JP2001136416A - Signal processing unit and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a signal processing unit that can obtain an optimum noise reduction effect depending on a noise level in the case of reducing noise from a video signal. SOLUTION: The signal processing unit consists of a motion vector detection circuit 308 that detects a motion vector from an input video signal, a memory controller 306 that generates a motion correction control signal on the basis of the motion vector corresponding to the input video signal, a noise level detection circuit 310 that detects a noise level of the input video signal on the basis of the result of detection by the motion vector detection circuit 308, a nonlinear processing circuit 303 that applies nonlinear processing with the strength correspondent to the noise level and a 2nd sbutractor 302 that composites the signal subject to nonlinear processing and the input video signal.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a signal processing apparatus and method for performing noise reduction processing on a video signal.

[0002]

2. Description of the Related Art It is known in the field of digital video signal processing that a cyclic noise reduction apparatus using a frame memory reduces noise by performing averaging processing between fields. By extracting a low-level portion of the difference from the video signal one frame before from the memory as noise and subtracting the extracted noise component from the input video signal, noise is reduced and noise is reduced. One that writes a signal into a frame memory has been proposed. If a field memory is used instead of the frame memory, the capacity of the memory can be reduced.

FIG. 15 shows an example of a conventional noise reduction device. An input video signal Vin converted into a digital signal is supplied from an input terminal 201, and this input video signal Vin is supplied to a first subtractor 202 and a second subtractor 204.
Supplied to Output video signal Vou of first subtractor 202
t is taken out to the output terminal 207 and written into the frame memory 205. A memory controller 206 is provided in association with the frame memory 205. The memory controller 206 controls a write operation and a read operation of the frame memory 205.
The read data from the frame memory 205 is one frame delayed from the write data. When one frame delay is represented by F ⁻¹ , the output signal Vout · F ⁻¹ of the frame memory 205 is supplied to the second subtractor 204.
Since the input video signal Vin is supplied to the second subtractor 204, a frame difference is generated.

The output video signal of the second subtractor 204 is supplied to the first subtractor 202 via the nonlinear processing circuit 203. The non-linear processing circuit 203 multiplies the feedback coefficient K according to the level of the input signal, and is constituted by a ROM (read only memory). The input / output characteristics of the non-linear processing circuit 203 are such that the input (frame difference) is small
The input is output as a noise component with K = 1, the output is limited to a predetermined value in the range of the intermediate level, the output is reduced in the range of the large input, and when the input is large, the frame difference is , The output is set to 0 as it is caused by motion.

[0005] As described above, the nonlinear processing circuit 203 extracts the noise component by utilizing the characteristic that the correlation between frames is small and the amplitude is small. In the first subtractor 202, noise can be reduced by subtracting the extracted noise component from the input video signal.

The output video signal Vout from the noise reduction device
Is Can be represented by

As described above, in the conventional noise reduction apparatus, a small-amplitude component is extracted as noise from the difference between the input video signal and the video signal one frame before read from the frame memory, and is extracted from the input video signal. Noise is reduced by subtraction.

Further, in the conventional noise reduction device,
The video signal with the reduced noise is written to the frame memory and used for processing the next frame. If an autocorrelation between fields is used instead of a frame, a similar configuration can be made using a field memory instead of a frame memory.

[0009]

By the way, the above-described noise reduction device uses the same non-linear processing circuit irrespective of the input noise level and frequency distribution. There are problems that the noise reduction effect is low or the noise reduction effect in a certain frequency band is low. Conversely, when there is almost no noise, there is a problem that motion blur of a block or a field correlation in which a motion vector cannot be detected is conspicuous, or noise increases conversely.

Further, depending on the magnitude of the dynamic range of the input signal, the value of the high amplitude or the low amplitude of the video signal may stick to a certain value. When a signal stuck to such a fixed value exists, there is no noise in the stuck portion, so that it is erroneously detected that there is no noise in the original video signal. The present invention has been proposed in view of the above-described circumstances, and reduces noise even from a video signal having a large noise level, reduces noise from each frequency band, and achieves image quality even when there is almost no noise. It is an object of the present invention to provide a signal processing apparatus and method capable of reducing noise so that the signal does not deteriorate and preventing erroneous detection that there is no noise when a signal is stuck.

[0011]

In order to solve the above-mentioned problems, a signal processing apparatus according to the present invention comprises: a motion vector detecting means for detecting a motion vector from an input video signal; A motion compensator for performing motion compensation based on the motion vector detected by the motion vector detector, a difference unit for obtaining a difference signal between the signal motion-compensated by the motion compensator and the input video signal, A noise level detecting means for detecting a noise level of the input video signal based on a detection result of the vector detecting means, and a noise level detected by the noise level detecting means for the differential signal obtained by the differential means. A non-linear processing means for performing a non-linear processing of the same intensity, and a combination of the signal subjected to the non-linear processing by the non-linear processing means and the input video signal. And has a means.

Further, a signal processing device according to the present invention comprises a motion vector detecting means for detecting a motion vector from an input video signal, A motion correcting means for performing motion correction, a difference means for obtaining a difference signal between the signal corrected in motion by the motion correcting means and the input video signal, and converting the difference signal obtained by the difference means into a frequency band. Orthogonal transform means, a noise level detecting means for detecting a noise level of the input video signal based on a detection result in the motion vector detecting means, and a signal converted into a frequency band by the orthogonal transform means, A non-linear processing means for performing, for each frequency band, a non-linear processing having a strength corresponding to the noise level detected by the noise level detecting means; And it has a synthesizing means for synthesizing the non-linear processing is applied signal and the input video signal in the form processing unit.

In the signal processing method according to the present invention, a motion vector is detected from an input video signal, and the input video signal is subjected to motion correction based on the detected motion vector.
Obtaining a difference signal between the motion-compensated signal and the input video signal; detecting a noise level of the input video signal based on a detection result of the motion vector; Is performed, and the signal subjected to the non-linear processing is combined with the input video signal.

In the signal processing method according to the present invention, a motion vector is detected from an input video signal, and the input video signal is subjected to motion correction based on the detected motion vector. And a difference signal between the input video signal and the input video signal, the difference signal is converted to a frequency band, the noise level of the input video signal is detected based on the detection result of the motion vector detection means, and converted to the frequency band. A non-linear process having a strength corresponding to the noise level detected by the noise level detecting means for each frequency band, and synthesizing the signal subjected to the non-linear process and the input video signal. is there.

