JPS5927145B2 - Noise removal device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to noise removal in interframe coded transmission of color television signals, particularly color television signals in which color signals and luminance signals are frequency multiplexed.

Conventionally, as a method for noise removal in interframe coding, an isolated point removal method is used. There is a method of treating the pixel as an invalid pixel by regarding it as being caused by noise.

However, according to the isolated point removal method, in cases such as when the amplitude of the isolated pixel is large or when the isolated pixel actually contains color information, the isolated pixel is not due to noise but is significant signal information. Therefore, there is a drawback that isolated point removal also removes important signal information other than noise components. On the other hand, even if significant pixels occur due to noise, if they occur consecutively, they are considered significant pixels, so there is also the drawback that transmission code information increases. As a method similar to noise removal other than isolated point removal, there is a method of inserting a low-pass filter into the frame difference signal to reduce high frequency components. However, when this method is applied to a color television signal, the color information is removed by a low-pass filter, resulting in a drawback that the color signal is significantly degraded. An object of the present invention is to provide a noise removal device that does not have the above-mentioned drawbacks in interframe coding of frequency-multiplexed color television signals, preserves significant signal information, and efficiently removes noise components. The object of the present invention is to realize the transmission of color television signals with less deterioration in image quality and less influence of noise.

A first configuration of the present invention is a color television signal encoding/transmission apparatus that uses a first function {1-F(Z
)}, and a second function f whose value is approximately 1 at least in the vicinity of the zero frequency and the vicinity of the subcarrier frequency, and whose absolute value is smaller than 1 in the frequency domain excluding the vicinity of the zero frequency and the vicinity of the subcarrier frequency. (K, Z) and the signal X(Z? transfer function {1-F(Z)}.
This is a noise removal device consisting of means for removing noise components by passing f(K,Z).

Here, the degree to which the function f(K, Z) becomes smaller than 1 changes depending on the value of K. According to this configuration, the frame difference signal obtained by passing through the function {1-F(Z)} passes through the second function f(K,Z) to reach the zero point where the power spectrum of the color television signal is concentrated. Since it is attenuated except for the vicinity of the frequency and the vicinity of the subcarrier frequency, the noise component can be removed without affecting the main part of the signal. In other words, while the main part of the television signal is concentrated near the zero frequency and the subcarrier frequency, the noise component is almost uniformly distributed over the entire frequency range, so the function f as described above is
The signal-to-noise ratio after passing through (K,Z) will be improved. A second configuration of the present invention realizes the first and second functions {1-F(Z)} and f(K,Z) described in the first purchase in a color television transmission signal. For the means and signal X(Z), {1-F(Z)}・{1f(K
, Z)}・X(Z) is smaller than the threshold δ, R(δ, K, Z)=f(K, Z), and if the absolute value is larger than δ, R(δ, K, The third function R(
δ, K, Z) and means to remove the noise component by passing the signal X(Z) through the transfer function {1-F(Z)}·R(δ, K, Z) It is a removal device.

In the first configuration, all frame difference signals are f(K,Z)
Since the noise is removed, the signal is also distorted in the attenuation region of f(K,Z), although it is not the main part of the signal. The second configuration utilizes the fact that the noise amplitude is generally sufficiently small compared to the signal amplitude to suppress signal distortion as much as possible and to remove noise. That is, {1-F(Z)}{1-f(K
, Z)} is regarded as noise, and the amplitude of this component is suppressed by passing it through the function f(K, Z). If the amplitude is larger than the threshold value δ, it is regarded as a signal, and the amplitude is not suppressed. Therefore, the signal component is not distorted and only the noise component is suppressed, so that image quality can be improved. A third feature of the present invention is a color television signal transmission device in which the first, second and third functions {1-F(Z
)}・f(K,Z) and R(δ,K,Z), and transfer function {1F(Z)}R(δ,
K, Z), and means for automatically changing K or δ or K and δ according to the amount of information generated in the encoding/transmission device.

