JPH05145802A - Impulse noise elimination device - Google Patents

Abstract

PURPOSE:To reduce mis-detection of an impulse noise in a moving picture in the impulse noise elimination device. CONSTITUTION:A luminance signal (Y) 100 from a Y/C separation circuit 14 is inputted to a frame memory 101 and an adder 115, and a difference from a frame delay luminance signal 102 delayed by one frame period is calculated by the frame memory and a frame difference signal 105 is extracted through an absolute value circuit 104. The frame difference signal 105 is given to a comparator circuit 106 in which the level of the signal 105 and the level of a motion detection threshold level signal 107 are compared and a 1H after motion detection signal 108 is obtained in response to the quantity and a 1H preceding motion signal 110 is obtained through line memories 109-1, 109-2 and both the signals are inputted to a NOR circuit 111, from which a motion detection signal 112 is outputted.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an impulse noise eliminator suitable for satellite broadcasting receivers and the like, and more particularly to a structure of impulse noise detecting means for reducing false detection of impulse noise in a moving image.

[0002]

2. Description of the Related Art Since the radio waves of satellite broadcasting are in the microwave band, if the reception level is lowered due to rainfall or snow on the antenna, the C / N (carrier / noise) of the received signal is lowered. As the C / N decreases, the S / N (signal / noise) of the video signal after FM demodulation decreases,
Furthermore, when C / N decreases and exceeds the threshold value, impulse noise called medaka noise occurs.

FIG. 6 shows a waveform of medaka noise generated in a video signal. For a video signal A whose luminance changes stepwise as shown in (a), C / N as shown in (b). Is lowered and medaka noise B is generated. When this medaka noise is generated, the brightness is significantly increased or decreased for several hundreds of nanoseconds. When viewed on a monitor screen, it looks like a white or black medaka, and the image quality is greatly impaired. In order to remove such medaka noise, for example, a noise removing circuit as shown in FIG. 7 described in Technical Report of the Television Society of Japan, August, 1990, Vol. 14, No. 42, P16-18 is used. Are known.
In FIG. 7, 51 is a digital noise filter, 52
Is a difference circuit, 53 is a comparison circuit, and 54 is an erroneous detection prevention circuit. First, the demodulated signal is AD-converted and the digital notch filter 51 removes the color subcarrier and the digital audio subcarrier. Next, the difference circuit 52 extracts the high frequency component of the video signal. The level comparison circuit 53 compares the signal with a threshold value and determines a pixel exceeding the threshold value as a candidate for impulse noise. Further, in order to prevent erroneous detection, the circuit 54 determines that the upper and lower pixels are impulse noise only when they are not candidates for impulse noise.

Using this noise removing circuit, medaka noise is detected from the demodulated signal of the video signal, and the pixel containing the detected medaka noise is replaced with the pixel containing no medaka noise before or after one field or one scanning line. The medaka noise is removed by. Specifically, when a pixel to be replaced is selected, one field or one field as shown in FIG. 3 is set so that the phase of the color subcarrier frequency-multiplexed in the video signal becomes continuous after the pixel is replaced. It is necessary to replace it with diagonal pixels before and after one scanning line.

[0005]

The conventional impulse noise detection circuit for detecting noise based on the horizontal and vertical correlations in the field has a drawback that medaka noise is easily erroneously detected. Also, in the conventional circuit, since the composite video signal is interpolated as it is, it is necessary to match the phases of the color subcarriers, and therefore it is necessary to interpolate.
Since the interpolation is performed by the diagonal pixels adjacent to each other, there is a problem that the accuracy of the interpolation is poor.

Therefore, the present applicant has previously performed accurate and reliable detection and removal of impulse noise in a video signal when the C / N of the received signal is lowered and impulse noise is generated as described above. Japanese Patent Application No. 3-16 of Japanese Patent Application No. 3-16
Proposed in No. 3626. FIG. 8 shows a television receiver including the noise detection circuit, which will be described in detail below.

