JPH08331421A - Picture quality enhancer - Google Patents

Abstract

PURPOSE: To provide the ringing effect of a little distortion to a signal by a separating means for the respective high and low frequency components of a video signal for reproducing two-dimensional horizontal and vertical images, envelope extracting means for the high frequency, and window function forming means. CONSTITUTION: The inpulse waveform of most ringing is inputted as an input signal from an input line La, and the vertical low frequency component is separated while using a low-pass filter 11 of 100% roll-off characteristics including an input frequency band. On the other hand, the input signal is inputted to a subtracter 13, the vertical low frequency component is subtracted and the vertical high frequency component is provided and inputted through a delay circuit 16 to a multiplier 17. The output of the multiplier 17 for converting a vertical envelope component, vertical window function Kv based on a first reference value and vertical high frequency component into a second vertical high frequency component is added to an adder 15 together with the vertical low frequency component and synthesized. An output Lb is provided by the synthesizing means 15 and passed through a horizontal signal reproducing and synthesizing means by a means similar to the vertical component, and an output Lc is provided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is a device for reducing the ringing and noise of a video signal of a television or the like and improving the image quality. It can be applied to equipment.

[0002]

2. Description of the Related Art An example of the configuration of a conventional ringing reduction device will be described with reference to FIGS. 2 (a) and 2 (b). Here, a luminance signal of the NTSC system is taken up as an example of a signal to be subjected to ringing reduction. Although it can be applied not only to the luminance signal but also to a color difference signal or the like, processing for the luminance signal will be taken up as an example for explaining the operation. This luminance signal is a signal having a frequency band fm of about 4 MHz as shown in FIG. 4 (a). (See the following formula)

[0003]

[Equation 1]

The signal processing is performed by a digital circuit whose sampling frequency is fs in the following equation.

[Equation 2]

First, a first conventional configuration example, FIG. 2A, will be described. This device comprises a low pass filter 21 and a subtractor 2
It consists of four blocks: 2, adder 23 and non-linear converter 24. The input signal from the line La is as shown in FIG.
It is an impulse signal with flat characteristics up to the frequency fm as shown in (a). This signal is a signal with a lot of ringing in which the spectrum sharply becomes zero at the band upper limit frequency. FIG. 7 is a characteristic diagram of an input signal. 7 (a) is a time waveform diagram of amplitude linear display, FIG. 7 (b) is a time waveform diagram of amplitude dB display, FIG. 7 (c) is a spectrum diagram of amplitude linear display, and FIG. 7 (d) is amplitude d display.
It is a spectrum diagram of B display. 8, 9 and 14 to 22 which will be cited later are also drawn in the same display method. This input signal is applied to the low pass filter 21 and the subtractor 22. The characteristic of the low-pass filter 21 is 4 as shown in Go (f) in the following equation.
It has a 100% roll-off characteristic that attenuates like a cosine wave up to MHz.

[0006]

(Equation 3)

The input signal and the signal from the low-pass filter 21 are supplied to the subtractor 22 in the next stage, and the high-frequency component is obtained by subtraction. This high frequency component is added to the non-linear converter 24 in the next stage. The input / output characteristic of the non-linear converter 24 is a non-linear characteristic as shown in the following equation in which an output signal is difficult to be output when the input signal level is low as shown in FIG. 4 (c).

[0008]

[Equation 4]

In this circuit, a component such as ringing having a small amplitude in the high frequency region is compressed by the fourth power of the ratio with the reference value po. The output of the non-linear converter 24 is the adder 2 of the next stage.
In 3, the low frequency component is added and the output signal as shown in FIG. 8 is obtained via the line Lc. Figures 8 (a) and (b)
As far as the time waveform diagram of is seen, ringing surely decreases, and it seems to be good at first glance, but looking at the linear display in Fig. 8 (c) that converts this to the frequency domain, the frequency band within the signal band below fm The spectrum waviness is large, and it can be seen from Fig. 8 (d) in dB that the meaningless spectral components added to the outside of the signal band where the frequency is fm or higher are added in a considerable amount. Since the spectrum outside this band has weak correlation with the signal, it operates in the same manner as noise, and in general, a low-pass filter is often attached to remove it. The low-pass filter for removing the out-of-band noise needs to have a characteristic that the gain is flat up to the frequency fm and the gain sharply drops at frequencies higher than that in order to prevent the signal component from being affected.

However, the filter for obtaining such characteristics is a complicated circuit, and the cost is considerably high. Even if unnecessary components outside the signal band are removed by such an expensive low-pass filter, it is not possible to remove the spectral disturbance within the signal band, and this remains as signal distortion. . Since such signal distortion cannot be removed by post-processing in general, the characteristic of the nonlinear converter shown in FIG. 4 (c) was devised, but the expected characteristic was not obtained. Next, a second conventional configuration example shown in FIG. 2 (b) will be described. This is a low pass filter 25 for the input signal.
It is based on the idea that the high-frequency component of the signal can be dropped by multiplying by, and the ringing component having the main energy in the high-frequency component can be suppressed. However, when an input signal having the characteristic shown in FIG. 7 is added and the device is actually operated, an output signal having the characteristic shown in FIG. 9 is obtained. 9 (a) and (b)
Comparing the time waveform diagram of Fig. 7 (a) and 7 (b), it is difficult to say that the level of ringing is reduced due to the decrease in the high-frequency component, but the feeling of sickness spread over the entire waveform is reduced. I understand.

[0011]

The ringing component which appears as a striped pattern on the edge of a steep signal is a factor of impairing the image quality, and the coding distortion which is often conspicuous in the area where the background is flat due to the encoding and decoding is the image quality. Although it is a factor of deterioration, there is also a cause of this that ringing accompanying the sharpening of the spectrum is also a cause. On the other hand, various adverse effects caused by performing the ringing reduction processing as in the conventional example, that is, adverse effects such as distortion within the signal band, addition of noise-like harmonics outside the signal band, and deterioration of the signal high frequency component In order to obtain a comprehensive image quality improvement effect, the amount of noise that has the same properties as ringing is reduced by reducing the amount of natural ringing with less distortion while keeping the resolution as low as possible. Is what kind of measures should be taken.

[0012]

The above-mentioned problems have been solved by constructing a noise reduction device by the following means. (1) In an image quality improving device that performs ringing and noise reduction processing on a video signal for reproducing a horizontal and vertical two-dimensional image, in the vertical direction, the input signal is divided into a vertical low frequency component and a vertical high frequency component. A signal separating means, a vertical envelope extracting means for obtaining an envelope component of the vertical high-frequency component, and a vertical window function which forms a vertical window function based on the vertical envelope component and a predetermined first reference value. Window function forming means, vertical high-frequency conversion means for converting the vertical high-frequency component to a second vertical high-frequency component using the vertical window function, and the second vertical high-frequency component for the second vertical high-frequency component. Vertical signal recombining means for forming a vertical recombined signal by adding and combining the vertical low-frequency components,
With respect to the horizontal direction, a horizontal signal separating means for dividing the vertical re-synthesis signal into a horizontal low-frequency component and a horizontal high-frequency component, a horizontal envelope extracting means for obtaining an envelope component of the horizontal high-frequency component, and Horizontal window function forming means for forming a window function in the horizontal direction based on the horizontal envelope component and a predetermined second reference value, and the horizontal high-frequency component is changed to a second horizontal height by using the horizontal window function. Image quality improvement by a horizontal high-frequency converting means for converting into a high-frequency component and a horizontal signal re-synthesizing means for adding and synthesizing the second horizontal high-frequency component with the horizontal low-frequency component to form a horizontal re-synthesis signal. Configured to do.

(2) In (1), the vertical signal separating means separates the vertical low-pass components by using a low-pass filter having a 100% roll-off characteristic that covers the frequency band of the input signal, and then separates the above-mentioned vertical low-pass components. A vertical high frequency component is obtained by subtracting the vertical low frequency component from the input signal, and in the horizontal signal separating means,
First, a horizontal low-pass component is separated by using a low-pass filter having a 100% roll-off characteristic that covers the frequency band of the input vertical re-synthesis signal. A horizontal high frequency component is obtained by subtracting the high frequency component.

(3) In (1) or (2), the vertical envelope is calculated from the vertical standard deviation value obtained from the square root of the root mean square of the difference between the vertical low-frequency component and the values of a plurality of sampling points of the input signal. A vertical envelope extracting means for obtaining a line component,
And a horizontal envelope extracting means for obtaining a horizontal envelope component from a horizontal standard deviation value obtained from the square root of the square sum average of the horizontal high frequency components.

(4) In (1) or (2), vertical envelope extraction means for determining a vertical envelope component by band limiting the horizontal standard low pass filter with respect to the vertical standard deviation value. The horizontal standard deviation value is further limited by a horizontal low-pass filter, and horizontal envelope extracting means for determining a horizontal envelope component is provided.

(5) In any one of (1) to (4), a phase angle is obtained from the vertical or horizontal envelope component and a predetermined reference value, and the range of the phase angle is 0 to π.
/ 2, and this is applied to a trigonometric function such as a sine function or a cosine function, and the trigonometric function is multiplied by a predetermined power to obtain a window function in the range of 0 to 1 Vertical or horizontal window function formation Equipped with means.

(6) In (5), the phase angle is the sum of the vertical or horizontal envelope component and a predetermined reference value with respect to the vertical or horizontal envelope component or the reference value. It is configured to be obtained from the ratio.

(7) In an image quality improving device that performs a ringing reduction process and a noise reduction process on a video signal by three-dimensional processing of horizontal and vertical times, in the time direction, the input signal is time low frequency component and time high frequency component. A time signal separating means for dividing the time high-frequency component, a time envelope extracting means for obtaining an envelope component of the time high frequency component, the time envelope component and a predetermined first
Time window function forming means for forming a window function in the time direction based on the reference value of, and a time high frequency component for converting the time high frequency component into a second time high frequency component using the time window function. Conversion means, time signal recombining means for adding and synthesizing the second time high-frequency component to the time low-frequency component to form a time recombined signal, and the time recombined signal in the vertical direction. A vertical signal separating means for dividing a vertical low-frequency component and a vertical high-frequency component, a vertical envelope extracting means for obtaining an envelope component of the vertical high-frequency component, the vertical envelope component and a predetermined second criterion. Vertical window function forming means for forming a window function in the vertical direction based on the value, and vertical high frequency component conversion means for converting the vertical high frequency component into a second vertical high frequency component using the vertical window function. And adding the second vertical high frequency component to the vertical low frequency component A vertical signal recombining means for forming a vertical recombined signal, a horizontal signal separating means for dividing the vertical recombined signal into a horizontal low-frequency component and a horizontal high-frequency component, and the horizontal high-frequency component. Horizontal envelope extracting means for obtaining an envelope component of the area component; horizontal window function forming means for forming a horizontal window function based on the horizontal envelope component and a predetermined third reference value; A horizontal high-frequency component converting means for converting the horizontal high-frequency component into a second horizontal high-frequency component using a window function, and the second horizontal high-frequency component is added and synthesized with the horizontal low-frequency component. And a horizontal signal recombining means for forming a horizontal recombined signal to improve the image quality.