The third configuration is characterized by adaptively removing noise. In other words, the degree of noise removal varies depending on how K and δ are selected, but it is generally desirable for the amount of transmitted information to be approximately constant in a signal transmission device.
K or δ depending on the amount of information generated in the encoder and transmitter
Alternatively, by adaptively changing both of these, balanced noise removal can be achieved. In other words, when the amount of generated information is large, K is selected so that the attenuation amount in the attenuation region of f (K, Z) is large, and the amplitude judgment threshold δ is also increased to increase the amount of noise removed and reduce the amount of generated information. and when the amount of generated information is small, the opposite control is performed. Increasing δ or increasing the amount of attenuation will remove some signal components other than noise components, but in interframe coding, if the amount of generated information is large, the screen may be working hard. At this time, since the visual characteristics of the eye have become dull when it comes to details, distortion due to removal of signal components representing details is permissible. Hereinafter, the present invention will be explained in detail with reference to Examples.

FIG. 1 is a block diagram showing a specific example of the configuration of the first embodiment of the present invention. In this example, the sampling frequency F
A case will be described in which an NTSC color television signal sampled at a frequency approximately three times the subcarrier frequency F8O is subjected to composite predictive coding. The input color television signal X(t) is converted by an A/D converter (analog-to-digital converter) 10 into a PCM (pulse code modulation) signal sent to.
The subtracter 21 calculates the difference between the output of the A/D converter 10 and the output of the frame memory 27. That is, the subtracter 21 calculates the interframe prediction error. The interframe prediction error calculated by the subtracter 21 has its noise component removed by a noise removal circuit 30 and is sent to the subtracter 22 . The subtracter 22 calculates the difference between the interframe prediction error from which noise components have been removed by the noise removal circuit 30 and the output of the intraframe predictor 26 .

That is, the intra-frame prediction error of the inter-frame prediction error,
That is, a composite prediction error is calculated by the subtractor 22. The composite prediction error calculated by the subtracter 22 is quantized by a quantization circuit 25 according to a predetermined quantization characteristic. The sum of the composite prediction error quantized by the quantization circuit 25 and the output of the intra-frame predictor 26 is calculated by the adder 23.

That is, the locally decoded signal of the interframe prediction error is obtained as the output of the adder 23. The interframe prediction error locally decoded by the adder 23 is sent to the adder 24 and the intraframe predictor 26.
sent to. The intra-frame predictor 26 calculates an intra-frame predicted value using the locally decoded inter-frame prediction error, and sends the calculated intra-frame predicted value to the subtracter 22 and adder 23. The adder 24 calculates the sum of the locally decoded interframe prediction error sent from the adder 23 and the output of the frame memory 27. That is, a locally decoded color television signal is obtained at the output of the adder 24. The color television signal locally decoded by the adder 24 is sent to the frame memory 27, where it is delayed by a time equivalent to one frame time, and the delayed output is sent to the subtracter 21 and the adder 24.
sent to. The composite prediction error quantized by the quantization circuit 25 is sent to the encoder 40, and the encoder 40 generates a code corresponding to each quantization level and sends it to the transmission path.

By the way, the color television signal format of the NTSC system is such that the subcarrier phase is inverted between successive frames, so even if the input to the encoding device is a still image, the interframe prediction error of the subtractor 21 will occur. Although the amplitude of the carrier color signal component is doubled, it is only necessary to perform efficient intra-frame predictive coding on this carrier color signal component.

A prediction function P(2
) is a well-known example.

Therefore, P(ZZ shown in equation (1)) may be used as the prediction function used in the intra-frame predictor 26, and a predictor having such characteristics can be easily constructed using a non-recursive digital filter. As is clear from the above explanation of the operation of the composite predictive coding device, the important meaning of the interframe prediction error output from the subtractor 21 is that the low frequency components and sub-frame prediction errors of the interframe prediction error are important for motion. The components near the carrier wave frequency, and even in the case of a still image, the components near the subcarrier signal have an important meaning.