In the figure, the reception signal received by the antenna 1 is frequency-converted into a first IF signal 3 in the LNB 2 and sent to the tuner 4. The tuner 4 selects a desired channel and performs FM demodulation. This tuner 4
Outputs a video signal 5, a detection signal 7, and an audio signal 11. The detection signal 7 is a signal immediately after the FM demodulation. After the detection signal 7 is de-emphasized, the 15 Hz triangular wave is removed and the LPF in the video signal band is passed to obtain the video signal 5. The noise reduction circuit 6 detects and removes medaka noise, the detailed configuration of which will be described later.

When the medaka noise is detected in the noise reduction circuit 6, a portion which is not medaka noise may be erroneously detected as medaka noise (erroneously detected), and a noise removing operation is performed on this erroneously detected portion. If you do, the image quality of that part will be impaired. Therefore, the occurrence probability of false detection must be made sufficiently smaller than the occurrence probability of medaka noise. Further, when C / N (for example, about 6 dB or less) in which medaka noise is not generated is obtained, it is necessary to prohibit noise removal in the noise reduction circuit 6.

Therefore, a noise detection circuit 8 is provided, and C
/ N is detected. The noise detection circuit 8 detects only the noise component included in the detection signal 7 and obtains a voltage corresponding to the noise level (for example, about 1 to 2 V). From the voltage, C / It is determined whether or not it is N, and the determination result is the noise reduction circuit 6
In addition, the noise reduction control signal 9 is output.

When the control signal 9 indicates C / N in which medaka noise is generated, the noise reduction circuit 6 detects the medaka noise portion in the video signal and interpolates the luminance of the portion. And supplies it as a noise reduction video signal 10 to the monitor 12. When the control signal 9 indicates C / N in which no medaka noise is generated, the video signal 5 from the tuner 4 is supplied to the monitor 12 without being interpolated. On the other hand, tuner 4
The audio signal from the monitor is supplied to the monitor 12 as in the conventional case.

FIG. 9 shows a schematic structure of the noise reduction circuit. In the figure, the FM demodulated video signal 5 from the tuner 4 is 4 fsc by the A / D converter 13.
A / D at a sampling frequency of (approximately 14.318 MHz)
After conversion, the YC separation circuit 14 separates the luminance signal (Y signal) and the chroma signal (C signal). The Y signal is applied to the impulse noise detection circuit 6-1 and impulse noise (medaka noise) is detected by the Y signal, and the noise interpolation circuits 6-2 and 6-3 perform Y detection according to the detection result.
For the signal and the C signal, the pixel in which noise is detected is interpolated by the adjacent pixel that does not contain noise. The interpolated Y and C signals are mixed by the YC mix circuit 39, D / A converted by the D / A conversion circuit 40, and output.

FIG. 5 is a block diagram showing a concrete configuration example of the noise reduction circuit 6, 13 is an A / D converter, and 14 is
Is a YC separation circuit, 15 is a luminance signal, 16-1 to 16-4
Is a line memory, 17 is a 1H delayed luminance signal, and 18 is 2H.
Delayed luminance signal, 19 is a frame memory, 20 is a frame delayed luminance signal, 21-23 are difference signals, 24-1-24-
3 is an absolute value circuit, 25-1 to 25-3 are comparison circuits, 26
Is a threshold value, 27 is a 3-input AND circuit, 28 is a medaka noise detection signal, 29 is an interpolation luminance signal, 30, 31 are switches, 32 is a chroma signal, 33 is a 1H delay chroma signal,
34 is a 2H delayed chroma signal, 35 is a phase inversion circuit, 36
Is an interpolated chroma signal, 37 and 38 are switches, 39 is YC
Mix circuit, 40 is a DA converter, 41-1 to 41-3
Is a subtractor, 42-1 to 42-2 are adders, 43-1 to 4-4
3-2 shows a 1/2 multiplier.

In FIG. 5, the video signal 5 from the tuner 4 is passed through an A / D converter 13, and then a YC separation circuit 14 is provided.
The luminance signal 15 and the chroma signal 32 are separated by and are supplied to the line memories 16-1 to 16-4, respectively. And
1H and 2H delayed luminance signals 17 and 18 are obtained from the line memories 16-1 and 16-2, respectively. On the other hand, 1H and 2H chroma signals 33 and 34 are obtained from the line memories 16-3 to 16-4, respectively. Further, the 1H delayed luminance signal 17 is passed through the frame memory 19 to obtain the frame delayed luminance signal 20.

