error detection circuit 19 in FIG.

In the configuration of the apparatus of the present invention described above, the speed information of the object is stored in the register 17, and only the error between it and the movement amount of the input signal is detected by the error detection circuit 1.
9 and the error amount is fed back as negative feedback, so even if the subject moves rapidly, it is possible to follow it without increasing the circuit scale.

As described above, the correlation calculation circuit 21 can be realized with the same circuit configuration as the conventional circuit 11 shown in FIG. 1, and the number of calculation circuit structures can be reduced.

As described above, according to the present invention, it is possible to realize a motion compensation means capable of following the motion of a subject regardless of its motion speed, with a circuit scale significantly smaller than that of the conventional art. Furthermore, according to the present invention, since the speed of the object is stored in the register, it is possible to detect the speed using this, which not only improves the efficiency of interframe encoding, but also for various purposes such as noise removal using the speed of the object. available for use.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of a conventional interframe coding device that follows motion, Fig. 2 is a principle block diagram of an image memory device having a motion compensation function of the present invention, and Fig. 3 is a block diagram of an image memory device having a motion compensation function of the present invention. A configuration diagram of an applied interframe encoding device is shown. 13... Address counter, 14... Frame memory, 15... Delay circuit, 17... Register,
18...Addition circuit, 19...Error detection circuit, 20
...Delay circuit, 21...Correlation calculation circuit.

Claims (1)

[Claims] 1. In an image memory device for calculating the amount of movement of the object per frame from the television signal obtained by imaging a moving object, and obtaining the television signal of one frame before the movement amount is compensated. storage means capable of storing at least one frame of television signal as means for compensating the amount; and an address counter generating a first address signal for storing the input television signal in a predetermined location of said storage means; reading means for generating a second address signal for reading out a television signal stored in a location independent from the predetermined location; an error detection means for calculating the amount of compensation error; a register for storing the amount of movement per frame corrected by the amount of motion compensation error obtained from the detection means; 1
an adder circuit for adding the address signals of the readout means, and the output of the adder circuit is added to the second
An image memory device having a motion compensation function characterized in that an address signal of