(8) In (7), the time signal separating means first separates the time low-pass components using a time low-pass filter having a 100% roll-off characteristic that covers the frequency band of the input signal, and thereafter, The time high-frequency component is obtained by subtracting the time low-frequency component from the input signal, and the vertical signal separation means first uses a 100% roll-off characteristic including the frequency band of the time re-synthesis signal to be input. The vertical low-pass component is separated by using the vertical low-pass filter of, and then the vertical high-pass component is obtained by subtracting the vertical low-pass component from the time recombined signal, and in the horizontal signal separating means, 100% including the frequency band of the vertical re-synthesis signal input first
A horizontal low-pass filter having a roll-off characteristic is used to separate the horizontal low-pass component, and thereafter the horizontal low-pass component is subtracted from the vertical re-synthesis signal to obtain a horizontal high-pass component. .

(9) In (7) or (8), the time envelopment is performed from the time standard deviation value obtained from the square root of the root mean square of the difference between the values of the sampling points of the input signal with respect to the time low frequency component. A vertical standard deviation value obtained from the square root of the square sum average of the difference between the time envelope extracting means for obtaining the line component and the values of the plurality of sampling points of the input time recombined signal for the vertical low frequency component. A vertical envelope extracting means for obtaining a vertical envelope component from the horizontal envelope extracting means for obtaining a horizontal envelope component from a horizontal standard deviation value obtained from the square root of the square sum average of the horizontal high-frequency components and the like. .

(10) In (7) or (8), a time envelope extracting means for defining a time envelope component by further band limiting the horizontal standard low pass filter with respect to the time standard deviation value. Vertical envelope extraction means for determining a vertical envelope component by band-limiting the vertical standard deviation value with a horizontal low-pass filter, and horizontal low-pass filtering for the horizontal standard deviation value. And a horizontal envelope extracting means for determining a horizontal envelope component by performing band limitation with a container.

(11) In any one of (7) to (10), a phase angle is obtained from the time, vertical and horizontal envelope components and a predetermined reference value, and the range of the phase angle is set to 0. ~ Π
/ 2, and this is applied to a trigonometric function such as a sine function or a cosine function.
A time window vertical and horizontal window function forming means for determining a window function in the range of 1 is provided.

(12) In (11), the phase angle is the time, vertical or horizontal envelope component with respect to the sum of the time, vertical or horizontal envelope component and a predetermined reference value, or It is configured to be obtained from the ratio with the reference value.

[0024]

1 and 3 show a first embodiment of an image quality improving apparatus of the present invention.
It is an example of. 4 to 26 are operation explanatory diagrams. In the waveform diagrams for explaining the operation, the time delay due to the processing time in the block or the circuit is omitted, and the notation according to the reference position of the presented waveform is used. For example, when handling a pulse signal, the peak position is used as a reference (time t =
0), and the waveform after processing in each block or circuit is also drawn on the same time base. In the present invention, horizontal and vertical 2
An image signal having a three-dimensional arrangement is dealt with, and FIG. 5A shows an arrangement diagram of sample points of the image signal arranged in a square lattice. x00 (□
Mark) as the central pixel and 2N + 1 from -N to N in the horizontal direction
The number () indicates a pixel array of three () from m-1 to m + 1 in the vertical direction. In the vertical direction, since a field image will be handled, three scan line data will be handled. The range of pixels handled by the functional blocks quoted in the operation description is a maximum of 3 pixels in the vertical direction and a maximum of 15 pixels in the horizontal direction.

FIG. 1 shows an embodiment of the image quality improving device according to the present invention, and FIG. 3 shows a concrete example of the configuration of the main blocks of FIG. FIG. 1 can be functionally divided into two large block groups. Blocks 11 to 1-a are blocks that perform image quality improvement processing in the vertical direction of the input image signal, and 1-b to 1-k are blocks that perform image quality improvement processing in the horizontal direction. First,
The configuration of the image quality improvement block group in the vertical direction will be described. Block 11 is a first low pass filter, 13 is a first subtractor, 1
5 is a first adder, 17 is a first multiplier, 18 is a first standard deviation circuit, 19 is a second low-pass filter, 1-a is the first
In the non-linear converter, 12, 14, 16 are delay circuits for processing time correction in the circuit system. Next, the configuration of the image quality improvement block group in the horizontal direction will be described. 1-b is a third low-pass filter, 1-d is a second subtractor, 1-f is a second adder, 1
-H is a second multiplier, 1-i is a second standard deviation circuit, 1
-J is a fourth low-pass filter, 1-k is a second nonlinear converter, and 1-c, 1-e, and 1-g are delay circuits for processing time correction.

As an input signal from the input line La of FIG. 1, a luminance signal having a signal band fm (approximately 4 MHz) as shown in FIG. 4A is handled. As an example of the waveform, FIG.
An impulse waveform such as, that is, sinx / x pulse waveform is taken up. Figure 7 (a) is a waveform diagram of amplitude linear display,
FIG. 7B is a waveform diagram in amplitude dB display, FIG. 7C is a spectrum diagram in amplitude linear display, and FIG. 7D is a spectrum diagram in amplitude dB display. 7 (a) is a linear time-varying waveform diagram, while FIG. 7 (b) is a temporal waveform diagram for viewing a minute ringing component in an easy-to-understand range, and FIG. 7 (c) is within the signal frequency band. FIG. 7D is a spectrum diagram for observing the degree of smoothness of the frequency characteristic, and FIG. 7D is a spectrum diagram for making it easy to see the state outside the signal band.

This sinx / x pulse waveform is input in a two-dimensional format as shown in FIG. 5 (b). In FIG. 5 (a), the horizontal axis represents time display in the horizontal direction, and the vertical axis represents scanning line display in the vertical direction. Although the waveforms of the scanning lines from the scanning line number m-3 to m + 4 are drawn, it shows that there is a sinx / x pulse in which the ringing spreads horizontally in the scanning line numbers m-3 to m. It is the figure which cut out a part of image which spreads two-dimensionally, and was represented by the signal waveform. Waveform diagrams quoted in the following description of the operation of each block, Fig. 5 (c), Fig. 10 to Fig. 13, Fig.
The same two-dimensional representation method is used in FIGS. 24 and 25.
Fig. 10 shows the waveform diagram of the schematic diagram of Fig. 5 (b) drawn with higher accuracy.
FIG. 6A is a waveform diagram of two-dimensional display in the range of ± 5 μs in the horizontal direction, which is (a). This input signal is supplied to the first low-pass filter 11 and the first delay circuit 12. First low pass filter 11
The characteristic G1 (f) is 100 up to 525/4 cph in the band as shown in Equation 5.
% Low-pass filter with roll-off characteristic.

[0028]

(Equation 5)

This is 0 at the upper limit signal frequency fmax of the video signal band as shown by the thick solid line in FIG. 4 (e) when viewed in interlaced scanning of the television, that is, in field units.
The characteristics are as follows. When one image is formed on a frame-by-frame basis, the vertical direction of the field image is usually not band-limited. The output waveform diagram of the first low-pass filter is shown in FIG. 10 (c), which shows a waveform in which the high-frequency component of the vertical waveform change portion is suppressed. In terms of data, changes are seen in the waveforms of scanning line numbers m and m + 1. FIG.
(a) is a circuit configuration example of the first low-pass filter 11, which is a digital transversal filter including a delay circuit group, an amplifier group (weighting circuit group), and a combiner. La-a1 is an input line and La-a2 is an output line. Each of the blocks aa1 and aa2 is a group of delay circuits connected in series with a delay time of a horizontal scanning period Th (the following equation).

[0030]

(Equation 6)

This delay circuit group outputs three signals at Th intervals. This delayed signal group is the next-stage amplifier group a-b.
1, ab2 and ab3 are supplied to the signals input by this amplifier group, but with predetermined magnifications (1/4, 1/2, 1/4) respectively.
Are weighted. The three signals output from this amplifier group are added and combined by the next combiner a-c1 and output. If the replacement of fmax = fh / 2 (fmax = upper band upper limit frequency) is performed in Equation 6, the filter characteristic of the spatial frequency display, that is, Equation 5, that is, the characteristic shown by the bold solid line in FIG. 4 (e), can be obtained. The output of the first low-pass filter 11 is sent to the first subtractor 13 and the second delay circuit 14 in the next stage.

The input signal from the line La is processed by the first delay circuit 12 at the processing time T in the first low pass filter 11.
It is given a time delay of h and sent to the subtractor 13 in the next stage. The first subtractor 13 is a circuit that finds the difference between the signal from the first delay circuit 12 and the signal from the first low-pass filter 11. The output waveform diagram is shown in Fig. 10 (b), which is the vertical high-frequency component of the input signal. Scan line numbers m and m +
A change in the waveform can be seen at position 1. The output of this subtractor is added to the third delay circuit 16 in the next stage. The first standard deviation circuit 18 obtains the square root of the sum of squared averages of the differences between a plurality of (three) sample points before and after each horizontal scanning period added from the block 12 and the low-frequency component added from the block 11. Although it is a circuit, this is a calculation like the following equation using the values of three pixels in the vertical direction spatially.

[0033]

(Equation 7)

A concrete configuration example of the standard deviation circuit 18 is shown in FIG. 3 (b). In the figure, lines Lb-a1 and Lb-a2
Is an input terminal, and the line Lb-a3 is an output terminal. ba
Reference numerals 1 and b-a2 are delay circuits each having a delay time Th (horizontal scanning period), which are connected in series and three input signals having a time interval Th after being time-corrected by the delay circuit 12 in the preceding stage. The sample value of can be taken out. The three sample values from this delay circuit group are the following subtractor groups b-b1 to b-
It is added to b3. These blocks have line Lb-a2
The signal sent from the low-pass filter 11 is commonly added via. Here, the difference between the three signals which are input from the block 12 and are temporally preceding and following, and the signal from the low-pass filter 11 is obtained. This subtractor group b-b1 to b-
The three signals from b3 are respectively the following multiplier groups b-c1 to b-c1.
b-c3.

This multiplier group is a circuit for obtaining the square value of input signals, and is a circuit for multiplying input signals. The three signals obtained from this group of multipliers are
The following combiner b-d1 performs addition combining. The square roots of the signals obtained by the addition and synthesis are calculated by the next converter b-e2 and output from the line Lb-a3. This converter is a table lookup type circuit that calculates the relationship between the input value and the square root in advance, writes the data in a ROM or the like in advance, and reads and outputs the data when necessary. Can be achieved with. Output waveform diagram σ of the first standard deviation circuit 18
v is shown in FIG. 10 (d). This is calculated as N = 1 in the equation (7). Due to the function of the standard deviation circuit, the envelope component of the vertical high frequency component that is input is extracted. Since the square root of the sum of squares from adjacent sample points is obtained, the function of the low-pass filter is also included, and the harmonic components in the vertical direction are also suppressed. The first standard deviation value σv is applied to the second low pass filter 19 in the next stage. This low-pass filter 19
The characteristic G2 (f) is a low-pass filter having the same configuration as the block 11 represented by the following equation and a 100% roll-off characteristic in the frequency band 7.16 MHz (= fs / 2).