Next, the 1H delayed luminance signal n (0) 17 to the 2H delayed luminance signal n (A) 18 and the luminance signal n (B).
15 and the frame delay luminance signal n (C) 20 are subtracted by subtractors 41-1 to 41-3, respectively, and their difference signals,
That is, n (0) -n (A) signals 21, n (0) -n
The (B) signal 22 and the n (0) -n (C) signal 23 are calculated, the difference signals thereof are passed through the absolute value circuits 24-1 to 24-3 to obtain the absolute values, and the comparison circuits 25-1 respectively. ~ 25
The medaka noise detection signal 28 is obtained by determining the magnitude with the threshold value 26 at -3 and inputting the result to the 27 of the 3-input AND circuit. Here, the threshold value 26 is set to an optimum value through experiments.

When medaka noise is detected, it is necessary to perform an operation of replacing the detected portion with some signal, that is, interpolation. Therefore, in the embodiment of FIG. 4, the luminance signal 15 and the 2H delayed luminance signal 18 are added to the adders 42-1 and 1/2.
The signal is passed through the multiplier 43-1 and interpolated by the average value of these signals, and this is input to the switch 30 as the interpolated luminance signal 29.

In the switch 30, the 1H delayed luminance signal 17
And the interpolated luminance signal 29 are selected by the medaka noise detection signal 28 output from the AND circuit 27 and output. The switch 31 is controlled by the noise reduction control signal 9 so as to select the 1H delayed luminance signal 17 in the C / N where medaka noise does not occur and to select the output of the switch 30 in the C / N where medaka noise occurs. ..

Also for the chroma signal 32, it is necessary to interpolate when medaka noise is detected. Therefore, the chroma signal 32 and the 2H delayed chroma signal 34 are passed through the adder 42-2 and the 1/2 multiplier 43-2 to obtain the average value of these signals, and the phase inversion circuit 35 inverts the phase. The interpolated chroma signal 36 is obtained by
The H-delayed chroma signal 33 is input to the switches 37 and 38, and the switch 37 that operates in synchronization with the switch 30 selects and outputs the medaka noise detection signal 28. The switch 38 is synchronized with the switch 31, and its operation is exactly the same.

The phase of the interpolated chroma signal 36 is inverted by the phase inversion circuit 35 because the phase of the chroma signal is inverted line by line in the NTSC system, so the interpolated chroma signal 36 is delayed by 1H. This is to make the phase the same as the chroma signal 33. The outputs of the switches 31 and 38 are combined by the YC mix circuit 39, converted into an analog signal by the D / A converter 40, and output as the noise reduction video signal 10.

When viewed on a pixel-by-pixel basis, a television image has a high correlation with adjacent pixels in the horizontal and vertical directions, and shows a high correlation between adjacent frames. On the other hand, since the impulse noise is temporally random, the correlation between lines (scanning lines) and frames is very low.

The impulse noise detector in FIG. 5 utilizes these characteristics to detect impulse noise, and its outline will be described below. First, in the Y signal, the target pixel is subjected to level comparison with upper and lower pixels in the same field, and when the absolute value of the level difference exceeds the set threshold value for both upper and lower pixels. Determined as the first candidate for impulse noise. This will be described with reference to FIG.

In the figure, the target pixel is S₁₁,Up
The pixel at the same horizontal position on the line is S_Ten, The same water on the bottom line
Pixel in flat position is S₁₂And the threshold value is R₁Then, | S₁₁-S_Ten|> R₁ , And | S₁₁-S₁₂|> R₁  When the condition of is satisfied, S₁₁Is the impulse noise
It is assumed to be a candidate of 1. In this way vertical correlation
The noise was detected when the impulse noise was constant.
Noise is horizontal when viewed in pixel units because it has a space
This is because they have a correlation.