Therefore, by suppressing components other than the low frequency component and subcarrier frequency component of the interframe prediction error using the noise removal circuit 30, it is possible to increase the encoding efficiency in composite predictive encoding without substantially degrading the image quality. can. Therefore, the transfer characteristics required for the noise removal circuit 30 are those shown in FIG. 2, that is, the gain is almost 1 near the zero frequency and the subcarrier frequency F8O, and the gain is smaller than 1 in other frequency regions. All you have to do is stay there. FIG. 3 shows a block diagram showing a specific configuration of the noise removal circuit 30. In FIG. 3, the input signal to the noise removal circuit 30 is sent to a subtracter 101 and a noise extraction filter 103 given by a transfer function H(Z), and the noise is extracted by the noise extraction filter 103. It is sent to a subtracter 101 and a multiplier 104. The subtracter 101 calculates the difference between the input to the noise removal circuit 30 and the output of the noise extraction filter 103. That is, from the output of the subtracter 101, only significant signal components from the input to the noise removal circuit 30 are extracted and sent to the adder 102. The multiplier 104 multiplies the noise extracted by the noise extraction filter 103 by K and sends it to the adder 102 . However, K is O<. The value is selected to satisfy K<1. Therefore, the noise extracted by the noise extraction filter 103 is suppressed according to the value of K. As is clear from the above explanation, the transfer function of the noise removal circuit is also a function of K, so if this is expressed as f(K,Z), the following equation is obtained.

Ingredients\▲h'61-ノ? ~? --: ノ●ANG-' And the transfer function H(Z) of the noise extraction filter 103 is designed to block low frequency components and components near the subcarrier frequency, while allowing other components that are not very important for composite predictive coding to pass through as is. An example is a function that only needs to have the following characteristics.

FIG. 4 shows the specific purchase of the noise extraction filter 103 that realizes the characteristics given by equation (3).

In FIG. 4, 201, 202, and 203 are shift registers, 204 is a subtracter, and 205 is a multiplier with a magnification of 0.5. Shift registers 201, 202 and 20
3 operates at a clock frequency F, (?3f8.), shift registers 201, 202, and 203 delay the input signal by one sampling period, and the input signal of shift register 201 is three-sampled at the output of shift register 203. Output is delayed by one cycle. A subtracter 204 calculates the difference between the input signal to the noise extraction filter 103 and the output of the shift register 203, that is, a signal delayed by three sampling periods, and the multiplier 205 multiplies the difference by 0.5. To summarize the above, the input to the noise extraction filter 103 is 0.5(1-
The calculation Z-3) is performed. As is clear from the above explanation, the transfer function f(K, Z) of the noise removal circuit has the characteristics shown in Figure 2, and the value of K suppresses signal components other than low frequency components and signal components near the subcarrier frequency. The degree of

If the value of K is selected to be small, the noise component will be suppressed more, and if K-1 is selected, f(K,Z)-1 will be obtained, and the noise component will not be suppressed. FIG. 5 is a block diagram showing another configuration of the noise removal circuit 30.

In FIG. 5, 301 is a subtracter, 302 is an adder, and 303 is an adder.
is the transfer function S(Z) of the signal extraction filter 1 and 30
4 is a multiplier with a multiplication factor K (0≦K≦1). As clearly understood from FIG. 5, the noise removal circuit 30 shown in FIG.
The transfer function f(K, Z) is as follows.

S(Z) is the prediction function P(
Similarly to Z), if you choose a function that efficiently predicts the low frequency component and carrier color signal component, f(K, Z) shown in equation (4) will have the frequency characteristics as shown in Figure 2, and if K is chosen small, The suppression of noise components becomes greater. Figure 6 shows S(Z) as (
1) A block diagram showing a specific configuration when the same function as P(Z) shown in equation 1 is selected is shown.