[0036]

(Equation 8)

FIG. 3C shows an example of the circuit configuration of the second low-pass filter 19, which is similar to the first low-pass filter 11 and includes a delay circuit group, an amplifier group (weighting circuit group), and It is a transversal filter having a digital configuration including a synthesizer. Lc-a1 is an input line and Lc-a2 is an output line.
Each of c-a1 and c-a2 is a group of delay circuits connected in series with a delay time of sampling interval Ts (Equation 2). This delay circuit group outputs three signals at Ts intervals. This delayed signal group is supplied to the amplifier groups c-b1 to c-b3 at the next stage, and the signals inputted by this amplifier group are respectively given predetermined magnifications (1/4, 1/2, 1/4). ) Is weighted. The three signals output from this amplifier group are added and combined by the next combiner c-c1 and output. An output waveform diagram of the second low pass filter 19 is shown in FIG. 11 (e). An unnecessary horizontal high frequency component remaining in the input first standard deviation value is removed to form a smooth waveform. The banding reduction range can be narrowed down by strongly applying band limitation. The output of this second low pass filter 19 is applied to the next first non-linear converter 1-a. An example of the characteristic of the first non-linear converter 1-a is shown in the following equation.

[0038]

[Equation 9]

Here, the reason why x is the absolute value of the input signal is that the first low-pass filter 11 in the preceding stage may act to input a small negative value. In the first term of this equation, the ratio of x to x + xo, that is, x / (x + xo) is used. This is seen in the waveform example of x and x + xo shown in Fig. 4 (d). And at time t = tb, which is the peak of the waveform, x>
Since> xo, the value of this ratio is close to 1, and at t = ta, which is a little away from the peak, the value of the ratio is in the range of 1 to 1/2, and t = tc, which is further away from the peak. At x
= Xo, the ratio value becomes 1/2, and when x is sufficiently small, the ratio value should be close to 0. Since a value between 0 and π / 2 can be obtained by multiplying the value of this ratio by π / 2, a sine function having this as a phase angle can be used.

It is also possible to use this sine function and the cosine function of 90 ° phase difference. At this time the phase angle is
As shown in the section, it is obtained from the ratio of xo / (x + xo). The r-th power of a trigonometric function such as a sine function or a cosine function based on these phase angles and a coefficient Ko for slight fine adjustment
The window function is formed from and. xo = 0.01, Ko = 1 and r =
FIG. 6 shows an example of calculation of the window function, that is, the converter output K when 1, 2, 4, 8, and 16 are set. In Fig. 6 (a), the horizontal axis x = 0 to 0.02
6B is an example of calculation of K for the horizontal axis x = 0 to 0.2. The characteristics of the window function are the parameters xo, r, Ko
The characteristics can be changed depending on the method of giving, but the standard value used to obtain the operation waveform diagram is the parameter of the following equation.

[0041]

[Equation 10]

Such a non-linear characteristic can be constituted by a table lookup type ROM or the like. FIG. 11 (f) shows an output waveform diagram obtained based on the parameters of the above equation. It goes without saying that this waveform is a case where the input signal is a sinx / x waveform, and when another waveform is an input signal, it becomes a different waveform corresponding to it. The value K of the equation 10 determined in this way is used as a window function for ringing and noise reduction. A window function in the time domain is effective for the purpose of reducing ringing, that is, narrowing the existing range of ringing. The present invention is characterized by this window function forming method. Regarding the vertical window function, the noise reduction function appears rather than the ringing reduction function on the front surface, as will be described later. This is because when the television video signal of the interlace scanning system is handled, band limitation is not normally performed in the vertical direction of the field image. However, when there is coding distortion due to compression / expansion or the like, ringing equivalent to a sharp band limitation may occur in the vertical direction as well, so that the present invention functions effectively in such a case. Become.

Second delay circuit 14 and third delay circuit 1
Reference numeral 6 is a delay circuit for correcting the processing time in the window function forming circuit system (blocks 18 to 1-a). The delay time is Th + Ts, and each delays the input signal and outputs it. The output of the first non-linear converter 1-a is added to the next first multiplier 17 and serves as a window function for obtaining the product of the signal from the third delay circuit 16. This multiplier 17
Figure 11 (g) shows the output waveform diagram of the. Compared to Fig. 10 (b) of the signal before applying the window function, the waveform has horizontal ringing suppressed. Next first adder 15
Then, the low frequency component from the delay circuit 14 and the second vertical high frequency component from the multiplier 17 are added. The result of the addition synthesis is output via the line Lb. This output waveform diagram is shown in Figure 11.
It shows in (h). The processing of the lines La to Lb is the image quality improving processing in the vertical direction. Compared with the input signal of FIG. 10 (a), the high frequency component of the waveform of the scanning line number m is slightly dropped, and the waveform of the scanning line number m + 1 is the high frequency component of its rebound. The output from line Lb is the third low pass filter 1-
b and the fourth delay circuit 1-c. The characteristic G3 (f) of this third low-pass filter is that the pass band as shown in Eq.
It is a low-pass filter with 100% roll-off characteristics up to z.

[0044]

[Equation 11]

This is a characteristic which becomes 0 at the upper limit signal frequency fm of the video signal band. The output waveform diagram of the third low-pass filter is shown in FIG. 12 (j). It is a waveform in which the high-frequency component in the horizontal direction is suppressed and the ringing component is largely suppressed. FIG. 3 (e) shows an example of the circuit configuration of the third low pass filter. It is a transversal filter having a digital configuration including a delay circuit group, an amplifier group, and a combiner. Le-a1 is an input line and Le-a2 is an output line. Blocks e-a1 to e
Reference numeral -a14 is a group of delay circuits connected in series, each having a delay time of sampling period Ts (equation 2).

This delay circuit group outputs 15 signals at Ts intervals. This delay signal group is the amplifier group e- of the next stage.
The signals input to the amplifier groups b1 to e-b15 are weighted by a predetermined magnification to obtain the characteristic of the equation (11) in each amplifier. The 15 signals output from this amplifier group are added and combined by the next combiner e-c1 and output from the line Le-a2. In the number 11,
If the replacement of fmax = fm is performed, the characteristics shown by the thick solid line in FIG. 4 (e) can be obtained. The output of this third low pass filter is
It is sent to the second subtractor 1-d and the fifth delay circuit 1-e in the next stage. The second subtractor 1-d includes a fourth delay circuit 1-c.
And a signal from the third low-pass filter 1-b. Figure 12 (i) shows the output waveform diagram, which is the horizontal high frequency component of the input signal. The output of the second subtractor is the second standard deviation circuit 1-i of the next stage.
And the sixth delay circuit 1-g. The second standard deviation circuit 1-i is a circuit for obtaining the square root σh of the square sum average of the horizontal high-frequency components from the second subtractor 1-d. The following calculation using the pixel value is performed.

[0047]

(Equation 12)

A concrete configuration example of the second standard deviation circuit is shown in FIG. 3 (d). In the figure, line Ld-a1 is an input terminal and Ld-a2 is an output terminal. The block d-a1 is a squaring circuit that squares the input signal. b-b1 to b-
b4 is a circuit in which delay circuits each having a delay time Ts (sampling period) are connected in series, and sampled values of five signals at time Ts intervals input from the squaring circuit d-a1 can be taken out. There is. The five signals from this delay circuit group are added and combined by the next combiner d-c1. The square roots of the signals obtained by the addition and synthesis are obtained by the next converter d-d1 and output from the line Ld-a2. This converter calculates in advance the relationship between the input value and the square root, and
It can be realized by a circuit of a table lookup system in which the data is written in advance and read out and output when necessary.

The output waveform diagram of the second standard deviation circuit is shown in FIG.
As shown in 12 (k), this is obtained from five sample points with N = 2 in the equation 12. Due to the function of this standard deviation circuit, the envelope component of the horizontal high frequency component that is input is extracted. Since the square root of the sum of squares from adjacent sample points is obtained, the function of the low-pass filter is also included and the harmonic components are also suppressed. This second standard deviation value σ
h is added to the low pass filter 1-j of the next stage. The characteristic G4 (f) of this low-pass filter is 100% of the frequency band 7.16MHz (= fs / 2) of the same function as the block 11 as expressed by the following equation.
% Low-pass filter with roll-off characteristic.

[0050]

(Equation 13)

An output waveform diagram of the fourth low-pass filter 1-j is shown in FIG. 13 (l). A gentle waveform is obtained by removing unnecessary horizontal high-frequency components remaining in the second standard deviation value that is input. The ringing reduction range can be narrowed down by strongly applying the band limitation. A circuit example of the fourth low-pass filter 1-j is shown in FIG. The line Lf-a1 is an input terminal, and the line Lf-a2 is an output terminal. Line Lf-a1
The signal input from the delay circuit group f connected in series is
-A1 to f-a4. The delay time of the delay circuit is set to the sampling interval Ts. The outputs of the five delay circuit groups at time Ts intervals are added and averaged by the next combiner f-b1 and output from the line Lf-a2. Thus, the fourth low pass filter is an averaging circuit with five sample values.

The output of the fourth low pass filter 1-j is applied to the second non-linear converter 1-k. The characteristics of the second non-linear converter are the same as those of the first non-linear converter, including the parameters, but it is not an absolute requirement that they have the same characteristics. Figure 13 (m) shows the output waveform of the second non-linear converter. Since there is little ringing in the vertical direction, narrowing down also works strongly, and there is much ringing in the horizontal direction, so there is a window function that protects the vicinity of the pulse peak. You can see that it is formed. As for the horizontal direction, the noise reduction function that behaves similarly to this ringing function plays a major role in addition to the ringing reduction function, and the noise reduction function mainly works for signals with little ringing. . The fifth delay circuit 1-e and the sixth delay circuit 1-g are both window function forming circuit systems (blocks 1-i ...
1-k) is a delay circuit for correcting the processing time. The delay time is (2 + 2) Ts, and the input signal is delayed and output.

The output of the second non-linear converter 1-k is added to the next second multiplier 1-h, and the sixth delay circuit 1-
It is a window function for obtaining the product of the signal from g. An output waveform diagram of the second multiplier 1-h is shown in FIG. 13 (n). In the next second adder 1-f, the horizontal low-frequency component from the delay circuit 1-e and the second horizontal high-frequency component from the multiplier 1-h are added. The result of the addition synthesis is output via the line Lc. The waveform diagram of this final output signal is shown in Fig. 13 (o). Comparing this with the input signal waveform diagram 10 (a), the ringing of the waveforms from scanning line number m-3 to m is greatly reduced,
It can be seen that there is no unnecessary distortion that can be detected in the scan line numbers m + 1 to m + 4.

The operation of each functional block of the device of the present invention has been seen in a two-dimensional waveform diagram with reference to FIGS. 10 to 13. Next, the waveform of the two-dimensional waveform of FIG. The change of the waveform at the scanning line numbers m-1, m, and m + 1 which change to 1 will be analyzed and viewed. Scan line number m-3 in Figure 10 (a)
FIG. 7 shows a detailed characteristic diagram of the pulses existing up to m.
On the other hand, the scanning line number m- in the vertical processing output line Lb
Detailed characteristic diagrams of 1, m, and m + 1 are as shown in FIGS. 14 to 16, respectively, and the scanning line number m in the final processing output line Lc.
Detailed characteristic diagrams of -1, m, and m + 1 are as shown in FIGS. 17 to 19, respectively.

According to this, in the vertical processing output and the final processing output, the high frequency characteristics in the horizontal direction are slightly reduced at the scanning line number m, and the remnant waveform remains unbroken at the scanning line number m + 1. As far as the final output is seen, the signal of the scanning line number m + 1 also shows a sufficient ringing reduction effect to leave distortion below the detection limit. Also,
Comparing FIG. 17 which is the final processed output waveform diagram and spectrum diagram with FIG. 8 which is the output waveform diagram and spectrum diagram of the conventional example corresponding to this, and FIG. 9, there is almost no distortion in the signal band, and It can be seen that the component added outside the signal band also has a regular spectrum having a correlation with the signal. Since the spectrum outside this band is the harmonic component of the transient of the waveform, it is utilized as effective signal energy. If the width of the ringing is narrowed, the spectrum region of the signal is widened, and the out-of-band region, which is originally empty and has no utility, is effectively used.