[0015] Here, it is determined whether or not the input signal is sticky for each image unit (field or frame) or for each block when the input image signal is divided into blocks. It is preferable to mask the input of noise level detection for each video unit (field or frame) or to use only blocks without sticking for noise level detection.

[0016]

Embodiments of the present invention will be described below with reference to the drawings. First, an example of an optical disk recording / reproducing apparatus to which the present invention can be applied will be described with reference to FIG.

In FIG. 1, an optical disk recording / reproducing apparatus 1
The recording system includes an A / D conversion circuit 102 to which an image signal is input from an input terminal 101, and an NTSC (National Television) to which image data is input from the A / D conversion circuit 102.
System Committee) A decoder 103 and a noise reduction circuit 104 to which an image is input from the NTSC decoder 103
And an MPEG (Moving Picture Experts Group) encoder 105 to which image data is input from the noise reduction circuit 104, an ECC (Error Correction Codes) encoder 106 to which image data is input from the MPEG encoder 105, and an image data from the ECC encoder 106 And an RF amplifier 108 to which image data is input from the 8-14 modulation circuit 107.
It is composed of

The A / D conversion circuit 102 has an input terminal 101
Receives an image signal of the NTSC system, and performs A / D conversion processing. The A / D conversion processing circuit 102
By performing the conversion process, an analog image signal is converted into digital image data. Then, this A / D conversion circuit 1
02 inputs the image data to the NTSC decoder 103.

The NTSC decoder 103 receives image data of the NTSC system from the A / D conversion circuit 102. The NTSC decoder 103 performs a decoding process on an NTSC composite signal. The NTSC decoder 103 performs a decoding process to convert the image data into a baseband signal (luminance signal, RY signal, BY signal).
Signal). The NTSC decoder 10
3 outputs the image data to the noise reduction circuit 104.

Image data is input from the NTSC decoder 103 to the noise reduction circuit 104. The noise reduction circuit 104 performs a process of reducing random noise included in the input video signal. This noise reduction circuit 104
Increases the image compression efficiency of the subsequent MPEG encoder 105 by performing a noise reduction process,
Improve motion compensation prediction accuracy. The noise reduction processing is performed on all components or only on the luminance signal. The noise reduction circuit 104 performs a noise reduction process on the image data sent from the NTSC decoder 103 by performing a filtering process. The noise reduction circuit 104 is connected to the control circuit 120 and operates according to a control signal sent from the control circuit 120. The control circuit 120 has a configuration of a microcomputer. And this noise reduction circuit 1
04 outputs the image data to the MPEG encoder 105. The noise reduction circuit 104 according to the present embodiment reduces noise by extracting a noise component excluding motion and selecting an optimal nonlinear circuit according to the noise level.

The MPEG encoder 105 performs motion compensation inter-frame prediction coding on the image data sent from the noise reduction circuit 104, and performs block DCT (Discrete Cosine Transformation) coding processing on the prediction error. Is applied. This MPEG encoder 1
05 performs an encoding process on the image data so that the MPE
It is assumed that the image data is of G system. At this time, the MPEG encoder 105 adds encoding information such as a quantization scale to the image data to form a bit stream. The MPEG encoder 105 converts the image data into E
Output to CC encoder 106.

The ECC encoder 106 adds redundant data of an error correction code to the bit stream sent from the MPEG encoder 105. Then, the ECC encoder 106 outputs the bit stream to the 8-14 modulation circuit 107.

The 8-14 modulation circuit 107 performs signal processing such as 8-14 modulation on the bit stream sent from the ECC encoder 106. The 8-14 modulation converts an 8-bit code into 14-bit data in order to reduce low frequency components of a recording signal. 8-14 modulation circuit 10
7 outputs the bit stream subjected to 8-14 modulation and other processing to the RF amplifier 108. RF amplifier 10
8 amplifies the bit stream sent from the 8-14 modulation circuit 107 and outputs the result to the optical pickup 109.

The recording system of the optical disc recording / reproducing apparatus 1 records a bit stream indicating an image on the optical disc 110 via the optical pickup 109. The optical disk 110 is a recordable, magneto-optical disk,
An optical disk such as a phase change disk can be used.

The reproducing system of the optical disk recording / reproducing apparatus 1 records a bit stream indicating an image from the image data recorded on the optical disk 110 via the optical pickup 109. The optical disk 110 is a recordable one, and an optical disk such as a magneto-optical disk or a phase change disk can be used.

The reproducing system of the optical disk recording / reproducing apparatus 1 includes an RF amplifier 111 to which image data recorded on the optical disk 110 is input via an optical pickup 109.
8-, image data is input from the RF amplifier 111
An ECC decoder 113 to which image data is input from the 14 demodulation circuit 112 and the 8-14 demodulation circuit 112, an MPEG decoder 114 to which image data is input from the ECC decoder 113, and a noise reduction circuit to which image data is input from the MPEG decoder 114 115 and the noise reduction circuit 1
15, an image quality correction circuit 116 to which image data with noise suppressed is input, an NTSC encoder 117 to which image-corrected image data is input from the image quality correction circuit 116, and NTSC image data from the NTSC encoder 117. And a D / A conversion circuit 118 to be inputted.

The RF amplifier 111 is connected to the optical pickup 10
The image data sent from the optical disk 110 detected in step 9 is subjected to amplification processing. Although not shown in FIG. 1, a tracking error signal and a focus error signal are generated in the RF amplifier 111 for tracking servo and focus servo. These tracking error signal and focus servo signal are supplied to the servo circuit. The RF amplifier 111 outputs the amplified image data to the 8-14 demodulation circuit 112.

The 8-14 demodulation circuit 112 is an RF amplifier 1
The image data sent from 11 is subjected to 8-14 demodulation processing. The 8-14 demodulation circuit 112 converts 14-bit data into an 8-bit code, contrary to the 8-14 modulation of the recording system. The 8-14 demodulation circuit 112 outputs the demodulated image data to the ECC decoder 113.

The ECC decoder 113 performs a decoding process on the image data sent from the 8-14 demodulation circuit 112. That is, an error included in the reproduction data is detected, and a correctable error is corrected. And this ECC
The decoder 113 outputs the error-corrected image data to the MPEG decoder 114. In MPEG, the coder 114 performs MPEG decoding and outputs a baseband signal. The MPEG decoder 114 outputs a baseband signal to the noise reduction circuit 115.