In FIG. 6, 401, 402, 403 and 404 are shift registers, 405 and 406 are multipliers with a magnification of 0.5 and β-1-2-N (N: positive integer), 407 is a subtracter, and 408 is an adder. It is a vessel. Since each operation is clear from the diagram, the explanation will be omitted. As is clear from the above description of the first embodiment, by passing the interframe prediction error through the transfer function f(K,Z) having the characteristics shown in FIG. 2, there is no effect on the main components of the signal. It is possible to remove noise components without adding

Note that the noise extraction filter 103 and the fifth noise extraction filter in FIG.
The function of the signal extraction filter 303 in the figure is not particularly limited, and may take any form as long as f(K, Z) has the characteristics shown in FIG.

FIG. 7 shows the function R(δ, K
, Z) is a block diagram showing a specific configuration for realizing.

In FIG. 7, the subtracter 101, adder 102, and noise extraction filter 103 are replaced by the subtracter 101 in FIG.
It has exactly the same function as the adder 102 and the noise extraction filter 103.

If the absolute value of the input amplitude is less than or equal to a predetermined threshold δ, the nonlinear circuit 504 multiplies the input signal by K and outputs δ.
If the above is the case, the circuit has input/output characteristics as shown in FIG. 8, in which the output is approximately multiplied by 1. If the nonlinear circuit 504 having such characteristics is used instead of the multiplier 104 shown in FIG. 3, f(K, Since the signal passes through a function with the characteristic Z), generally only small amplitude noise components can be selectively suppressed. As described above in the second embodiment, if the transfer function of the noise removal circuit 30 in FIG. Noise removal with very little deterioration can be achieved.

Note that the input/output characteristics determined by S and K of the nonlinear circuit 504 in the second embodiment are not particularly limited.

FIG. 9 is a block diagram showing the configuration of a third embodiment of the present invention.

In FIG. 9, the A/D converter 10 subtracters 21 and 22,
Since the adders 23 and 24, the quantization circuit 25, the intra-frame predictor 26, and the frame memory 27 have exactly the same functions as those shown in FIG. 1, a detailed explanation of their operations will be omitted. In this embodiment, the composite prediction error quantized by the quantization circuit 25 is converted into unequal length codes corresponding to the respective quantization levels by the unequal length encoding circuit 41 and sent to the buffer memory 50.

The unequal length encoding circuit 41 encodes the quantized composite prediction error using a predetermined unequal length code so that the average code length is the shortest on average according to the composite prediction error distribution. Since equal-length encoding is performed, the amount of information sent to the buffer memory 50 changes from moment to moment depending on the color television signal that is the input signal to the encoding device. Therefore, the buffer memory 50 temporarily stores the constantly changing amount of information sent from the unequal length encoding circuit 41, converts the speed to match the transmission speed of the transmission path, and transfers the information to the transmission path. send out. The buffer memory 50 monitors the amount of information stored in the buffer memory 50 and sends the information storage amount to the control circuit 60. Control circuit 6
0 changes the parameters K and δ that determine the characteristics of the noise removal circuit 31 according to the amount of information stored in the buffer memory.
However, the configuration of the noise removal circuit 31 is similar to the block diagram shown in FIG. 7, and the input/output characteristics of the nonlinear circuit 504 can be controlled from the outside. That is, the control circuit 6
0, if the amount of information stored in the buffer memory increases and an overflow is about to occur, the value of K is decreased and the value of δ is increased according to predetermined rules. Control.

If such control is performed, not only noise components but also signal components are suppressed, thereby suppressing the generation of information in composite predictive coding, and overflow of the buffer memory 50 can be prevented. On the other hand, if the amount of good news stored in the buffer memory 50 decreases and an underflow is about to occur, the control circuit 60 changes the value of K to 1 or 1.
, and control δ to be zero or a value close to zero. If such control is performed, the noise removal circuit 31 will not apply any suppression operation to the signal component and the noise component, which will increase the amount of information generated in composite predictive coding and prevent underflow of the buffer memory 50. can do.