Comparing the spectrum diagram of the input signal of FIG. 7 (c) and the spectrum diagram of the corresponding output signal of FIG. 17 (c), as shown schematically in FIG. 4 (f), the upper limit frequency fm of Instead of deleting the frequency component of the horizontal stripe on the left side, the frequency component of the vertical stripe on the right side of fm is formed and added. The component on the right side of fm is a component newly formed by non-linear processing, but it can be considered as an effect of band expansion. Due to the effect of expanding the band, the steepness of the waveform sloped portion is maintained even if the ringing is suppressed. The waveform quoted in the above description was a pulse waveform with a lot of ringing. Next, let's see how the waveform slope is affected. As an input signal for observing the state of this waveform slope part, Fig. 23 (a) (time waveform diagram),
(b) (enlarged time waveform diagram), (c) (spectrum diagram) bar waveforms (pulse width = approximately 36 μs, waveform with sharp edge with amplitude = 1) are used. The waveform diagram and spectrum diagram of the final output signal on the line Lc corresponding to this one-to-one are obtained as shown in FIGS. 26 (a), (b) and (c).

Comparing the input signal waveform FIG. 23 (a) and the enlarged view (b) with the output signal waveform FIG. 26 (a) and (b), the existence range of the ringing attached before and after the waveform slope is significantly increased. You can see that it is decreasing. Even if such a ringing reduction process is performed, it can be seen from the comparison between the time-enlarged waveforms in FIG. 23 (b) and FIG. 26 (b) that the waveform slope portion hardly changes. This is the feature of the present invention. This is as mentioned above
And as can be seen from the comparison of the spectrum diagrams in Fig. 23 (c) and Fig. 26 (c), the high-frequency spectrum is reduced by the nonlinear processing of the window function, and at the same time, the out-of-band component is formed and added, and the effect of band expansion is increased. It is thought to be due to. Figure 5 (c)
Corresponds to Fig. 5 (b) in which pulses are arranged two-dimensionally.
FIG. 6 is a waveform diagram in which a waveform having such a waveform sloping portion is arranged in scanning line numbers m-3 to m. FIG. 24 (a) is a two-dimensional waveform diagram drawn with accuracy from FIG. 5 (c). 24 (b) and 24 (c) are output waveform diagrams of the vertical processing output line Lb and the final output line Lc when FIG. 24 (a) is an input signal, respectively. In this waveform diagram, no band limitation is applied in the vertical direction, but the steepest change is given.However, in Figure 24 (b), the scan line number m +
Fig. 24 shows a slight break in the waveform that appears temporarily in 1.
It can be seen that the final output signal waveform of (c) returns to the original state and the ringing component generated due to the sharp band limitation up to the frequency fm in the horizontal direction is greatly reduced. FIG.
Is the window function Kv (Fig. 25)
(a)) and the window function Kh acting in the horizontal direction (Fig. 25 (b)
) Shows a waveform diagram. The waveform diagrams of these window functions provide an intuitive understanding of the edge-preserving mechanism of the present invention for narrowing ringing and reducing noise.

The device of the present invention has not only the ringing reduction processing but also a noise reduction function which is in some sense common to ringing. Figure 20 shows S / N for the input signal of Figure 7.
FIG. 7 is a waveform diagram and a spectrum diagram on the input line La when noise of about 20 dB is added. Figure 21 is the line for it
It is a waveform diagram and a spectrum diagram after vertical processing in Lb.
21 is a waveform diagram and a spectrum diagram of the output signal on the final output line Lc. The comparison between Fig. 20 and Fig. 21 shows that a noise reduction effect of several dB is obtained, and further the comparison with Fig. 22 shows that in addition to ringing, noise in the high band and out of the band is significantly reduced. I understand. The slightly remaining noise is a low-frequency noise component that is not processed in the present invention. In the comparison between FIG. 20 and FIG. 21, noise reduction of several dB is performed by the vertical processing, which is an effect obtained by the vertical correlation and is a positive effect of the two-dimensional processing. Although such a noise reduction effect can be realized by the conventional technology, it is possible to provide a ringing reduction function capable of removing an unnecessary minute high frequency component without giving an abnormal waveform distortion to a main signal component. This is the greatest feature of the present invention.
The characteristic (Equation 9) can be changed by changing some parameters in the nonlinear converter, which is a central function in the ringing reduction processing process. Increasing the coefficient Ko increases the gain of the converted high frequency component and emphasizes the high frequency component. On the contrary, if Ko value is decreased, high frequency components will decrease. When the value of the reference value xo is increased, the width of the window function is narrowed, and the narrowing range of ringing is narrowed. Conversely, as the value of xo becomes smaller, the ringing reduction effect becomes weaker.

When the value of r is increased, the ringing is narrowed down strongly, and the existence range of the ringing is narrowed. Conversely, if the value of r is reduced, the ringing reduction effect becomes weaker.
By using such characteristics, it is possible to obtain optimum characteristics according to the application. It is also possible to perform adaptive control by dynamically controlling these parameters. In the present invention, after performing the image quality improving process in the vertical direction,
Image quality improvement processing is being performed in the horizontal direction. This processing procedure is important, and if the reverse procedure is taken, there arises an adverse effect such that unnecessary harmonic components in the horizontal direction remain. Also, the first low-pass filter 11 and the third low-pass filter 1-b are set to 100% roll-off characteristics up to the upper limit frequency of the contained frequency band of the passing signal as described above. However, this is a characteristic selected because a good frequency characteristic with extremely small waviness of the spectrum in the signal band can be obtained even if the ringing reduction and the noise reduction processing by the window function are performed. Furthermore, in the description of the operation of the present invention, the luminance signal has been described as an example of an image (video) signal to be handled, but it can be applied to other video signals, that is, RGB signals, baseband signals such as color difference signals. Further, it can be applied to other than NTSC video signals. Further, it effectively works to reduce various coding distortions that occur during the compression / expansion process of image data.
It is also effective for pre- and post-processing such as PEG.

27 and 3 show the second embodiment of the image quality improving apparatus of the present invention. 4 and 28 to 52 are explanatory diagrams of the operation. In this embodiment, an image signal having a three-dimensional arrangement of horizontal and vertical times is handled, and FIG. 28 (a) shows an arrangement diagram of sample points of the image signal arranged in a cubic lattice. Sampling point number i in the horizontal direction with X000 (□) as the central pixel (processing target sampling point)
A total of 27 sample points at the intersections of the three coordinate axes of −1 to 1, vertical scanning line number j = −1 to 1 and time direction frame number k = −1 to 1 are shown. In the figure, three frames with frame numbers k = -1 to 1 are displayed in order to perform calculations in the frame interval Ts (= 1/30 sec). ● is a sample value of a frame number of k = −1, ■ and □ is k = 0, and ▲ is a frame number of k = 1. In the vertical direction, calculation is performed between horizontal scanning line intervals Ys (= 1 / fh, fh = 15.73426 KHz = horizontal scanning frequency), but in the figure, scanning line numbers j = -1 to 1
3 scanning lines are displayed. In the horizontal direction, since the sampling interval Xs (= 1 / (4fsc), fsc = 3.579545KHz = color subcarrier frequency) is calculated, the horizontal sampling point number i = −
Three sample points 1 to 1 are displayed.

FIG. 28 (b) shows an array of sample values when viewed in a horizontal and vertical two-dimensional plane. The calculation is performed in the field, but the calculation in the vertical direction in the even field indicated by ○ uses points A and C for the calculation target point B, and the calculation target point in the odd field indicated by ●. For E, points D and F are used. FIG. 28 (c) is an array of sample points when viewed in a two-dimensional plane perpendicular to time. The calculation is performed on a frame-by-frame basis, but in the time-direction calculation in the even field indicated by ◯, the points G and I are also used for the calculation target point H, and the calculation target point K is indicated in the odd field indicated by ●. Indicates that points J and L are used.

FIG. 29A shows the arrangement of all scanning lines when viewed in the vertical time plane. Circles and circles represent scanning lines of even fields, and squares and squares represent scanning lines of odd fields. In this
The test waveforms such as impulse and bar are superimposed on and, and the data on which the DC bias waveform which is the background reference level is superimposed are set in ○ and □, and they are used to investigate the effect of image quality improvement. ing. Since the calculation in the time direction is performed at the frame intervals in which the sampling points on the scanning line corresponding to the horizontal axis direction exist, see the effect of processing on one field, for example, even field, as shown in Fig. 29 (b). Thus, it is possible to determine the effect on all fields, that is, all scan lines. Therefore, in the following description, the waveform diagram quoted and described in the description of the effect of the processing uses the processing result of one field as shown in FIG. 29 (b).

In the following explanation of the operation, a method of displaying a three-dimensional image signal waveform by a combination of two-dimensional waveform diagrams is adopted. FIG. 30 shows six scan lines (number j
= -2 to 3) as a set, waveform diagrams for four frames in the time direction (number k = -1 to 2) (Figs. 30 (a) to (d))
Is a method of displaying. This waveform display method is used when displaying the processing results in the horizontal and vertical directions. FIG. 31 is a waveform diagram of four scanning lines (number j = −1 to 2) in the vertical direction (FIG. 31 (a) to (( d)) is displayed. This waveform display method is used when displaying the result of the time direction processing. Figure
30 and FIG. 31 are representation methods for understanding the processing effect in each direction. FIG. 27 is a configuration example of the image quality improving device according to the present invention. FIG. 3 is a specific configuration example of the main blocks of FIG.

FIG. 27 can be functionally divided into three large block groups. Blocks 1 to X-a are a first block group for performing image quality improvement processing in the time direction of the input image signal, and blocks 11 1-a are image quality improvement processing in the direction perpendicular to the image signal after the time direction processing. In the second block group, the blocks 1-b to 1-k are a third block group that performs the image quality improving process in the horizontal direction on the image signal after the process in the vertical direction. First, the configuration of the first block group that performs the image quality improving process in the time direction will be described. Block 1 is the first X
Low pass filter, 3 is a 1X subtractor, 5 is a 1X adder,
7 is the 1X multiplier, 8 is the 1X standard deviation circuit, and 9 is the 2X
, A first X non-linear converter, and blocks 2, 4 and 6 are delay circuits for correcting the processing time in the circuit system. Next, the configuration of the second block group for performing the image quality improving process in the vertical direction will be described. Block 11 is a first low pass filter, 13 is a first subtractor, 15 is a first adder,
17 is a first multiplier, 18 is a first standard deviation circuit, 19
Is a second low pass filter, 1-a is a first non-linear converter, and blocks 12, 14 and 16 are delay circuits for correcting the processing time in the circuit system.