However, the correlation between the upper and lower lines is
The first candidate of impulse noise detected by the
The vertical high-frequency component in the signal is included.
Vertical edges such as letters may be falsely detected as noise.
There is. To prevent such erroneous detection, one frame
Pixel S before frame_{twenty one}Level comparison with
The absolute value of the difference is the threshold value R₁Impulse is exceeded
Use it as a candidate for noise. In other words, the first of the impulse noise
Among the candidates for₁₁-S_{twenty one}|> R₁  If the condition of is satisfied, the pixel S₁₁Impulse noise
To be a candidate.

According to the above impulse noise detection method, erroneous detection can be almost completely prevented in a still image. By the way, the noise detection method described above is noise extraction by a kind of filter, and if the frequency domain extracted by this filter is shown on the vertical-time frequency axis, it becomes the shaded area in FIG.

However, if this noise detection method is applied to a moving image, erroneous detection will be performed on the image signal in the frequency domain shown by the shaded area in FIG. If interpolation is performed using adjacent pixels in response to this erroneous detection, that portion will become image distortion and will deteriorate image quality.

An object of the present invention is to propose an impulse noise removing apparatus which improves image quality by preventing image distortion due to erroneous detection of impulse noise in a specific frequency region in a moving image.

[0026]

In order to achieve the above object, an impulse noise removing apparatus of the present invention demodulates a modulated reception signal from a reception antenna to obtain a video signal including a detection signal and luminance information. And a noise component included in the detection signal, and based on the amount of the noise component, the presence or absence of an impulse noise component with a single large luminance change included in the video signal is determined, and a noise determination signal is output. By means of the noise detection determination means and the noise determination signal, when the impulse noise component is present, the impulse noise component in the video signal is detected, and the luminance of the detected portion is interpolated to thereby remove the impulse noise component from the video signal. And a noise detection / removal means for outputting the video signal from the demodulation means when there is no impulse noise component. A frame difference component extracting means for extracting a frame difference component signal from a video signal including the luminance information and a signal obtained by delaying the video signal for a predetermined frame period, and comparing the levels of the frame difference component signal and a predetermined reference signal. Then, based on the comparison result, the presence or absence of motion in the upper and lower adjacent pixels in the same field is detected, and a motion detection unit that outputs a motion detection signal, and the motion detection signal, only when there is motion detection, It is considered that there is no impulse noise component, and an impulse noise determination unit that outputs only the video signal from the demodulation unit as an output of the noise detection and removal unit is provided.

[0027]

In the apparatus of the present invention, the noise component included in the detection signal obtained from the received signal is detected, and the noise determination signal for determining the presence / absence of the impulse noise component is output based on the amount of the noise component. With this noise determination signal, when the impulse noise component is present, the impulse portion in the video signal is detected, the luminance of the detected portion is interpolated, and the video signal from which the impulse noise portion is removed is output.

When there is no impulse noise component,
The video signal is output. The levels of the frame difference component signal extracted from the video signal including the brightness information and the signal obtained by delaying this signal for a predetermined frame period and the predetermined reference signal are compared. Based on the comparison result, a motion detection signal that detects the presence or absence of motion in vertically adjacent pixels in the same field is output. Only when motion is detected by this motion detection signal, it is considered that there is no impulse noise component and only the video signal is output.

[0029]

The present invention will be further described below with reference to the drawings.
To prevent erroneous detection of impulse noise in the moving image, the motion of the pixel is detected in the time-frequency region of the shaded portion in FIG. 9 where the erroneous detection occurs, and the pixel for which motion is detected is excluded from candidates for impulse noise. This is possible. However, the motion detection based on the frame correlation of one pixel, which is usually performed, is exactly the same as the above-described impulse noise detection method and is not suitable because the motion of noise cannot be discriminated. Therefore, the present invention adopts a motion detection method as described below.

