JPH0239716A - Prediction device - Google Patents

Abstract

PURPOSE:To satisfactorily regenerate an original signal by presuming and restoring the frequency band of the lost original signal by executing the prediction and presumption with an applied type outer insertion interpolating method. CONSTITUTION:An outer insertion presumed value calculator 20 obtains an absolute value ¦alpha¦ and an outer insertion presumed value SA of the inclination due to real sample values S1 and S2 and an outer insertion presumed value calculator 14 obtains an absolute value ¦beta¦ and an insertion presumed value Sa of the inclination due to real sample values S3 and S4. Next, absolute values ¦alpha¦and ¦beta¦ of the inclination are outputted to an inclination absolute value comparator 22 and compared and a smaller pair is outputted to an outer insertion presumed value selector 24. Consequently, the real sample and the predictive sample value are alternately inputted to a monitoring television set 32 and the display by the signal of a faithful high band is executed to the original signal.

Description

[Detailed Description of the Invention] [Industrial Application Field] The unfinished application relates to a prediction device that performs signal interpolation on a sampling signal sequence, and is particularly useful for improving the quality of reproduced display images by expanding the bandwidth of video signals. The preferred FJII device is described below.

[Prior Art] For example, the bandwidth of an NTSC television broadcast video signal is limited to approximately 4.2 MHz.

Since this is not limited by an image input device such as a television camera, high quality images can be reproduced or displayed if the frequency band of the transmission path is expanded.

Considering this in terms of digital sampling values of video signals, according to Nyquist's theorem or sampling theorem,
If a signal with a bandwidth of 4.2 MHz is sampled at 8.4 M5 PS (mega sample 7 seconds) or more, the information of the original signal is preserved. However, if the transmission frequency band is limited, even if the sampling density of the received signal is increased any further, it will not be possible to improve the quality.

Therefore, the sampling density is made to correspond to the limited band, and by estimating and interpolating different values between each sampling gain 1, a signal sequence of 16.8M5PS or more in total is obtained. That is, if the sampling density is doubled and this is used as a display target, it is possible to improve the quality of the raw image.

Traditionally, doubling the sampling density like this.

This has mainly been carried out by determining the 'f-measured value of the sample to be estimated by interpolating its adjacent sample values or surrounding samples.

[Problem to be solved by the invention] However, such conventional means cannot be used to restore the frequency band lost from the original signal by generating a new frequency band in the signal of the displayed image. Even if it is said that the quality of a single image is improved, it is only expected that it will improve the visibility of the image.

The present invention has been made in view of this point, and by performing predictive estimation using an adaptive extrapolation method instead of the conventional interpolation method, the lost frequency a band of the original signal is recovered. It is an object of the present invention to provide a prediction device suitable for improving the quality of an image, which can estimate and restore the original signal and perform good reproduction of the original signal.

[Means for Solving the Problems] The present invention uses two actual sample values in the past from the position to be interpolated, and calculates vi at that position, the line-extrapolated predicted value A, and the slope α of the extrapolated line. A first arithmetic processing means and a second arithmetic processing means for calculating a predicted value B linearly extrapolated to that extent δ and a slope β of the extrapolated line using two actual sample values in the future from the position to be interpolated. and the weighting coefficient γ is set to the above α
and third vi calculation processing means for determining the predicted sample value by taking into account the values of It is something.

[Operation] According to the present invention, the predicted value A linearly extrapolated to the position and the slope α of the extrapolated line are obtained by using two actual sample values located in the past from the position to be interpolated. Also, Supplement U
Using two real sample values located in the future from the n power position l,
The predicted value B linearly extrapolated to that position and the slope β of the extrapolated line are determined.

Next, the slopes α and β of the extrapolated line obtained as follows
The weighting coefficient .gamma. is determined from the value of , and then the calculation of AX.gamma.+Bx (1-.gamma.) is performed to finally obtain the predicted sample value.

If γ=1 or 0, either predicted value A or B is selected as −1−M sample f.

[Examples] Examples of the present invention will be described below with reference to the accompanying drawings. First, the basic principle of the present invention will be explained with reference to FIGS.