Next, the configuration of the third block group for performing the image quality improving process in the horizontal direction will be described. Block 1-b is a third low pass filter, 1-d is a second subtractor, 1-f is a second adder, 1-h is a second multiplier, 1-i is a second multiplier. Standard deviation circuit, 1-j is a fourth low-pass filter, block 1-k is a second non-linear converter, and 1-c, 1-e and 1-g correct the processing time in the circuit system. It is a delay circuit. As an input signal from the input line La 'of FIG. 27, a luminance signal having a signal band up to fm (approximately 4 MHz) is treated as shown in FIG. 4 (a).
As an example of the signal waveform, an impulse waveform having the most ringing as shown in FIG. 48, that is, a sinx / x pulse waveform will be taken. Figure 48 (a) is a waveform diagram of amplitude linear display,
(b) is a waveform diagram in amplitude dB display, FIG. 48 (c) is a spectrum diagram in amplitude linear display, and FIG. 48 (d) is a spectrum diagram in amplitude dB display. The input signal is the impulse waveform (sin
(X / X pulse) is arranged in the middle of the horizontal scanning period, and FIG.
As shown in (b), it is placed as a test waveform on the upper left of the reference position (j = k = 0), that is, on the scanning line as in the following formula (scan line marked with ●), and the value on other scanning lines is 0 (○). It is a three-dimensional image signal which is placed as a background with the scanning line).

[0066]

[Equation 14]

This input signal is a set of six signal waveforms on the scanning line numbers j = -2 to 3, and the frame numbers k = -1 to
Waveform diagrams drawn for No. 2 are shown in FIGS. Figure 32
Is based on the waveform display method (1) in FIG. 30, but the horizontal axis represents time (sec) in the horizontal direction, the vertical axis represents the amplitude scale, and the scan line parameter represents the vertical direction. The waveform of a part of the scanning lines with the scanning line numbers j = -2 to 3 is drawn, but the scanning line numbers j = -2 to 0 and the frame number k≤
Sinx with horizontal ringing spread in the range of 0
/ X pulse is present. It is the figure which cut off one part of the image expanded three-dimensionally, and was represented by the signal waveform. Waveform diagrams and diagrams cited in the explanation of the operation of each block below
33 and FIG. 36 to FIG. 38 are similar expression methods. Note that Fig. 34 and Fig.
35 uses the waveform display method (2) in Fig. 31. This input signal is first supplied to the 1X low-pass filter 1 and the 1X delay circuit 2. The characteristic G1X (f) of this 1X low-pass filter is a 100% roll-off characteristic up to a band of 15 Hz as shown in the equation (15).

[0068]

(Equation 15)

Looking at the scanning lines G, H, I of FIG. 28 (c) relating to the calculation of this low-pass filter, the sampled values on each scanning line are weighted by 1/4, 1/2, 1/4. Is given, the output of the low-pass filter is obtained at the sampling point at the position of the scanning line H. This is as shown by the thick solid line in Fig. 4 (e) when the scanning line arrangement of the interlace scanning type television signal is viewed in one field.
The characteristic is such that it becomes 0 at the upper limit signal frequency fmax = 15 Hz of the video signal band. If you look at the image in one field,
Since information of up to half of the frame image can be expressed, it is usual to see that the band is not limited in the time direction. FIG. 3A shows an example of the circuit configuration of the 1X low-pass filter, which is a digital transversal filter including a delay circuit group, an amplifier group (weighting circuit group), and a combiner. La-a1 is an input line and La-a2 is an output line. The blocks aa1 and aa2 are a group of delay circuits connected in series with a delay time of a frame period Tf (the following equation).

[0070]

[Equation 16]

This delay circuit group outputs three signals at Tf intervals. This delayed signal group is the next amplifier group a-b1, a-
The signals supplied to b2 and a-b3 are weighted by a predetermined magnification (1/4, 1/2, 1/4) to the signals input by this amplifier group. The three signals output from this amplifier group are added and combined by the next combiner a-c1 and output. This 1X
The output of the low pass filter is sent to the following 1X subtractor 3, 2X delay circuit 4 and 1X standard deviation circuit 8. Further, the input signal from the line La 'is given a time delay of the processing time Tf in the 1X low pass filter by the 1X delay circuit 2 and sent to the next subtractor 3. The 1X-th subtractor 3 is a circuit for obtaining a difference between the signal from the 1X-th delay circuit 2 and the signal from the 1X-th low-pass filter 1. Here, the high frequency component of the input signal in the time direction is obtained. The output of this subtractor is applied to the next 3X delay circuit 6.

The 1 × X standard deviation circuit 8 is the square root of the sum of squared averages of the differences between a plurality of (three) sample points before and after each frame period added from the block 2 and the low frequency component added from the block 1. Although it is a circuit for obtaining σt, spatially, the operation is performed as in the following equation using the values of three pixels in the time direction, and the standard deviation value σt in the time direction is obtained.

[0073]

[Equation 17]

A concrete configuration example of this standard deviation circuit is shown in FIG.
Shown in (b). This standard deviation circuit has the same circuit configuration as that of the standard deviation circuit described above, and therefore its description is omitted.
Due to the function of this standard deviation circuit, the envelope component of the input high frequency component in the time direction is extracted. Since the square root of the sum of squares from adjacent sample points is obtained, the function of the low-pass filter is also included, and the harmonic component in the time direction is also suppressed. The 1X standard deviation value σt is applied to the 2X lowpass filter 9 in the next stage. The characteristic G2X (f) of this low pass filter is
It is a 100% roll-off characteristic of the band f2 = fs / 2 as expressed by the equation (18).

[0075]

(Equation 18)

FIG. 3C shows an example of the circuit configuration of this 2X low-pass filter. Similar to the 1X low-pass filter, a delay circuit group, an amplifier group (weighting circuit group), and a synthesizer are used. It is a transversal filter with a digital configuration consisting of. Lc-a1
Is an input line, and Lc-a2 is an output line. Block c
Reference symbols -a1 and c-a2 are delay circuit groups connected in series, each having a delay time of sampling interval Ts (Equation 2). This delay circuit group outputs three signals at Ts intervals. This delayed signal group is supplied to the next amplifier groups c-b1 to c-b3, and the signals input by this amplifier group are each given a predetermined magnification (1/4, 1/2, 1/4). Are weighted.

The three signals output from this amplifier group are added and combined by the next combiner c-c1 and output. Due to the action of the 2X low-pass filter, the input first standard deviation value becomes a smooth waveform in which the unnecessary unnecessary high-frequency component in the horizontal direction is removed. The ringing reduction range can be narrowed down by strongly applying the band limitation. The output of this 2X low-pass filter is the next 1X non-linear converter X
-Added to a. An example of the characteristic of this 1X non-linear converter is shown in the following equation (Equation 9).

Here, the reason why x is the absolute value of the input signal is that there is a possibility that a small negative value may be input due to the action of the 1X low-pass filter in the preceding stage. In the first term of this equation, the ratio of x to x + xo, that is, x / (x + xo) is used. This is shown in the schematic waveform example of x and x + xo shown in Fig. 4 (d). Looking at the time t = tb, which is the peak of the waveform,
Since x >> xo, this ratio value is close to 1, and at t = ta, which is a little away from the peak, the ratio value is 1 to 1/2.
The value of the ratio is within the range of, and at t = tc farther from the peak, the ratio value becomes 1/2 at x = xo, and when x is sufficiently small, the ratio value should be close to 0. Π to the value of this ratio
Since a value between 0 and π / 2 can be obtained by multiplying by / 2, a sine function having this as a phase angle can be used. It is also possible to use this sine function and the cosine function of 90 ° phase difference. The phase angle at this time is calculated from the ratio of xo / (x + xo) as shown in the second term of the equation 9. Sine function (sin) or cosine function (cos) based on these phase angles
R-th power of trigonometric function such as, and coefficient for slight fine adjustment
The window function is formed from Ko and.

When xo = 0.01 and Ko = 1, r = 1, 2, 4,
FIG. 6 shows an example of calculation of the window function, that is, the converter output K when 8 and 16. FIG. 6A shows K for the horizontal axis x = 0 to 0.02,
FIG. 6B is an example of calculation of K for the horizontal axis x = 0 to 0.2. The characteristic of the window function can be changed by the way of giving the parameters xo, r, Ko, but the standard value used to obtain the operation waveform diagram is the parameter of the following equation (Equation 10). Such non-linear characteristics can be configured by a table lookup type ROM or the like.

FIG. 33 is a waveform diagram of the window function Kt in the time direction obtained based on the parameters of the above equation. Figure 33 (a) ~
(d) has a one-to-one correspondence with the display methods of FIGS. 32 (a) to (d). It goes without saying that this waveform is obtained when a sinx / x waveform is used as the input signal, and when another waveform is an input signal, it becomes a different waveform corresponding to it.
The value Kt of the equation 9 determined in this way is used as a window function in the time direction for ringing and noise reduction. The present invention is characterized by this window function forming method. As for the window function in the time direction, the noise reduction function appears in front of the ringing reduction function, as will be described later. This is a case where an interlace scanning type TV video signal is handled, and in the case of an image signal for which calculation is performed between fields, only a half of the signal band of the image signal in frame units can be handled, so the band limitation in the time direction is performed. This is because it is normal that no mark is applied. However, when there is coding distortion due to compression / expansion or the like, ringing equivalent to a sharp band limitation may occur in the time direction as well, so that the present invention functions effectively in such a case.

Both the 4th 2X delay circuit and the 6th 3X delay circuit are delay circuits for correcting the processing time in the window function forming circuit system (blocks 8 to Xa). Delay time is Tf +
Ts, which delays the respective input signals and outputs the delayed signals. The output of the 1X non-linear converter, that is, the window function K
t is added to the next 1X multiplier 7 and the 3X delay circuit 6 is added.
The product with the high frequency component in the time direction from is obtained. As a result, a second high frequency component in the time direction is formed. The next
The 1X adder 5 adds the low-frequency component in the time direction from the delay circuit 4 and the second high-frequency component in the time direction from the multiplier 7. The result of the addition is output via the line Lb '.
This output waveform diagram is shown in FIG. Input signal waveform Figure 32 (a) ~
(d) is based on the waveform representation method (1) in Figure 30, while Figure 34
(a) to (d) are based on the waveform representation method (2) of FIG. 31, and the six scanning line signals of frame numbers k = -2 to 3 are set to 1
As a set, four waveform charts of scanning line numbers j = -1 to 2 (Fig.
34 (a) to FIG. 34 (d)). The input signal of FIG. 32 is included in the high frequency component in the time direction by the function of the time window function,
It is a waveform obtained as a result of suppressing ringing and noise components.

The process between the line La 'and the line Lb' is the image quality improving process in the time direction. Figure 32 (a) and (b)
34 (a) and 34 (b), the high frequency component of the waveform of frame number k = 0, which is the edge of the test waveform, is slightly dropped, and the waveform of scanning line number k = 1 is It bounces back into a waveform with high frequency components added. The output from the line Lb 'is the first low-pass filter 11 and the first delay circuit 12
Added to. The description of each block below is the same as that of the first embodiment already described in FIG. 1, and the description thereof will be simplified. The characteristic G1X (f) of this 1X low-pass filter is a 100% roll-off characteristic up to a pass band of 525/4 cph as shown in Eq.

Looking at the sample points A, B and C of FIG. 28 (b) relating to the calculation of this low pass filter, the value of each sample point is 1/4,
By giving weights of 1/2 and 1/4, the output of the low-pass filter can be obtained at the position of the sample point B. This is because when the pixel arrangement of the television signal of the interlaced scanning system is viewed in field units, it is 0 at the upper limit signal frequency fmax = 525/4 cph of the video signal band, as shown by the thick solid line in FIG. The characteristics are as follows. When an image is viewed in field units, the number of scanning lines is half the number of scanning lines in frame units, and it is common to see that band limitation is not applied in the vertical direction. FIG. 3A shows a circuit configuration example of the third low-pass filter, which is a digital transversal filter including a delay circuit group, an amplifier group (weighting circuit group), and a combiner. La-a1 is an input line and La-a2 is an output line. Blocks aa1 and aa2 are delay circuit groups connected in series, each having a delay time of horizontal scanning period Th (equation 6).