The noise reduction circuit 115 performs a noise reduction process on the image data sent from the MPEG decoder 114 by performing a filtering process. The noise reduction circuit 115 performs a noise reduction process to obtain M
Mosquito noise and block distortion caused by performing the decoding processing by the PEG decoder 114 are reduced. The noise reduction circuit 115 is connected to the control circuit 120, and is controlled according to a control signal sent from the control circuit 120. And this noise reduction circuit 1
At 15, the image data subjected to the noise reduction processing is output to the image quality correction circuit 116.

The noise reduction circuit 115 of the present embodiment
Noise is reduced by extracting a noise component excluding motion and selecting an optimal nonlinear circuit according to the noise level.

The image quality correction circuit 116 includes a noise reduction circuit 1
Image quality correction processing is performed on the image data sent from 15. The image quality correction circuit 116 performs, for example, an outline correction process or the like as the image quality correction process. The image quality correction circuit 116
Is connected to a control circuit 120 described later, and is controlled in accordance with a control signal sent from the control circuit 120. Then, the image quality correction circuit 116 outputs the image data subjected to the image quality correction processing to the NTSC encoder 117.

The NTSC encoder 117 subjects the image data sent from the image quality correction circuit 116 to processing such as adding a synchronization signal and modulating a color difference signal. The NTSC encoder 117 converts the image data into an NTSC decoded video signal by performing an encoding process. The NTSC encoder 117 converts the NTSC video signal into a D
/ A conversion circuit 118.

The D / A conversion circuit 118 performs a D / A conversion process on the image data sent from the NTSC encoder 117. The D / A conversion circuit 118 performs a D / A conversion process to generate an analog NTSC decoded video signal. Then, the D / A conversion circuit 118
The A-converted image signal is output to the output terminal 119.

In connection with the control circuit 120 for supplying a control signal to the above-described noise reduction circuit 104, the noise reduction circuit 115, and the image quality correction circuit 116, for example, an operation input which is operated by a user to supply an input signal to the control circuit 120 A part 121 is provided.

The control circuit 120 is composed of, for example, a microcomputer or the like, and supplies a control signal to the noise reduction circuit 104, the noise reduction circuit 115, and the image quality correction circuit 116 described above. The control circuit 120 supplies a control signal for reducing, for example, block distortion to the noise reduction circuit 115 according to the input signal transmitted from the operation input unit 121. Further, the control circuit 120 supplies a control circuit indicating whether or not to perform image quality correction and a control signal indicating the degree of image quality correction to the image quality correction circuit 116.

The operation input unit 121 generates and outputs an input signal by, for example, selectively pressing a switch or the like when a user or the like controls on / off of block distortion reduction. The operation input unit 121 is also provided with a switch or the like that allows the user to control the degree of image quality correction, and generates and outputs an input signal when the user selectively presses the switch.

In the present embodiment, the noise reduction circuit 104 and the noise reduction circuit 115
Is applied to reduce noise. Note that the present invention can be applied to a recording / reproducing apparatus using a recording medium other than an optical disk. further,
The present invention can also be applied to the case of communicating image data.

FIG. 2 shows a first embodiment of a noise reduction circuit to which the present invention is applied. The first embodiment of the noise reduction apparatus is a noise level adaptive field recursive noise reduction circuit provided with a noise level detection circuit 310 for detecting a noise level of an input video signal. In the noise level detection circuit 310, the optimal nonlinear processing circuit 303 is automatically selected according to the detected noise level, so that the feedback amount can be controlled according to the strength of the input noise.

This noise reduction device includes a first field memory 309 for storing a video signal input from an input terminal 301 for one field, and a video signal sent from a terminal 301 and a first field memory 309 for storing one field. It has a motion vector detection circuit 308 for detecting a motion vector from a video signal delayed by + β.

The input terminal 301 is supplied with a video signal converted to a digital signal. Then, the input video signal Vin is written into the first field memory 309 and is input to the motion vector detection circuit 308 at the same time. In the motion vector detection circuit 308, the input terminal 301
And the first field memory 30
Then, a motion vector of each block is obtained from 1 field read from 9 and a video signal delayed by β.

Further, the noise reduction device has an output terminal 307.
Field memory 305 for storing one field of the video signal output from
8, a memory controller 30 that controls the second field memory 305 based on the motion vector detected in
6.

The memory controller 306 receives a motion vector and the like obtained from the motion vector detection circuit 308 and outputs the motion vector as a motion correction control signal.
6 is delayed by one field before being output. The video signal is input to the first field memory 309 in order to match the delay of one field of the motion correction control signal, and is read out with a delay of one field + β. This one
In accordance with the video signal delayed by field + β, the second field memory 305 outputs one field + α according to the motion compensation control signal sent from the memory controller 306.
The delayed video signal is read.

Further, the noise reduction device includes a first subtractor 304 for subtracting the video signal sent from the second video signal 305 from the video signal sent from the first field memory 309 to obtain a difference signal. , A non-linear processing circuit 30 that performs non-linear processing on the difference signal sent from the first subtractor 304
3 and a second subtractor 302 that combines the video signal sent from the first field memory 309 by subtracting the video signal sent from the non-linear processing circuit 303 from the video signal sent from the first field memory 309.

In the first subtractor 304, the video signal delayed by 1 field + β delayed from the first field memory 309 and the video signal delayed by 1 field + α delayed from the second field memory 305 are output. Obtain a field difference signal.

The field difference signal is subjected to nonlinear processing in a nonlinear processing circuit 303 to extract a noise component. The signal processed by the non-linear processing circuit 303 is converted by the second subtractor 302 into the first field memory 3.
09 and the output terminal 307
Output from

The noise reduction device has a noise level detection circuit 310 which detects a noise level based on the detection result of the motion vector detection circuit 308 and controls the nonlinear processing circuit 303 according to the noise level. .

To the noise level detection circuit 310, the minimum value of the difference of the block matching of the motion vector for each block obtained from the motion vector detection circuit 308 and the validity / invalidity determination of the motion vector detection are input.

The noise level detection circuit 310 measures an average noise level of the video signal, converts the average noise level into an N-level limiter level, and outputs the converted level to the non-linear processing circuit 303. The processing in the noise level detection circuit 310 will be further described later.