First, the original signal obtained from the imaging bag ri (not shown) is
The curve LA is shown in the same figure (A), and the signal that has been subjected to a predetermined band limit is shown in the same figure (A) as the curve LA.
Assume that it is as shown in B. Furthermore, this LB) Sl, S2. S3... is a sampling signal.

For example, if the reference level or “0” level is the same [4(B)
Assuming that the position t is as shown in , the sampling signals Sl, S2°S3 . ... has the size shown in the figure.

Here, as shown in U;J(D), the actual sample value S2. The sample value SA is T-D inserted during S3, and the actual sample value is S3. The case where the sample value SB is predicted and inserted during S4 will be explained as an example.

The predicted sample value SA is the actual sample value Sl.

Linear extrapolation of S2 or real sample value S3. It can be estimated by linear extrapolation of S4. Therefore, the predicted sample value SA is
The weighting of both of these is determined as the average value.

However, in the uninvention, the following processing is performed due to the need to adopt a simple determination method that reduces noise generation due to fn errors and restores the missing frequency band as much as possible.

That is, (γ) and (l
-γ), two sample values 31. Inclinations (or differences) α and β (same figure (
The absolute value of (see C) shall be used.

Here, an example in which the weight (γ) of the extrapolation point based on the set of the smaller actual sample value 1 among 1α1 and 1β1 is set to rlJ is shown in FIG.

fI! The case of '11 sample value SA will be explained as an example. In the same figure (C), the slope α due to the actual sample value 31°S2 and the actual sample value S3. Comparing the absolute values of the slope β by S4, 1α1 is smaller than 1β1.

Therefore, the actual sample value Sl and the 32 straight-line extrapolated values are adopted, and the weight (γ) thereof is rlJ. On the other hand, the difference (l−γ) on the actual sample value S3, 34 side is “0”.

In the case of the child side sample value SB, the actual sample value S2. S3 and the actual sample value S4. The absolute values of the slopes obtained in S5 are compared. The absolute value of the slope in this case is the actual sample value S4. Since the S5 side is small, the extrapolation point resulting from this is treated as a ton.

The sequence of actual sample values and predicted sample values adopted using the method described above is as shown in FIG. In this way, the signal waveform obtained by the present invention is shown in the same figure (A
) contains higher frequency components than the band-limited signal shown by the curve LB, and has a waveform that is more similar to the original signal LA.

Next, an embodiment of the present invention will be described with reference to No. 21]. In the figure, an input terminal 10 is connected to a shift register 12. Each is connected to the input side of the extrapolation estimate value calculator 14. The output side of the shift register 12 is connected to a second shift register 16 . Each is connected to the input side of the extrapolation estimate value calculator 14. Furthermore, shift register 1
The output side of 6 is connected to the third shift register 18.6. They are each connected to the input sides of other extrapolation estimate value calculators 20.

The output side of the shift register 18 is also connected to this extrapolation estimation (Iffn output 20).

The output side of the extrapolation estimate calculator 14.20 described above is a slope absolute E comparator 22.20. The output side of the slope absolute value comparator 22 is connected to the other input side of the extrapolation estimate value selector 24 .

Next, the output side of the extrapolation estimated value selection a24 is connected to the input side of the double-speed shift register 26, and the output side of the Fa shift register 16 is connected to the double-speed shift register 26.
and the input side of another double-speed shift register 28, respectively. This double-speed shift register 28
The output side of is connected to the input side of a monitor television 32 via a D/A converter 30.

Of these, the input terminal 1O is configured to receive a quantized sample sequence of a signal band-limited and transmitted from the outside.

Further, the extrapolation estimated value calculator 14.20 calculates and outputs the absolute value of the slope and the extrapolated estimated value based on the two sample values, and the absolute value of the slope is input to the slope absolute value comparator 22. , the extrapolated estimated value is manually entered into the extrapolated estimated value selector 24. Note that the extrapolation estimated value calculation fi14 is an extrapolation in the future direction (in the example of FIG. 1, the actual sample value S3
The extrapolation estimate calculator 20 performs extrapolation in the past direction (in the example of FIG. 1, extrapolation to actual sample values S3 and S4).

Next, the operation of the embodiment configured as follows will be described using the above-mentioned case 11A as an example.