From this delay circuit group, three signals at Th intervals are output. This delayed signal group is the next amplifier group a-b.
1, ab2 and ab3 are supplied to the signals input by this amplifier group, but with predetermined magnifications (1/4, 1/2, 1/4) respectively.
Are weighted. 3 output from this amplifier group
These signals are added and combined by the next combiner a-c1 and output. The output of the first low pass filter is the following first subtractor 13, second delay circuit 14 and first standard deviation circuit 18
Sent to Further, the signal in the time direction from the line Lb ′ is given a time delay of the processing time Th in the first low pass filter by the first delay circuit 12 and sent to the next subtracter 13.

The first subtractor 13 is a circuit for obtaining the difference between the signal from the first delay circuit 12 and the signal from the first low pass filter 11. The high frequency component in the vertical direction of the signal input here is obtained. The output of this subtractor is applied to the third delay circuit 16. The first standard deviation circuit 18 obtains the square root σv of the sum of squared averages of the differences between a plurality of (three) sample points before and after the horizontal scanning period added from the block 12 and the low-frequency component from the block 11. Circuit,
This is spatially a calculation like the following equation 7 using the values of three pixels in the vertical direction. With this, the standard deviation value σv in the vertical direction is obtained.

A concrete configuration example of this standard deviation circuit is shown in FIG.
Shown in (b). This standard deviation circuit has the same circuit configuration as that of the standard deviation circuit described above, and therefore its description is omitted.
Due to the function of this standard deviation circuit, the envelope component of the vertical high-frequency component that is input is extracted and the square root of the sum of squares from adjacent multiple sampling points is obtained, so the function of the low-pass filter is also included. , The vertical harmonic components are also suppressed. This first standard deviation value .sigma.v is applied to the second low pass filter 19 which follows. Characteristics of this low pass filter G2 (f)
Is the same as the above-mentioned second X low-pass filter as expressed by
= 100% roll-off characteristic of fs / 2.

[0087]

[Formula 19]

FIG. 3C shows an example of the circuit configuration of the second low-pass filter, which is similar to the first low-pass filter, and includes delay circuit group, amplifier group (weighting circuit group) and combiner. It is a transversal filter with a digital configuration consisting of. Lc-a1
Is an input line, and Lc-a2 is an output line. Block c
Reference symbols -a1 and c-a2 are delay circuit groups connected in series, each having a delay time of sampling interval Ts (Equation 2). This delay circuit group outputs three signals at Ts intervals. This delayed signal group is supplied to the amplifier groups c-b1 to c-b3 of the next stage, and the signals input by this amplifier group are respectively given predetermined magnifications (1/4, 1/2, 1/4). ) Is weighted. The three signals output from this amplifier group are the following combiners c-c1.
Are combined and output.

By the action of the second low-pass filter, the input first standard deviation value becomes a gentle waveform in which the unnecessary unnecessary high-frequency component in the horizontal direction is removed. The ringing reduction range can be narrowed down by strongly applying the band limitation. The output of this second low pass filter is applied to the next first non-linear converter 1-a. As the characteristic of the first non-linear converter, the same number 10 as in the case of the block Xa is used. The parameters for obtaining the operation waveform diagram of the window function were the same 11 parameters as those for the time window function. FIG. 35 shows an output waveform diagram obtained based on the parameters of the above equation.
35 (a) to (d) are the same as FIG. 34, and the waveform display method of FIG. 8 (2)
It is drawn based on. The value Kv of the equation 9 determined in this way is used as a vertical window function for ringing and noise reduction.

This is because when the television video signal of the interlace scanning system is handled, it is common to see that band limitation is not performed in the vertical direction in the case of an image signal which is calculated in field units. However, when there is coding distortion due to compression / expansion or the like, ringing equivalent to a sharp band limitation may occur in the time direction as well, so that the present invention functions effectively in such a case. Block 14
The second delay circuit of 16 and the third delay circuit of 16 are both delay circuits for correcting the processing time in the window function forming circuit system (blocks 18 to 1-a), and the delay time is Th + Ts, and each is input. The delayed signal is output. The output of the first non-linear converter, that is, the window function Kv is
Of the third delay circuit 16 and the product with the high-frequency component in the vertical direction are obtained to form the second high-frequency component in the vertical direction.

In the next first adder 15, the vertical low-frequency component from the delay circuit 14 and the second vertical high-frequency component from the multiplier 17 are added and combined. The result of the addition synthesis is output via the line Lc '. This output waveform diagram is shown in FIG. 36 (a) to 36 (d) are based on the waveform display method (1) in FIG. 30, and the scanning line number k is a set of six scanning line signals of scanning line numbers j = -2 to 3 FIG. 36 is a waveform diagram of four sheets (−1 to 2) (FIG. 36 (a) to FIG. 36 (d)). 32 is a waveform obtained as a result of the ringing and noise components included in the high frequency component in the vertical direction being suppressed by the action of the vertical window function Kv of the input signal in FIG. 32.

The process between the lines Lb 'and Lc' is the image quality improving process in the vertical direction. Compared to the input signals in Fig. 32 (a) and (b), the scanning line number k in Fig. 36 (a) and (b)
The high frequency component of the waveform at the boundary portion of = 0 is slightly dropped, and the high frequency component of the rebound is added to the waveform of the scanning line number k = 1. The vertically processed signal from line Lc 'is applied to low pass filter 1-b and delay circuit 1-c. The characteristic G3 (f) of the third low-pass filter is a low-pass filter with a 100% roll-off characteristic up to a pass band of 4 MHz as shown in the following equation. It is the same as GO (f) (Equation 3).

[0093]

(Equation 20)

This is the upper limit signal frequency fm of the video signal band.
The characteristic is that it becomes 0 in (Equation 1). By the action of the third low-pass filter, the high-frequency component in the horizontal direction is suppressed, and the low-frequency component in the horizontal direction in which the ringing component and the noise component are largely suppressed is output. FIG. 3 (e) shows an example of the circuit configuration of the third low pass filter. The transversal filter has a digital structure including a delay circuit group, an amplifier group (weighting circuit group), and a combiner. Le-a1 is the input line, Le-
a2 is an output line. Blocks e-a1 to e-a14
Is a group of delay circuits connected in series, each having a delay time Ts (sampling period, see equation 2). The 15 signals at Ts intervals output from this delay circuit group are the next amplifier group eb
1 to e-b15, where weighting of a predetermined scale factor for obtaining the characteristic of the equation (20) is performed. The 15 signals output from this amplifier group are added and combined by the combiner e-c1 and output from the line Le-a2. If the replacement of fmax = fm is performed in the equation (20) (= the equation (3)), the characteristic of the thick solid line in FIG. 4 (e) can be obtained. The output of the third low-pass filter is sent to the next second subtractor 1-d and fifth delay circuit 1-e.

The second subtracter 1-d is provided in the fourth delay circuit 1
It is a circuit for obtaining the difference between the signal from -c and the signal from the third low-pass filter 1-b, and outputs the high-frequency component in the horizontal direction of the input signal. The output of this second subtractor is
The following second standard deviation circuit 1-i and sixth delay circuit 1-g
Is added to The second standard deviation circuit 1-i calculates the square root σ of the square sum average of the horizontal high-frequency components from the second subtractor 1-d.
Although this is a circuit for obtaining h, it is obtained spatially here by the calculation of the following equation 12 using five pixel values in the horizontal direction.
With this, the standard deviation value σh in the horizontal direction is obtained. A concrete configuration example of the second standard deviation circuit is shown in FIG.
This standard deviation circuit has the same circuit configuration as that of the standard deviation circuit described above, and a description thereof will be omitted.

The horizontal standard deviation value σh is obtained from the five sample points with N = 2 in the equation (12). By the function of this standard deviation circuit, the envelope component of the input high frequency component in the horizontal direction is extracted. Since the square root of the sum of squares from adjacent sample points is obtained, the function of the low-pass filter is also included and the harmonic components are also suppressed. The output of this second standard deviation circuit is applied to the next fourth low pass filter 1-j. Characteristics of this low pass filter G4
(f) is an averaging low-pass filter as expressed by the following equation 13.

By the action of the fourth low-pass filter, the input second standard deviation value becomes a gentle waveform in which the unnecessary unnecessary high-frequency component in the horizontal direction is removed. The ringing reduction range can be narrowed down by strongly applying the band limitation. A circuit example of the fourth low-pass filter is shown in FIG. The line Lf-a1 is an input terminal, and the line Lf-a2 is an output terminal. The signal input from the line Lf-a1 is added to the delay circuit groups f-a1 to f-a4 connected in series.
The delay time of the delay circuit is set to Ts (sampling interval). The outputs of the five delay circuit groups at Ts intervals are added and averaged by the next combiner f-b1 and output from the line Lf-a2. Thus, the sixth low pass filter is an averaging circuit with five sample values.

The output of this fourth low pass filter is
Of the non-linear converter 1-k. The characteristics of the second nonlinear converter are the same as those of the 1x and first nonlinear converters including the parameters. FIG. 37 shows the output waveform of the third nonlinear converter, that is, the window function Kh in the horizontal direction.
Figures 37 (a) to 37 (d) are drawn based on the waveform display method (1) in Fig. 30 and have a one-to-one correspondence with the waveform diagrams in Figs. 33 (a) to 33 (d). It corresponds. From FIG. 37, it can be understood that there is little ringing in the time direction and the vertical direction, so the narrowing down also works strongly, and since there is a lot of ringing in the horizontal direction, the window function is formed so as to protect the vicinity of the pulse peak. . As for the horizontal direction, the noise reduction function that behaves similarly to this ringing function plays a major role in addition to the ringing reduction function, and the noise reduction function mainly works for signals with little ringing. .

The fifth delay circuit 1-e and the sixth delay circuit 1-g
Of the delay function of the window function forming circuit system (1-i to 1-
It is a delay circuit for correcting the processing time in k). The delay time is (2 + 2) Ts, and each input signal is delayed and output. The output of this second non-linear converter, that is, the window function Kh, is added to the next second multiplier 1-h, and the product with the high frequency component in the horizontal direction from the sixth delay circuit 1-g is obtained. And a second horizontal high frequency component is formed. In the next second adder 1-f, the horizontal low-frequency component from the delay circuit 1-e and the horizontal second high-frequency component from the multiplier 1-h are added and added. The result of the addition is output via the line Ld '. Waveform diagrams of this final output signal are shown in Figures 38 (a)-(d). Comparing this with FIGS. 32 (a) to 32 (d) showing the input signal waveform, the frame number k is -1.
It can be seen that at the positions 0 and 0, the ringing of the waveform from the scanning line number j = −2 to 0 is significantly reduced, and there is no unnecessary distortion that can be detected in the other waveform portions.

The operation of each functional block of the device of the present invention has been described with reference to the three-dimensional waveform charts shown in FIGS. 32 to 38. Let's look at the characteristics of the final output signal at a typical position as shown in the following equation, where it can be seen that a certain ringing is greatly reduced.