FIG. 3 shows the above-mentioned motion detection method.
Elephant pixel S₁₁The upper and lower pixels S_TenAnd S₁₂Match
The three pixels that have been set are considered as one block, and S_TenAnd S₁₂of
If motion is detected in either one of the pixels, the pixel S₁₁To
It is determined that there is movement. Pixel S_TenAnd S₁₂Motion detection
Is the pixel S one frame before₂₀, S_{twenty two}Level with
The comparison is performed and the level difference and the threshold value
U That is, the threshold for motion detection is R₂Then, | S_Ten-S₂₀|> R₂ Or | S₁₂-S_{twenty two}|> R₂  When the condition of is satisfied, S₁₁Was determined to have moved,
Exclude from impulse noise candidates.

FIG. 4 shows a principle circuit configuration for carrying out the above-described motion detecting method of the present invention. In the figure, 2
00 is the pixel S ₁₂ , 201 is S ₁₁ , 202 is S ₁₀ , 2
Reference numerals 03-1 to 203-4 are line memories, 204 is a frame memory, 205 is S ₂₂ , 206 is S ₂₁ , 207 is S ₂₀ , 208 is a motion detection signal before 1H, 209 is a motion detection signal after 1H, 210 Is a motion detection signal, 211 is an OR circuit, and 212-1 and 212-2 are frame correlation determination circuits.

S ₁₂ (200) is the line memory 203-1
And a line memory 2 added to the frame memory 204.
S ₁₁ (201) is obtained from 03-1 and S ₁₀ is output via the line memory 203-2. S ₂₂ is output from the frame memory 204, and the line memory 203-3
Then, S ₂₁ (206) is obtained and further line memory 2
S ₂₀ is output via 03-4.

The frame correlation determination circuit 212-1 is │S ₁₀
-S ₂₀ |> R ₂ is judged and S 1H before S ₁₁ (201)
The motion detection for ₁₀ (202) is performed, and the 1H previous motion detection signal 208 is output. Also, the frame correlation determination circuit 2
12-2 determines | S ₁₂ −S ₂₂ |> R ₂ and determines S ₁₁ (20
Motion detection for S ₁₂ (200) after 1H in 1) is performed, and a motion detection signal 209 after 1H is output.

The signals 208 and 209 are the OR circuits 211.
And the motion detection signal 210 is obtained. FIG. 1 is an embodiment based on the above-described principle of the present invention. A portion indicated by a broken line is an impulse noise detecting section similar to that in FIG. 5, and a portion indicated by a chain line is a motion detecting section added by the present invention.

FIG. 2 shows another embodiment of the present invention. The difference from FIG. 1 is that the structure of the impulse noise detector shown by the broken line is partly changed in order to save the frame memory.

In FIG. 1, the motion detecting section includes a frame memory 101, an adder 115, an absolute value circuit 104, a comparison circuit 106, line memories 109-1 and 109-2,
Comprising a NOR circuit 111 and an AND circuit 113, 102 is a frame delay luminance signal, 105 is a frame difference signal, 10
7 is a motion detection threshold signal, 108 is a 1H backward motion detection signal, 110 is a 1H forward motion detection signal, and 112 is a motion detection signal.

In the motion detecting section, the frame memory 101, the adder 115 and the absolute value detecting circuit 104 constitute the frame difference component extracting means. Further, the comparison circuit 1
06, the line memories 109-1 and 109-2 and the NOR circuit 111 constitute the motion detecting means, and the AND circuit 113 constitutes the impulse noise determining means.

The impulse noise detecting section includes line memories 116-1, 116-2, 116-3, adders 117-1, 117-2, comparing circuits 118-1, 11.
8-2, 118-3, absolute value circuits 119-1, 119-
2 and an AND circuit 121, which constitutes the noise detection determination means, and the noise detection and removal means has the same configuration as the noise interpolation sections (Y) and (C) in FIG.

In the embodiment of FIG. 2, the same reference numerals as those in FIG. 1 denote the same or similar circuits, and the motion detecting unit has the same configuration, but the impulse noise detecting unit is the frame memory 12.
2, adder circuit 123, absolute value circuit 124 and comparison circuit 1
25 is further provided and has a configuration similar to that of FIG.
The impulse noise detection unit in FIG. 1 shares a frame memory and the like with the motion detection unit, and has substantially the same function as in FIG.