In the non-linear processing circuit 303, a non-linear process corresponding to the N-stage limiter level input from the noise level detection circuit 310 is selected from N processes, and the non-linear process is performed. The characteristics of the N nonlinear processes are realized by a plurality of memory (ROM or RAM) tables.

The noise extracted as noise by being processed by the nonlinear processing circuit 303 is subtracted from the input video signal by the second subtractor 302. In this way, the noise level adaptive field cyclic noise reduction circuit 104 that reduces noise from the input video signal is configured.

Next, in the noise level detection circuit 310, an average noise level is measured from the minimum value of the difference of the block matching for each block and whether the detection is valid or invalid, and the limiter level corresponding to the noise level is measured. The output method will be described with reference to FIG.

The valid / invalid judgment result 401 for each block obtained from the motion vector detecting circuit 308 is input to the minimum value detecting section 403. Then, the minimum value of the minimum value of the difference of the block matching of the motion vector is obtained using only the minimum value of the difference of the block matching of the block determined to be valid.

By using only the minimum value of the difference of the block matching of the motion vector for which the motion vector detection is determined to be valid, it is possible to extract the difference of only the noise excluding the difference due to the motion.

As the detection of the minimum value in the minimum value detection unit 403, one minimum value may be obtained in the field, an average of a plurality of minimum values may be obtained, or each divided value in the field may be obtained. The average of the minimum values in the area may be taken.

The noise extracted by the minimum value detector 403 is input to the filter 405. The filter 405 is, for example, an infinite impulse response (infi
niteimpulse response; IIR).

The filter 405 shown in FIG. 4 includes a first amplifier 503 for amplifying the noise level 501 by 1-K times,
A delay unit 506 that delays the average noise level 507 by a unit time, a second amplifier 505 that amplifies the output from the delay unit 506 by K times, an output from the first amplifier 503, and a signal from the second amplifier 505 And an adder 504 for adding the outputs of the above to obtain an average noise level. In addition,
The first amplifier 503 can be regarded as a coefficient multiplier for multiplying the coefficient 1-K, and the second amplifier 505 can be regarded as a coefficient multiplier for multiplying the coefficient K.

Here, the amplification factor 1-K of the first amplifier 503 and the amplification factor K of the second amplifier 505 depend on the accuracy 502 of the noise level detection. The relationship between the filter coefficient K and the accuracy 502 of the noise level detection will be further described later. The filter 4 used here
05 is not limited to the IIR of the circuit shown in FIG. 4, but may be an IIR having a different transfer function or a median filter.

The noise level accuracy determined by the noise level detection accuracy determination unit 404 is input to the filter 405, and the value of the coefficient K of the first amplifier 503 and the second amplifier 505 is adaptively changed according to the accuracy. To change.

FIG. 5 shows the relationship between the accuracy of the noise level and the coefficient K. As shown in FIG. 5, when the accuracy of the noise level detection is 0, the filter coefficient K is 1. As the accuracy of noise level detection increases, the filter coefficient K monotonously decreases. When the accuracy of the noise level detection is 1, the filter coefficient K is almost 0.

The lower the accuracy of the noise level, the more
The information in that field is suppressed. The validity / invalidity determination result 401 for each block is input to the noise level detection accuracy determination unit 404, the number of valid blocks is counted, and accuracy = the number of valid blocks / the total number of blocks is output as the noise level detection accuracy. I do. The method for determining the accuracy of the noise level is not limited to this method.

The noise level output from the filter 405 is input to the limiter level converter 406, and is converted into a limiter level having N levels of strength.

As described above, in the first embodiment of the noise reduction device, the average noise level of the input noise is extracted, and the optimum nonlinear processing circuit corresponding to the extracted noise level is extracted. It is automatically selected and noise reduction is performed according to the strength of the input noise.

Next, a description will be given of a second embodiment of a noise reduction apparatus in which an orthogonal transform such as a Hadamard transform is used to extract noise. Second
Is a motion adaptive field recursive noise reduction apparatus that automatically measures the noise level of each frequency component and adaptively controls the feedback amount for each frequency component according to the noise level.

Here, the configuration of a noise reduction device (noise reducer) as a premise of the second embodiment of the noise reduction device will be described with reference to FIG. This noise reduction apparatus uses a frame memory that forms a recursive filter, performs an averaging process between fields, and reduces noise, which is conventionally known in the field of digital image signal processing.

The noise reduction apparatus, which is a premise of the present embodiment, further performs motion compensation in order to avoid motion blur when an image has motion and to remove independent noise components for each frequency component. Is obtained by dividing the difference signal from the current signal obtained in step (1) into a plurality of frequency components by orthogonal transform, performing nonlinear processing on each of them, and performing inverse orthogonal transform to obtain noise components.

A video signal converted into a digital signal is supplied from an input terminal 1 and this input video signal Vin is
The data is written to the first field memory 9 and is simultaneously input to the motion vector detection circuit 8. From the first field memory 9, a video signal delayed by about one field is read. To be precise, the video signal read from the first field memory 9 is further delayed by β in addition to one field. This time β is used for timing adjustment and the like.

The motion vector detecting circuit 8 obtains a motion vector of a block from the video signal input from the terminal 1 and the video signal read from the first field memory 9.

The motion vector and the like obtained from the motion vector detection circuit 8 are input to the memory controller 6 and are delayed by about one field until output from the memory controller 6 as a motion correction control signal.

The video signal output from the output terminal 7 is written to the second field memory 5 on a field basis. The video signal written in the second field memory 5 is read out about one field later than the video signal from the first field memory 9, and the first subtractor 4 outputs the first field memory 9. The difference from the video signal from is obtained.

That is, the video signal subjected to the motion compensation processing is read out from the second field memory in accordance with the video signal from the first field memory 9. The first subtracter 4 obtains a field difference signal from the video signal read from the first field memory 9 and the video signal read from the second field memory 5.

The memory controller 6 controls reading of the video signal from the second field memory 5 based on the motion vector detected by the motion vector detection circuit 8 so as to compensate for the motion of the input video signal.

Incidentally, in the second field memory 5, to be precise, the video signal is further delayed by a time α in addition to one field. This time α is used for timing adjustment and the like.