[0101]

[Equation 21]

FIGS. 39 (a) to 39 (d) are waveform characteristic diagrams at the position of equation 21 (1), and FIGS. 40 (a) to 40 (d) are waveform characteristic diagrams at the position of equation 21 (2). 41 (a) to 41 (d) are waveform characteristic diagrams at the position of equation 21 (3), and FIGS. 42 (a) to 42 (d) are waveform characteristic diagrams at the position of equation 21 (4). Compared to the input signal Fig. 48 (a) to (d), a significant reduction effect of ringing can be seen in all waveforms.
The position (4) has the greatest ringing reduction effect. This is because there is a rapid waveform change in the vertical and time directions, and the processing effects in both directions are superimposed. Number 21
The positions of (2) and Eq. 21 (3) are the positions where the ringing reduction effect is greatest next to Eq. 21 (4). This is because there is a position where there is a rapid waveform change in the time direction or the vertical direction. number
21 (1) is the position where a uniform signal waveform exists, with no sudden waveform change in either the time direction or the vertical direction. The effect of the ringing reduction processing performed in the three directions of the horizontal and vertical time is shown. This can be confirmed by comparing Fig. 39 (b) with Fig. 48 (b) of the input signal and observing how the signal amplitude sharply decreases with the distance from the pulse peak.

Also, comparing FIGS. 39 (c) to 41 (c) with FIG. 48 (c) of the input signal, there is almost no distortion within the signal band, and FIGS. 39 (d) to 41 (c). Comparing d) with Fig. 48 (d), it can be seen that the component added outside the signal band is also a regular spectrum having a correlation with the signal, and its value is extremely small. Since the spectrum outside this band is the harmonic component of the transient of the waveform, it is utilized as effective signal energy. By narrowing the width of existence of ringing, the spectrum region of the signal is widened, and the out-of-band region which is originally vacant and has no utility value is effectively utilized. When the spectrum diagram of the input signal in Fig. 48 (c) and the corresponding spectrum diagram of the output signal in Fig. 39 (c) are compared, as shown schematically in Fig. 4 (f), horizontal stripes on the left side of the upper limit frequency fm are shown. Instead of deleting the frequency component of, the vertical stripe frequency component on the right side of fm is formed and added. The component on the right side of fm is a component newly formed by non-linear processing, but it can be considered as an effect of band expansion.

Next, let us see how the corrugated slope is affected. 47 (a) (time waveform diagram) and 47 (b)
(Expanded time waveform diagram), a bar waveform having characteristics as shown in FIG. 47 (c) (spectrum diagram) (pulse width = approximately 36 μs, waveform with sharp edge with amplitude = 1) is used. This waveform is illustrated
29 (a) That is, the final output signal waveform diagram on the line Ld ′ is calculated by arranging it in the test waveform area of FIG.
It is taken out at the position of 21 and the characteristic diagram is obtained.
It becomes like. Fig. 49 is at the position of the number 21 (1), and is at the position of the number 21 (2).
50, Fig. 51 for the number 21 (3), and Fig. 52 for the position of the number 21 (4).
A characteristic diagram of is obtained.

Comparing the time waveform diagram (a) and the enlarged diagram (b) of these output signal waveform diagrams with the input signal waveforms of FIGS. 47 (a) and 47 (b), the ringing of the waveform before and after the sloped portion is observed. It can be seen that the range of existence is significantly reduced. Even if such a ringing reduction process is performed, as can be seen by comparing the time-expanded waveform diagrams, the waveform sloping portion hardly changes.
This is the feature of the present invention. This is as described above, and as can be seen from the comparison of the spectrum diagrams of FIG. 49 (c), FIG. 50 (c), FIG. 51 (c), and FIG. It is considered that this is due to the effect of band expansion in which out-of-band components are formed and added simultaneously with the decrease. Further, comparing FIG. 32, which is an input signal, the output diagram 34 after processing in the time direction, the output diagram 36 after processing in the vertical direction, and the final output diagram 38, when comparing FIG. Even if the waveform is a little distorted temporarily in the waveform changing part, the final output signal waveform in FIG.
It can be seen that the ringing component, which had been generated due to the steep band limitation up to that point, is also greatly reduced.

During the processing in these three directions, Kt (FIG. 33) for the time direction and Kv (FIG. 33) for the vertical direction.
35), window functions such as Kh in the horizontal direction (Fig. 37), etc. work for the high frequency components in each direction, but the waveform diagrams of these window functions are the ones that reduce ringing and reduce noise. It gives an intuitive understanding of the edge-preserving mechanism of the invention. Note that Kt in FIG. 33 and Kv in FIG. 35 have almost the same waveform, but as described in FIG. 29, the arrangement of significant waveforms (test waveforms) is symmetrical in the time direction and the vertical direction. It is due to something. The device of the present invention has not only the ringing reduction function, but also the noise reduction function which is in common with ringing in a sense. Figure 43 is Figure 48
Random noise of SN20 several dB is added to the input signal of and the signal is arranged as shown in FIG. 29, j = k = − on the input line La ′
It is a waveform diagram and a spectrum diagram of the position of 2. FIG. 44 is a characteristic diagram of the line Lb ′ after processing in the time direction, FIG. 45 is a characteristic diagram of the subsequent line Lc ′ after processing in the vertical direction, and FIG. 46 is a diagram of the final output line Ld ′. It is a characteristic view of an output signal.

When the input signal of FIG. 43 (d) is compared with that of FIG. 44 (d), a noise reduction effect of several dB is observed, and FIG. 45 (d)
In comparison with Fig. 46, a noise reduction effect of several dB is further seen, and in comparison with Fig. 46 (d) at the end, in-band high and out-of-band noise is significantly reduced in addition to ringing. I understand. The slightly remaining noise is a noise component in the spatiotemporal low frequency region when viewed in a three-dimensional coordinate system of horizontal and vertical time, which is not a target of processing in the present invention. The noise reduction effect of several dB is obtained by the processing in the time direction in the comparison of FIG. 43 and FIG. 44, and the noise reduction effect of several dB is further obtained by the processing in the vertical direction in the comparison with FIG. 45. And the effect based on the correlation of signals in the vertical direction, which is a positive effect of the three-dimensional processing. It is a feature of the present invention that such a noise reduction effect is provided with a ringing reduction function capable of removing an unnecessary minute high frequency component without giving an abnormal waveform distortion to a main signal component. .

The nonlinear converter, which plays a central role in the ringing reduction process, can have its characteristics arbitrarily changed by changing some parameters. Increasing the coefficient Ko increases the gain of the converted high frequency component and emphasizes the high frequency component. On the contrary, if Ko value is decreased, high frequency components are decreased. When the value of the reference value xo is increased, the width of the window function is narrowed, and the narrowing range of ringing is narrowed. Conversely, as the value of xo becomes smaller, the ringing reduction effect becomes weaker. When the value of r is increased, the ringing is narrowed down strongly and the ringing existence range is narrowed. Conversely, if the value of r is reduced, the ringing reduction effect becomes weaker.
By using such characteristics, it is possible to obtain optimum characteristics according to the application. It is also possible to perform adaptive control by dynamically controlling these parameters. In the present invention, the image quality improving process is performed in order of the time direction, the vertical direction, and finally the horizontal direction. If the reverse procedure is taken, there is a problem that unnecessary harmonic components, especially in the horizontal direction, remain. Further, the 1X low-pass filter 1, the first low-pass filter 11 and the third low-pass filter 1-b have 100% roll-off characteristics up to the upper limit frequency of the contained frequency band of the passing signal. It is set. This is a characteristic selected because a good frequency characteristic with extremely small spectrum distortion in the signal band can be obtained even if the ringing reduction and noise reduction processing by the window function are performed.

As described above, the image quality improving device of the present invention has the following effects. (A) It is possible to obtain the ringing reduction effect with less distortion on the signal including within the signal band and outside the signal band. (B) Even if the ringing existence region is narrowed, the sharpness of the waveform changing portion can be maintained and the sense of resolution is not impaired due to the effect of expanding the signal band. (C) Since it also has a noise reduction function, it is effective in reducing minute noise and coding distortion associated with coding and decoding, and is effective in improving overall image quality. (D) Since it is vertical and horizontal two-dimensional processing, it has a greater effect of reducing ringing and noise than one-dimensional processing. Further, since there is no directionality, it is effective in improving the overall image quality, such as not giving an unnatural feeling. (E) Since it is the three-dimensional processing of the horizontal and vertical times, the ringing and noise reduction effect is greater than that of the one-dimensional and two-dimensional processing. Low-frequency noise components that cannot be completely removed by one-dimensional processing are also reduced. Further, since there is no directionality, it is effective in improving the overall image quality, such as not giving an unnatural feeling. (F) The noise reduction function mainly works on a signal with little ringing, and in any case, it can contribute to image quality improvement. (G) The circuit configuration is not so complicated and can be realized by an ordinary digital circuit, so that the cost / performance is very good. (H) Since the characteristics can be changed by changing the parameters, it can be applied to adaptive control. (I) It can be applied to various image processing and signal processing fields, and particularly to the problem of image quality deterioration due to coding distortion due to coding and decoding.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a first embodiment of an image quality improving device of the present invention.

FIG. 2 is a configuration diagram of a conventional image quality improving device.

FIG. 3 is a detailed configuration diagram of a specific example of the image quality improving apparatus of the present invention.

FIG. 4 is an operation explanatory diagram.

FIG. 5 is an operation explanatory diagram.

FIG. 6 is a characteristic example of a window function.

FIG. 7 is an input signal characteristic diagram, (a) is an amplitude linear display waveform diagram, (b) is an amplitude dB display waveform diagram, (c) is an amplitude linear display spectrum diagram, and (d) is an amplitude dB display spectrum diagram. is there.

FIG. 8 is an output signal characteristic diagram of the first conventional example.

FIG. 9 is a characteristic diagram of an output signal of a second conventional example.

FIG. 10 is an operation waveform diagram (a) of each block of the device of the present invention.
(d).

FIG. 11 is an operation waveform diagram (e) of each block of the device of the present invention.
(h).

FIG. 12 is an operation waveform diagram (i) of each block of the device of the present invention.
(k).

FIG. 13 is an operation waveform diagram (l) of each block of the device of the present invention.
(o).

FIG. 14 is a vertical processing output signal characteristic diagram (scanning line number m-
1).

FIG. 15 is a vertical processing output signal characteristic diagram (scanning line number m
).

FIG. 16 is a vertical processing output signal characteristic diagram (scanning line number m +
1).

FIG. 17 is a final processing output signal characteristic diagram (scanning line number m-
1).

FIG. 18 is a final processing output signal characteristic diagram (scanning line number m
).

FIG. 19 is a final processing output signal characteristic diagram (scanning line number m +
1).

FIG. 20 is a characteristic diagram of an input signal when noise is added.

FIG. 21 is a characteristic diagram of a vertical processing output signal when noise is added.

FIG. 22 is a final processed output signal characteristic diagram when noise is added.

FIG. 23 is an input signal characteristic diagram of a bar waveform edge, where (a) is a waveform diagram, (b) is an enlarged waveform diagram, and (c) is a spectrum diagram.

FIG. 24 is a two-dimensional input / output waveform diagram of bar waveform edges,
(a) is the input signal, (b) is the vertical processing output, and (c) is the final processing output.

FIG. 25 is a waveform diagram of a window function of a bar waveform edge, where (a) is a vertical window function Kv and (b) is a horizontal window function Kh.