Next, the operation of the motion detector of the embodiment shown in FIG. 1 will be described. Luminance signal (Y) from the Y / C separation circuit 14
100 is input to a frame memory 101 and an adder 115, a difference from a frame delay luminance signal 102 delayed by one frame period is calculated by this frame memory, and a frame difference signal 105 is extracted via an absolute value circuit 104. ..

This frame difference signal 105 is sent to the comparison circuit 10.
6, the level is compared with the motion detection threshold value signal 107, the 1H backward motion detection signal 108 is obtained according to the magnitude, and the 1H forward motion signal 110 is obtained through the line memories 109-1 and 109-2. NOR circuit 111 for both signals
And the motion detection signal 112 is output.

The motion detection signal 112 becomes L level when motion is detected, and becomes H level when motion is not detected. Further, since the impulse noise candidate signal (the medaka noise detection signal) becomes H level when it is determined that the candidate is impulse noise, it is output from the AND circuit 113 to which the motion detection signal 112 and the impulse noise detection signal 120 are input. Impulse noise detection signal 114
Is excluded from candidates for impulse noise when motion is detected, and is regarded as having no impulse noise component.

In this way, the corresponding pixel is excluded from the impulse noise candidates according to the motion detection, and the false detection of impulse noise in the moving image can be reduced. The motion detection in the embodiment of FIG. 2 is also the same as that described above.

[0044]

As described above, according to the impulse noise removing apparatus of the present invention, the motion of the pixels on the upper and lower scanning lines in the same field of the pixels detected as the candidate of the impulse noise is detected to detect the motion. In that case, by excluding that pixel from candidates for impulse noise, it is possible to reduce image distortion due to false detection of impulse noise in a specific frequency region of a moving image to a level that does not visually disturb the image. It is possible to improve the image quality.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

FIG. 2 is a block diagram showing another embodiment of the present invention.

FIG. 3 is an explanatory diagram of a motion detection method of the present invention.

FIG. 4 is a block diagram showing a principle configuration for implementing the motion detection method of the present invention.

FIG. 5 is a block diagram showing a specific configuration of a noise reduction circuit.

FIG. 6 is a waveform diagram showing medaka noise generated in a video signal.

FIG. 7 is a block diagram showing a conventional noise removal circuit.

FIG. 8 is a block diagram showing a television receiver including a noise detection circuit.

FIG. 9 is a block diagram showing a schematic configuration of a noise reduction circuit.

FIG. 10 is an explanatory diagram of a method of detecting impulse noise in a still image.

FIG. 11 is a characteristic diagram showing a frequency region in which noise is detected.

[Explanation of symbols]

 101 Frame Memory 104 Absolute Value Circuit 106 Comparison Circuit 109-1, 109-2 Line Memory 111 NOR Circuit 113 AND Circuit

Claims (1)

[Claims] 1. A demodulation means for demodulating a modulated reception signal from a reception antenna to obtain a video signal including a detection signal and luminance information, and a noise component included in the detection signal is detected to obtain the noise component amount. Based on the determination of the presence or absence of impulse noise components included in the video signal large single-shot luminance change,
A noise detection determination unit that outputs a noise determination signal, and when the noise determination signal has the impulse noise component, the impulse noise component is detected by detecting the impulse portion in the video signal and performing the luminance interpolation processing on the detection portion. In addition to outputting the removed video signal, when there is no impulse noise component, noise detection and removal means for outputting the video signal from the demodulation means, a video signal containing the brightness information and the signal are delayed for a predetermined frame period. The frame difference component extracting means for extracting the frame difference component signal from the obtained signal and the level of the frame difference component signal and a predetermined reference signal are compared, and based on the comparison result, the upper and lower adjacent pixels in the same field are compared. Motion detection means for detecting the presence or absence of motion and outputting a motion detection signal; When there out only regarded as not even if the impulse noise component,
An impulse noise removing device, comprising: an impulse noise determining unit that outputs only the video signal from the demodulating unit as an output of the noise detecting and removing unit.