The field difference signal obtained by the first subtractor 4 is input to the Hadamard transform circuit 10. The Hadamard transform circuit 10 performs a Hadamard transform and performs an orthogonal transform for dividing the field difference signal into predetermined frequency components.

Each frequency component passes through the nonlinear processing circuit 13, and a noise component for each frequency band is extracted. The nonlinear processing circuit 13 multiplies the feedback coefficient K according to the level of each frequency band of the input signal.

The input / output characteristics of the non-linear processing circuit 13 are as follows. In the range where the input (field difference) is small, the input is output as a noise component with K = 1, and in the range where the input is an intermediate level, the output is limited to a predetermined value. In the range where the input is large, the output is reduced, and when the input is large,
The output of the field difference is assumed to be 0 due to motion.

The noise extracted by the non-linear processing in the non-linear processing circuit 13 is returned to a time-dependent signal by the inverse Hadamard transform circuit 11. Then, the second subtracter 2 combines the video signals sent from the first field memory 9 so as to remove noise.

This noise reduction device can reduce motion blur when there are a plurality of motions in a screen by using motion correction and orthogonal transformation.

The second embodiment of the noise reduction device is based on such a noise reduction device, by extracting a noise component excluding motion and selecting an optimal nonlinear processing circuit according to the noise level. , To reduce noise. That is, the present embodiment detects the noise level of the input video signal in the above-described noise reduction device,
The nonlinear processing is adaptively applied according to the noise level.

FIG. 7 is a block diagram of a second embodiment of the noise reduction device. In the second embodiment of the noise reduction device, the configuration in the first embodiment shown in FIG. 3 is such that the motion adaptive processing is performed and the average noise level is obtained by the average noise level detection circuit 712. Is the same as

This noise reduction device includes a first field memory 709 for storing a video signal input from an input terminal 701 for one field, and a video signal sent from the input terminal 701 and a first field memory 709 for storing the video signal. It has a motion vector detection circuit 708 for detecting a motion vector from a video signal delayed by + field.

Further, the noise reduction device has an output terminal 707.
A second field memory 705 for storing one field of the video signal output from the
8, a memory controller 706 that controls the second field memory 705 based on the motion vector detected.
And The second field memory 705 stores 1
The video signal delayed by + field is output.

Further, the noise reduction device comprises a first subtractor 704 for subtracting the video signal sent from the second field memory 705 from the video signal sent from the first field memory 709, Transform circuit 710 for transforming the difference signal sent from the device 704 into a frequency band by Hadamard transform, and a Hadamard transform circuit 71
And a non-linear processing circuit 703 for performing non-linear processing on the signal transmitted from 0 for each frequency band.

From the second field memory 705, a video signal delayed by one field + α is output based on the motion correction control signal from the memory controller 706 based on the motion vector obtained by the motion vector detection circuit 708. You.

In the first subtractor 704, one field + α output from the second field memory 705
The delayed video signal and the first field memory 709
And a field difference signal from the video signal delayed by 1 field + .beta. This field difference signal is supplied from a first subtractor 704 to a Hadamard transform circuit 710.
Is input to

The Hadamard transform circuit 710 performs a Hadamard transform process to divide the field difference signal into predetermined frequency components.

The non-linear processing circuit 703 operates in accordance with the limiter level of the frequency-specific noise level detection circuit 713,
Can be performed. The non-linear processing circuit 703 selects a non-linear process corresponding to the N-level limiter level input from the frequency-specific noise level detection circuit 713 from the N processes and performs the non-linear process.

The noise reduction device includes an inverse Hadamard transform circuit 711 for performing an inverse Hadamard transform on the signal in the frequency band subjected to the nonlinear processing by the nonlinear processing circuit 703 to convert the signal into real time, and a first field memory. And a second subtractor 702 for subtracting the video signal sent from the inverse Hadamard transform circuit 711 from the video signal sent from the 309.

The noise component for each frequency component that has passed through the nonlinear processing circuit 703 and is extracted as noise is returned by the inverse Hadamard transform circuit 711 to a signal on the time axis. The output signal of the inverse Hadamard transform circuit 711 is subtracted by a subtractor 702
And is subtracted from the input video signal by the second subtractor 702.

Then, the noise reduction device is sent from the average noise level detection circuit 712 for detecting the average noise level based on the detection result of the motion vector 708, and from the average noise level detection circuit 712 and the Hadamard conversion circuit 710. And a noise level detection circuit 713 for controlling the nonlinear processing circuit 703 by detecting a noise level for each frequency from the signal of the frequency band.

The field difference signal for each frequency component output from the Hadamard transform circuit 710 is input to the frequency-specific noise level detection circuit 713. The average noise level obtained by the average noise level detection circuit 712 is input to the noise level detection circuit 713 for each frequency.
From the motion vector detection circuit 708, the valid / invalid determination result of the detection for each block is input. Then, the noise level of each frequency component is measured, converted to an N-level limiter level, output, and input to the nonlinear processing circuit 703.

That is, the frequency-specific noise level detection circuit 713 re-extracts a block having only the noise level again using the noise level detected by the average noise level detection circuit 712 as a reference, and performs a Hadamard transform on the extracted block. Is calculated for each frequency.

On the other hand, the average noise level detection circuit 712 extracts a block containing only a noise component and detects one noise level in the field from the difference value.

Next, a frequency-specific noise level detection circuit 71
3, a method of outputting a limiter level according to the noise level of each frequency band will be described with reference to FIG.

The absolute value processing section 805 performs absolute value processing on the field difference signal 803 obtained from the Hadamard transform circuit 710.

The averaging section 806 is a block determined to be valid from the valid / invalid determination result 801 for each block obtained from the motion vector detection circuit 708 and obtained from the average noise level detection circuit 713. The value obtained by adding the offset α _F to the average noise level 802 is added to only the field difference signal of only the block having the smallest minimum value of the block matching difference obtained from the motion vector detection circuit 708, and the average value is calculated. Output. As a result, the noise component of only the block having the average noise component is output for each frequency component.

Here, the reason why the offset α _F is added to the average noise level 802 is that the magnitude of the noise is not constant but includes a variation, so that the value of the difference value due to the noise depends on the position of the matched block. Is different. α _F is a variation of the noise component, and the variation is proportional to the magnitude of the noise. For example, if A is a constant, then α _F = A × average noise level.