Now, for example, the quantized sample value (for example, 8 bits) by 8.4M5PS, that is, the real sample value 3 in FIG.
1. S2. S3. . . . has arrived at the input terminal lO. These actual sample values are sequentially shifted in shift registers 12, 16.18 using an 8.4 MHz clock. Then, at a certain moment, the actual sample value Sl in FIG.

S2. S3 appears on the output side of soft register 1B and 16.12, respectively, and the actual sample value S4 is output to shift register 12.
appears on the input side of The situation in this case is shown in FIG.

In this state, the extrapolation estimate value calculator 14 uses the actual sample value S3. S4 is inputted to the extrapolation estimated value calculator 20.
The actual sample values Sl and S2 are respectively input. Therefore, in the extrapolated estimated value calculation Im, 20, the absolute value Iαlj5 of the slope due to the actual sample value Sl, S2, and the extrapolated estimated value SA are calculated, and in the extrapolated estimated value calculator 14, the actual sample value S3. The absolute value lβ1 of the slope and the extrapolated estimated value Sa by S4 are determined.

Next, among these, the absolute value lαβ1 of the slope is outputted to the slope absolute value comparator 22 for comparison, and the smaller set, that is, the set of actual sample values St and S2, is output to the extrapolated estimated value selector 24. Output.

Based on the notification, the extrapolated estimated value selector 24 selects the appropriate extrapolated estimated value SA, which is the extrapolated estimated value SA, and outputs it to the double-speed shift register 26. Become.

On the other hand, at this time, the actual sample value S2, which is the output of the shift register 16, is input to the double-speed shift register 28, and the contents of these double-speed shift registers 26 and 28 are as follows.
The signals are shifted by a clock of 16.8 MHz, which is twice 8.4 MHz, and are sequentially output to the D/A converter 30.

In other words, following the actual sample value, the predicted sample value or D
/A converter 30.

Therefore, when the following operation E is repeated, the actual sample value and the predicted sample value will be input to the monitor television 32 for copulation. I want to do it, monitor TV 3
2, display is performed using a high-band signal that is more faithful to the original signal than shown in FIG. 1(E), and a high-quality image is reproduced.

Note that the present invention is not limited to the above embodiments; for example, in the above embodiments, the weighting coefficient γ is set to "1".
, or you can set it to any other value.

Further, the circuit configuration shown in FIG. 2 can also be modified in various ways so as to achieve the same effect.

Further, 1. Although the above embodiment is a case in which the present invention is applied to the transmission of an image signal, it may be applied to other signals.

[Effects of the Invention] As explained below, according to the present invention, the value obtained by extrapolating the v1 line to the position using two past actual sample values from the position to be interpolated is A, and the slope of the extrapolated line is α is the value linearly extrapolated to that position using two actual sample values in the future from the position to be interpolated, B is the slope of the extrapolation line, and β is the weighting coefficient.
and the above α.

Determine γ by considering the value of β, and also
Since we decided to obtain the predicted value as (1-γ), it is possible to successfully predict and restore the frequency band of the signal that is lost due to the band limitation of the transmission path with a comparatively simple configuration and with less noise generated due to errors. There is an effect.

[Brief explanation of the drawing]

FIG. 1 is an explanatory diagram showing the basic principle of the invention, and FIG. 2 is a device block diagram showing an embodiment of the invention. 12.16.18...Shift register, 14°20.
... Extrapolation estimated value calculator, 22... Slope absolute value comparator,
24... Extrapolation estimated value selection machine, 26.28-... Double speed shift register, 32... Monitor television, Sl. S2. S3...Actual sample value, SA, SB, SC...
- Predicted sample value. Patent Applicant: Japan Victor Co., Ltd. Representative: Kunio Kakiki Figure (Figure CD) Figure (E)

Claims (1)

[Claims] Using actual sample values before and after the position to be interpolated,
In a prediction device that calculates a predicted sample value, a first calculation process that uses two past actual sample values from the position to be interpolated to calculate a predicted value A linearly extrapolated to that position and a slope α of the extrapolated line. A second method for calculating the predicted value B linearly extrapolated to that position and the slope β of the extrapolated line using the means and two actual sample values in the future from the position to be interpolated.
and determines the weighting coefficient γ in consideration of the values of α and β, and calculates A×γ+B from that value.
A prediction device comprising: third calculation processing means for calculating the predicted sample value by calculating x(1-γ).