FIG. 26 is a final processed output signal characteristic diagram of a bar waveform edge, where (a) is a waveform diagram, (b) is an enlarged waveform diagram, and (c) is a spectrum diagram.

FIG. 27 is a configuration diagram of a second embodiment of the image quality improving device of the present invention.

FIG. 28 is an explanatory diagram of the operation (three-dimensional array of sampling points related to calculation).

FIG. 29 is an operation explanatory view (two-dimensional array of scanning lines related to calculation).

FIG. 30 is an operation explanatory view (waveform display method 1).

FIG. 31 is an operation explanatory view (waveform display method 2).

FIG. 32 is a three-dimensional display waveform diagram (input signal), where (a) k =
-1, j = -2 to 3 (b) k = 0, j = -2 to 3 (c)
k = + 1, j = -2 to 3 (d) k = + 2, j = -2 to 3
Is. Note that FIGS. 33 and 36 to 38 use the same expression method.

FIG. 33 is a three-dimensional display waveform diagram (time window function).

FIG. 34 is a three-dimensional display waveform diagram (time direction processing output),
(a) j = -1, k = -2 to 3 (b) j = 0, k = -2
˜3 (c) j = + 1, k = −2 to 3 (d) j = + 2, k
= -2 to 3. Note that FIG. 15 has the same expression method.

FIG. 35 is a three-dimensional display waveform diagram (vertical window function).

FIG. 36 is a three-dimensional display waveform diagram (vertical processing output).

FIG. 37 is a three-dimensional display waveform diagram (horizontal window function).

FIG. 38 is a three-dimensional display waveform diagram (horizontal processing output).

FIG. 39 is a characteristic diagram of an output signal (j = -2, k = -2).

FIG. 40 is a characteristic diagram of an output signal (j = −2, k = 0).

FIG. 41 is a characteristic diagram of an output signal (j = 0, k = -2).

FIG. 42 is a characteristic diagram of an output signal (j = 0, k = 0).

FIG. 43 is a characteristic diagram of an input signal when noise is added (j = k = −
2).

FIG. 44 is a characteristic diagram of time-direction processed output (j = k = −
2).

FIG. 45 is a characteristic diagram of vertical processing output (j = k = −
2).

FIG. 46 is a characteristic diagram of horizontal processing output (j = k = −
2).

FIG. 47 is a characteristic diagram of an input signal (bar waveform edge) (j =
k = -2), (a) waveform diagram, (b) enlarged waveform diagram, (c) spectrum diagram, and FIGS. 49 to 52 also use the same display method.

FIG. 48 is a characteristic diagram of an input signal (impulse waveform), (a)
It is an amplitude linear display waveform diagram, (b) amplitude dB display waveform diagram, (c) amplitude linear display spectrum diagram, (d) amplitude dB display spectrum diagram. Note that FIGS. 8, 9, and 39 to 46 have the same expression method.

FIG. 49 is an output signal characteristic diagram at a representative point (j = -2, k =
-2).

FIG. 50 is an output signal characteristic diagram at a representative point (j = -2, k =
0).

FIG. 51 is an output signal characteristic diagram at a representative point (j = 0, k =
-2).

FIG. 52 is a characteristic diagram of an output signal at a representative point (j = 0, k =
0).

[Explanation of symbols]

1, 9, 11, 19, 1-b, 1-j Low-pass filter 2, 4, 6, 12, 14, 16, 1-c, 1-e, 1-
g Delay circuit 3,13,1-d Subtractor 5,15,1-f Adder 7,17,1-h Multiplier 8,18,1-i Standard deviation circuit X-a, 1-a, 1- k Non-linear converter Kh Horizontal window function Kt Time window function Kv Vertical window function

Claims (12)

[Claims] 1. An image quality improving device for performing ringing and noise reduction processing on a video signal for reproducing a horizontal and vertical two-dimensional image, wherein an input signal is divided into a vertical low band component and a vertical high band component in the vertical direction. Vertical signal separating means for dividing, vertical envelope extracting means for obtaining an envelope component of the vertical high frequency component, and a vertical window function is formed based on the vertical envelope component and a predetermined first reference value. Vertical window function forming means for converting the vertical high frequency component into a second vertical high frequency component by using the vertical window function, and the second vertical high frequency component Vertical signal recombining means for adding and combining the vertical low-frequency components to form a vertical recombined signal; and a horizontal signal for horizontally dividing the vertical recombined signal into a horizontal low-frequency component and a horizontal high-frequency component. Separation means,
Horizontal envelope extracting means for obtaining an envelope component of the horizontal high-frequency component, and horizontal window function forming means for forming a horizontal window function based on the horizontal envelope component and a predetermined second reference value. A horizontal high frequency conversion means for converting the horizontal high frequency component into a second horizontal high frequency component using the horizontal window function,
An image quality improving apparatus comprising: a horizontal signal recombining means for adding and combining the second horizontal high frequency component to the horizontal low frequency component to form a horizontal recomposition signal. 2. The vertical signal separating means according to claim 1,
A vertical low-pass component is separated using a low-pass filter having a 100% roll-off characteristic that covers the frequency band of the input signal, and then the vertical low-pass component is subtracted from the input signal to obtain a vertical high-pass component. In the horizontal signal separating means, first, a horizontal low-pass component is separated by using a low-pass filter having a 100% roll-off characteristic that covers the frequency band of the input vertical re-synthesis signal. 2. An image quality improving device having a configuration in which a horizontal high-frequency component is obtained by subtracting the horizontal low-frequency component from the vertical re-synthesis signal of. 3. The vertical envelope component according to claim 1, wherein a vertical standard deviation value is obtained from a vertical standard deviation value obtained from the square root of the root mean square of the difference between the vertical low frequency component and the values of a plurality of sampling points of the input signal. Image quality characterized by comprising a vertical envelope extracting means for obtaining and a horizontal envelope extracting means for obtaining a horizontal envelope component from a horizontal standard deviation value obtained from the square root of the square sum average of the horizontal high-frequency components. Improvement device. 4. A vertical envelope extracting means for determining a vertical envelope component by further band-limiting the vertical standard deviation value with a horizontal low-pass filter according to claim 1 or 2, An image quality improving apparatus further comprising horizontal envelope extracting means for determining a horizontal envelope component by performing band limitation on the standard deviation value by a horizontal low-pass filter. 5. The phase angle according to claim 1, wherein a phase angle is obtained from the vertical or horizontal envelope component and a predetermined reference value, and the range of the phase angle is 0 to π. /
A vertical or horizontal window function forming means for obtaining a window function in the range of 0 to 1 by multiplying the trigonometric function by a predetermined power and applying it to a trigonometric function such as a sine function or a cosine function. An image quality improving device characterized by comprising: 6. The phase angle according to claim 5, with respect to a sum of a vertical or horizontal envelope component and a predetermined reference value,
An image quality improving device, characterized in that it is configured to obtain from the vertical or horizontal envelope component or a ratio with the reference value. 7. An image quality improving apparatus for performing ringing and noise reduction processing on a video signal for reproducing a three-dimensional image of horizontal and vertical time, in which the input signal is divided into a time low frequency component and a time high frequency component in the time direction. A time signal separating means, a time envelope extracting means for obtaining an envelope component of the time high-frequency component, and a window function in the time direction based on the time envelope component and a predetermined first reference value. A time window function forming means for forming, a time high frequency component conversion means for converting the time high frequency component into a second time high frequency component using the time window function, and the second time high frequency range A time signal recombining means for adding and synthesizing a component to the time low-frequency component to form a time recombined signal, and dividing the time recombined signal into a vertical low-frequency component and a vertical high-frequency component in the vertical direction. Vertical signal separating means and the vertical A vertical envelope extraction means for obtaining an envelope component of the frequency component, the vertical envelope component and a predetermined second
Vertical window function forming means for forming a window function in the vertical direction on the basis of the reference value of, and a vertical high frequency component for converting the vertical high frequency component into a second vertical high frequency component using the vertical window function. Conversion means, vertical signal recombining means for adding and combining the second vertical high-frequency component to the vertical low-frequency component to form a vertical recombined signal, and the vertical recombined signal in the horizontal direction. Horizontal signal separating means for dividing into a horizontal low-frequency component and a horizontal high-frequency component, a horizontal envelope extracting means for obtaining an envelope component of the horizontal high-frequency component, the horizontal envelope component and a predetermined third criterion. Horizontal window function forming means for forming a window function in the horizontal direction based on the value, and horizontal high frequency component conversion means for converting the horizontal high frequency component into a second horizontal high frequency component using the horizontal window function. When,
An image quality improving apparatus comprising: a horizontal signal recombining means for adding and combining the second horizontal high frequency component to the horizontal low frequency component to form a horizontal recomposition signal. 8. The time signal separating means according to claim 7,
A time low-pass component having a 100% roll-off characteristic that covers the frequency band of the input signal is used to separate the time low-pass component, and then the time low-pass component is subtracted from the input signal to obtain a time high-pass component. And the vertical signal separation means
First, a vertical low-pass filter having a 100% roll-off characteristic that covers the frequency band of the input time-resynthesized signal is used to separate the vertical low-pass components, and then the vertical re-synthesized signal is separated from the vertical low-pass signal. A vertical high-frequency component is obtained by subtracting the low-frequency component, and the horizontal signal separating means first uses a horizontal low-pass filter having a 100% roll-off characteristic that covers the frequency band of the input vertical re-synthesis signal. The image quality improving apparatus is configured to obtain a horizontal high-frequency component by separating the horizontal low-frequency component from the vertical low-frequency component and then subtracting the horizontal low-frequency component from the vertical re-synthesis signal. 9. The time envelope component according to claim 7, wherein a time standard deviation value obtained from a square root of a sum of squared averages of differences between values of a plurality of sampling points of the input signal with respect to the time low frequency component is obtained. Vertical envelope from the vertical standard deviation value obtained from the square root of the mean square sum of the difference between the time envelope extracting means to be obtained and the values of the plurality of sampling points of the input time recombined signal with respect to the vertical low-frequency component. A vertical envelope extracting means for obtaining a line component, and a horizontal envelope extracting means for obtaining a horizontal envelope component from a horizontal standard deviation value obtained from the square root of the square sum average of the horizontal high-frequency components and the like. Image quality improvement device. 10. The time envelope extracting means for defining a time envelope component by band limiting a horizontal low-pass filter for the time standard deviation value according to claim 7, or the vertical direction. Vertical envelope extraction means for determining a vertical envelope component by further band-limiting the standard deviation value with a horizontal low-pass filter, and a band with a horizontal low-pass filter for the horizontal standard deviation value. An image quality improving apparatus, comprising: a horizontal envelope extracting unit that limits the horizontal envelope component. 11. The phase angle according to claim 7, wherein a phase angle is obtained from the time, vertical and horizontal envelope components and a predetermined reference value, and the range of the phase angle is set to 0. ~
π / 2, and this is applied to a trigonometric function such as a sine function or a cosine function.
1. An image quality improving apparatus comprising window function forming means for obtaining a window function in the range from 1 to 1, vertical and horizontal. 12. The phase angle according to claim 11, wherein the phase angle is the sum of the time, vertical or horizontal envelope component and a predetermined reference value, the time, vertical or horizontal envelope component, or the reference. An image quality improving device characterized in that it is configured to obtain it from a ratio with a value.