The noise extracted by the averaging unit 806 is input to the filter 807 for each frequency component, and the same processing as the filter 405 for performing the IIR is performed. The noise level accuracy 804 determined by the noise level accuracy determination unit of the average noise level detection circuit 712 is input to the filter 807, and the characteristics of the filter change in the same manner as the filter 405 according to the accuracy.

The noise level output from the filter 807 is input to a limiter level converter 808, where it is converted to a limiter level 809 having N levels of strength.

Here, an example in which the method of outputting the limiter level is applied to the second embodiment of the noise reduction device has been described, but the same applies to the first embodiment of the noise reduction device. A limiter level can be output using the configuration. In the second embodiment of the noise reduction device, the input noise is extracted for each frequency band, the optimal nonlinear processing circuit corresponding to each noise level is automatically selected, and the noise intensity of each frequency band is increased. Noise reduction according to the noise level is performed.

In the second embodiment of the noise reduction device, after noise extraction, temporal filtering is performed to smooth temporal fluctuations in the noise level.

Next, another example of the noise level detecting circuit will be described with reference to FIG. This noise level detection circuit is applied to the noise level detection circuit 310 of the noise reduction device shown in FIG. 3 and the average noise level detection circuit 712 of the noise reduction device shown in FIG.

This noise level detection circuit uses the fact that the difference value of the vector (0, 0) by block matching in an area where the contrast of an image is small is a difference due to a noise component. Here, the vector (0, 0) is a zero vector for block matching in the search range in the (x, y) direction. For example, the average value 901 of the difference values of the block matching obtained from the motion vector detection circuit 308 and the vector (0,
The difference value 902 of the difference value of 0) is input to the minimum value detection unit 903. The minimum value detection unit 903 obtains the minimum value of the difference value of the vector (0, 0) from blocks satisfying the condition that the average value is smaller than a certain threshold value, that is, the image contrast is small.

The minimum value is output from the motion vector detection circuit 308 as the noise level in this field. As the detection of the minimum value in the minimum value detection unit 903, one minimum value may be obtained in a field, a plurality of minimum values may be obtained, or each area divided in the field may be detected. The average of the minimum values may be taken.

The relationship between the noise detection method and the noise level detection circuit shown in FIG. 3 is as follows. That is, the minimum value of both results may be taken.

When the number of effective regions of the motion vector is large, the detection method using the effective region of the motion vector detection result is performed, and when the number of effective regions is small, the detection method using a block with small contrast is used. You may go.

Further, when the number of blocks with low contrast is large, the detection method using blocks with low contrast may be used, and conversely, the detection method using the effective area of the motion vector detection result may be used. .

Next, the frequency-specific noise level detection circuit 71
3, another method of outputting a limiter level according to the noise level of each frequency band will be described with reference to FIG.

This method uses the fact that the difference value of the vector (0, 0) in an area where the contrast of an image is small is a difference due to a noise component.

The averaging unit 1005 includes an average value 1001 of the difference value of the block matching for each block obtained from the motion vector detection circuit 708, a difference value 1002 of the vector (0,0) of the block matching, and an average noise. The average noise level 1000 obtained from the level detection circuit 712 is input. A vector (0, 0) is obtained from a value obtained by adding the variation α of the noise level to the average noise level 1000.
0), the Hadamard-transformed field difference signal 1003 of only the block in which the difference value 1002 is small and the average value 1001 of the difference values is smaller than the threshold value is added for each frequency component, and an average is obtained. .

The value of the threshold is a value corresponding to the value of the average noise level 1000. For example, average noise level × motion vector search range (x × y) ×
It is a constant A.

The relationship with the noise level detection method for each frequency band shown in FIG. 8 is as follows. That is, addition and averaging may be performed for all blocks determined to be valid by both methods.

Further, these methods may be switched for each field depending on whether the number of blocks obtained by the detection method using the effective area of the motion vector detection result is large or small.

Further, these methods may be switched depending on whether the number of blocks obtained by the detection method using blocks with small contrast is large or small. As described above, in the present embodiment, the field recursive noise reduction circuit extracts a high-precision noise component excluding motion and automatically selects an optimal nonlinear processing circuit according to the noise level. The noise reduction effect can be automatically adjusted, and highly accurate adjustment can be performed.

Next, as a third embodiment of the present invention,
With reference to FIG. 11, a description will be given of a noise reduction device provided with a sticking detection for each video unit (field or frame) with respect to the noise level detection circuit in the noise reduction circuit. Since the overall configuration of the noise reduction device according to the third embodiment is similar to that of FIGS. 2 and 7 described above,
It is not shown and the description is omitted.

In the noise reduction device according to the third embodiment, the noise level detection circuit 310 shown in FIG.
The configuration shown in FIG. 11 is used as the average noise level detection circuit 712 and the frequency-specific noise level detection circuit 713 in FIG.

In FIG. 11, the valid / invalid judgment result 321 for each block obtained from the motion vector detection circuit 308 in FIG. 2 or the motion vector detection circuit 708 in FIG. 322 is input to the minimum value detection unit 325. Minimum value detector 3
In step 25, for example, the minimum value is further obtained using only the minimum value of the difference between the blocks determined to be valid in the motion vector detection. The minimum value output from the minimum value detection unit 325, that is, the average noise level, is input to the sticking determination circuit 330, and determines whether the sticking is present in one image unit, for example, in a field.

[0118] as an algorithm in the beam with the decision circuit 330, when the average noise level is smaller than a predetermined threshold value alpha _H, is intended to be there sticking, which, since it is taking field differential, static Even for pictures, the average noise level takes into account some non-zero value. That this determination condition is, for example, the average noise level <threshold alpha _H: sticking There average noise level ≧ threshold alpha _H: can be expressed as free sticking.