As described above in the third embodiment, by adaptively controlling the parameters K and δ of the noise removal circuit 31 according to the amount of information stored in the buffer memory 50, human visual characteristics can be adjusted. Adaptable information compression can be achieved.

Note that the configuration of the noise removal circuit 31 is not limited to the configuration shown in FIG. 7, and the configurations shown in FIGS. 3 and 5 may be used.

Furthermore, the composite predictive encoding device in the third embodiment can also use an encoding mode for controlling the generation of various types of information used in normal composite predictive encoding. Further, in the first, second, and third embodiments, only the case where the sampling frequency is selected to be approximately three times the subcarrier frequency has been described, but the sampling frequency is not particularly limited.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of the first embodiment, FIG. 2 is a diagram showing an example of the frequency characteristics of the noise removal circuit 30 shown in FIG. 1, and FIG. 3 is a specific configuration of the noise removal circuit 30. 4 is a block diagram showing an example of a specific configuration of the noise extraction filter 103 of the noise removal circuit of FIG. 3, and FIG. 5 is a block diagram of another example of the noise removal circuit 30 of FIG. 1. 6 is a block diagram showing an example of the configuration of the signal extraction filter 303 in FIG. 5, and FIG.
The figure is a block diagram showing the configuration of the noise removal circuit in the second embodiment, FIG. 8 is a diagram showing an example of the input/output characteristics of the nonlinear circuit 504 in FIG. 7, and FIG. 9 is the block diagram of the third embodiment. FIG.

Claims (1)

[Claims] 1. In a color television signal encoding and transmission device, X(Z), F(Z) and f(K,Z) are Z-transformed expressions in a sample value system, and Z^-^1 is A means for realizing a first function {1-F(Z)} whose absolute value is sufficiently smaller than 1 at least near the frame frequency, when K is a constant and represents a one-sample delay; Realizes a second function f(K,Z) whose value near the carrier frequency is approximately 1, and whose absolute value is smaller than 1 and whose degree depends on K in the frequency domain except near the zero frequency and near the subcarrier frequency. means to transfer the signal X(Z) to a transfer function {1-F(Z)}・f(K,Z)
1. A noise removal device comprising means for removing a noise component by passing through the noise component. 2 In a color television signal encoding and transmission device, X(Z), F(Z), f(K,Z) and R(δ,K
, Z) is the Z transformation representation in the sample value system, and Z^−
means for realizing a first function {1-F(Z)} whose absolute value is sufficiently smaller than 1 at least near the frame frequency, where ^1 represents a one-sample delay and K and δ are constants; A second function f() whose value near zero frequency and subcarrier frequency is approximately 1, and whose absolute value is smaller than 1 and whose degree depends on K in the frequency region except near zero solid wave number and near subcarrier frequency K
, Z) and means for realizing {1-
F(Z) {1-f(K,Z)} If the absolute value of X(Z) is smaller than the threshold δ, R(δ, K, Z) = f(K, Z),
means for realizing a third function R(δ, K, Z) in which R(δ, K, Z) becomes approximately 1 when the absolute value is larger than δ; −F(Z)}R(δ,
K, Z). 3 In a color television signal encoding and transmission device, X(Z), F(Z), f(K,Z) and R(δ,K
, Z) is the Z transformation representation in the sample value system, and Z^−
means for realizing a first function {1-F(Z)} whose absolute value is sufficiently smaller than 1 at least near the frame frequency, where ^1 represents a one-sample delay and K and δ are constants; The second function f(K
, Z) and means for realizing {1-
F(Z)}{1-f(K,Z) If the absolute value of X(Z) is smaller than the threshold δ, R(δ, K, Z) = f(K, Z),
means for realizing a third function R(δ, K, Z) in which R(δ, K, Z) becomes approximately 1 when the absolute value is larger than δ; −F(Z)}R(δ,
K, Z), and means for automatically changing K or δ or δ according to the amount of information generated in the encoder/transmission device.