The sticking presence / absence signal obtained by the sticking determination circuit 330 is output from the filter 327.
Sent to The filter 327 is connected to the minimum value detection unit 32
The filter control signal is used to filter the average noise level from the signal No. 5 and the noise level determined by the noise level detection accuracy determination unit 326 as the filter presence / absence signal from the filter determination circuit 330. The accuracy is entered. An example of this filter 327 is shown in FIG.
Shown in

In FIG. 12, the minimum value detecting section 325
Are sent to a coefficient multiplier 423, and an output from the coefficient multiplier 423 is taken out as a noise level 427 via an adder 424. The noise level 427 is sent to the adder 242 via the unit delay element 426 and the coefficient multiplier 425, and is added to the output from the coefficient multiplier 423. Here, the multiplication coefficient 1-K of the coefficient multiplier 423 and the multiplication coefficient K of the coefficient multiplier 425 are determined by the accuracy determination unit 326 of the noise level detection.
Is adaptively variably controlled by the accuracy 422 of the noise level and the signal 429 indicating the presence / absence of attachment from the attachment determination circuit 330. Specifically, for example, the relationship between the noise level accuracy 422 and the coefficient K is set such that the filter coefficient K monotonically decreases as the accuracy increases, as shown in FIG. When the sticking is determined to be present by the signal 429, the coefficient K is forcibly set to 1 and the coefficient multiplier 423
Of the average noise level 42
1 is not input into the filter (to mask the input signal).

Returning again to FIG. 11, the noise level 427 output from the filter 327 is sent to the limiter level converter 328, and this limiter level converter 32
8, N is determined according to the noise level from the filter 327.
The limiter level is converted to a step limiter level, and the limiter level 329 is extracted.

Next, as a fourth embodiment of the present invention,
With reference to FIG. 13, a description will be given of a noise reduction device that determines whether or not there is sticking in a block. In the noise reduction device according to the fourth embodiment, the noise level detection circuit 310 shown in FIG. 2 and the average noise level detection circuit 712 and the frequency-based noise level detection circuit 713 shown in FIG. Configuration is used. Note that the overall configuration of the noise reduction device according to the fourth embodiment is the same as that of FIG. 2 and FIG.

In FIG. 13, the valid / invalid judgment result 521 for each block obtained from the motion vector detection circuit 308 in FIG. 2 or the motion vector detection circuit 708 in FIG. 522 is input to the minimum value detection unit 525. The minimum value 522 of the difference of the motion vector detection is determined by the sticking determination circuit 5.
30, and the sticking determination circuit 530
The presence / absence of sticking of each block is determined based on the minimum value 522 of the difference in motion vector detection, and the determination result is sent to the minimum value detection unit 525.

The algorithm in the sticking determination circuit 530 is as follows.
When 22 is smaller than a predetermined threshold value β _H, it is determined that there is sticking. That is, the determination condition, the minimum value of the difference of the motion vector detection <threshold beta _H: sticking There minimum difference value ≧ the threshold of the motion vector detection beta _H: can be expressed as free sticking. This takes into account that the minimum value of the difference in the motion vector detection takes some non-zero value in the state without sticking.

The minimum value detection unit 525 uses only the minimum value of the difference in the motion vector detection of the block determined to be valid in the motion vector detection and has no sticking, for example, by using the minimum value. Ask for. The minimum value output from the minimum value detection unit 525, that is, the average noise level is sent to the filter 507.
FIG. 14 shows a specific configuration example of the filter 507.

In FIG. 14, the minimum value detector 5
25 from the coefficient multiplier 62
3 and the output from the coefficient multiplier 623 is
4 and is extracted as a noise level 627. This noise level 627 passes through the unit delay element 626,
The signal is sent to the adder 642 via the coefficient multiplier 625 and added to the output from the coefficient multiplier 623. Here, the multiplication coefficient 1-K of the coefficient multiplier 623, the coefficient multiplier 625
Of the noise level detection accuracy determination unit 6
26 is adaptively variably controlled based on the accuracy 622 of the noise level. Specifically, for example,
As shown in FIG. 5, the relationship between the noise level accuracy 622 and the coefficient K is such that the filter coefficient K monotonously decreases as the accuracy increases.

Returning to FIG. 13 again, the noise level 627 output from the filter 527 is sent to the limiter level converter 528, and this limiter level converter 52
8, N is determined according to the noise level from the filter 527.
The limiter level is converted to a step limiter level, and the limiter level 529 is extracted.

According to the third and fourth embodiments, the circuit for detecting the presence of the sticking of the input signal (sticking determination circuits 330 and 530) is added, whereby the sticking of the image signal is achieved. It is possible to reduce noise erroneous detection and realize a good automatic optimization noise reduction device.

In the first to fourth embodiments of the present invention, a recursive noise reduction device (noise reducer) for reducing noise in a digital video signal and an image on an optical disk or The present invention can be applied to recording / reproduction on a recording medium such as a magnetic tape, transfer of an image from a transmission side to a reception side via a transmission path, such as a video conference system, a video telephone system, and broadcasting equipment. Further, since the embodiment of the present invention uses a plurality of nonlinear processing circuits according to the input noise level and frequency distribution, even if a video signal with a large noise level is input, the noise is not affected in each frequency band. Can be effectively reduced.

That is, in the present embodiment, the same nonlinear processing circuit is used irrespective of the input noise level and frequency distribution. The problem that the noise reduction effect is low even when a video signal is input or the noise reduction effect in a certain frequency band is low is reduced.

In the present embodiment, when there is almost no noise, which is observed in the noise reduction apparatus which is the premise of the present embodiment, a block in which a motion vector cannot be detected or a block having a low field correlation is used. Problems such as noticeable motion blur and increased noise have been reduced.

In the present embodiment, the video signal is processed in units of fields, but the present invention is not limited to this. The present invention can also process a video signal on a frame basis.

In the present embodiment, the Hadamard transform is shown as an example of the orthogonal transform, but the present invention is not limited to this. The present invention may utilize other orthogonal transforms.

[0134]

According to the present invention, a high-precision noise component without motion is extracted from a field or inter-frame difference signal, and an optimal nonlinear processing circuit according to the noise level is automatically selected. In addition, the adjustment of the noise reduction effect can be automatically performed, and the adjustment can be performed with high accuracy.

According to the present invention, a field or inter-frame difference is divided into a plurality of frequency band components, a high-precision noise component excluding motion is extracted from each frequency band component, and the frequency component is extracted according to each noise level. By automatically selecting the optimal nonlinear processing circuit, the noise reduction effect according to the noise distribution can be automatically performed.

According to the present invention, after noise extraction, adaptive filter processing for changing the filter coefficient according to the noise level detection accuracy is performed, so that temporally smooth noise reduction can be performed.

Further, according to the present invention, when an input signal is stuck, the signal is not used for noise level detection, so that erroneous noise detection can be prevented, and good automatic optimization noise reduction can be achieved. Can be realized.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a schematic configuration of a recording / reproducing apparatus.

FIG. 2 is a block diagram showing a configuration of a first embodiment of a noise reduction device.

FIG. 3 is a block diagram illustrating a noise level detection circuit.

FIG. 4 is a circuit diagram illustrating filter processing in a noise level detection circuit.

FIG. 5 shows a filter coefficient K in a noise level detection circuit.
FIG. 6 is a diagram showing a relationship between the noise level and the accuracy of the noise level.

FIG. 6 is a block diagram illustrating a configuration that is a premise of a second embodiment of the noise reduction device.

FIG. 7 is a block diagram illustrating a configuration of a second embodiment of the noise reduction device.

FIG. 8 is a block diagram illustrating a limiter level in the noise reduction device.

FIG. 9 is a diagram illustrating another example of the noise level detection circuit.

FIG. 10 is a diagram showing another method for detecting a noise level for each frequency band.

FIG. 11 is a block diagram illustrating an example of a noise level detection circuit that is a main part of a noise reduction device according to a third embodiment of the present invention.

12 is a block diagram illustrating a specific configuration example of a filter in the noise level detection circuit in FIG.

FIG. 13 is a block diagram illustrating an example of a noise level detection circuit that is a main part of a noise reduction device according to a fourth embodiment of the present invention.

14 is a block diagram illustrating a specific configuration example of a filter in the noise level detection circuit in FIG.

FIG. 15 is a block diagram illustrating a configuration of a conventional noise reduction device.

[Explanation of symbols]

303,703 Non-linear processing circuit, 305,705
Second field memory, 306, 706 memory controller, 308, 708 motion vector detection circuit, 309, 709 first field memory, 3
10 noise level detection circuit, 325, 403, 52
5,903 minimum value detector, 326,404,526
Noise level detection accuracy determination unit, 327, 405,
527,905 filters, 328,406,52
8,906 Limiter level converter 330,530
Sticking judgment circuit

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Mami Tomita 6-35, Kita-Shinagawa, Shinagawa-ku, Tokyo Sony Corporation F-term (reference) 5C021 XA01 XA43 XB12 YA02 YC00 5C053 FA23 FA27 GA11 GB19 GB37 HA09 KA03 KA11 KA16 KA22 KA24 KA25

Claims (28)

[Claims] A motion vector detecting means for detecting a motion vector from an input video signal; a motion correcting means for performing a motion correction on the input video signal based on a motion vector detected by the motion vector detecting means; Difference means for obtaining a difference signal between the signal corrected in motion by the motion correction means and the input video signal; noise level detection for detecting a noise level of the input video signal based on a detection result in the motion vector detection means Means, nonlinear processing means for performing nonlinear processing on the difference signal obtained by the difference means according to the noise level detected by the noise level detection means, and nonlinear processing by the nonlinear processing means. A signal processing device comprising: synthesizing means for synthesizing the obtained signal and the input video signal. 2. The motion vector detection means detects a motion vector between video units constituting the input video signal, and the motion correction means detects a motion vector based on the motion vector detected by the motion vector detection means. 2. The signal processing apparatus according to claim 1, wherein a moving object is extracted from the input video signal before the image unit, and the extracted moving object is subjected to motion correction using the motion vector. 3. The signal processing apparatus according to claim 2, wherein the video unit is a frame or a field. 4. The signal processing apparatus according to claim 1, wherein the motion vector detecting means divides the input video signal into blocks, and detects a motion vector for each block. 5. The noise level detection means, comprising: an average level detection means for detecting an average noise level based on a motion vector detected for each block; and an average noise level detected by the average level detection means. Filter means for performing a filter process for smoothing a temporal variation of, and strength conversion means for converting the noise level smoothed by the filter means to the strength of the nonlinear processing in the nonlinear processing means. The signal processing device according to claim 4, wherein 6. The signal processing apparatus according to claim 5, wherein said filter means adapts characteristics to the accuracy of noise detection in said noise level detection means. 7. The noise level detecting means includes: an average level detecting means for detecting an average noise level based on a difference value of a motion vector detected for each block; A sticking determination means for determining the presence or absence of signal sticking within one video unit from the noise level of the image signal; and smoothing a temporal variation of an average noise level from the average level detecting means. 5. The signal processing apparatus according to claim 4, further comprising: a filter unit that performs a filter process for masking an input when the determination is made. 8. The sticking determination unit, wherein the noise level detecting unit determines whether there is signal sticking for each block based on a minimum value of a difference between motion vectors detected for each block, and the sticking determination. Means for detecting an average noise level based on a motion vector detected for each block for a block determined to have no sticking by the means; and a time-dependent average noise level detected by the average level detecting means. 5. The signal processing device according to claim 4, further comprising: a filter unit that performs a filter process for smoothing a fluctuation. 9. A motion vector detecting means for detecting a motion vector from an input video signal, a motion correcting means for performing a motion correction on the input video signal based on a motion vector detected by the motion vector detecting means, A difference means for obtaining a difference signal between the signal motion-compensated by the motion compensation means and the input video signal; an orthogonal transformation means for transforming the difference signal obtained by the difference means into a frequency band; A noise level detecting means for detecting a noise level of the input video signal based on a detection result by the means; and a noise level detected by the noise level detecting means for a signal converted into a frequency band by the orthogonal transform means. Nonlinear processing means for performing nonlinear processing of strength corresponding to each of the frequency bands, and nonlinear processing performed by the nonlinear processing means. A signal processing device comprising: synthesizing means for synthesizing the input signal and the input signal. 10. The motion vector detection means detects a motion vector between video units constituting the input video signal, and the motion correction means detects a motion vector based on the motion vector detected by the motion vector detection means. 10. The signal processing apparatus according to claim 9, wherein a moving object is extracted from the input video signal before the video unit, and the extracted moving object is subjected to motion correction using the motion vector. 11. The signal processing apparatus according to claim 10, wherein the video unit is a frame or a field. 12. The signal processing apparatus according to claim 9, wherein said motion vector detecting means divides the input video signal into blocks and detects a motion vector for each block. 13. The average noise level detecting means for detecting an average noise level based on a motion vector detected for each block, the average noise level detected by the average level detecting means. Filter means for performing a filter process for smoothing a temporal variation of, and strength conversion means for converting the noise level smoothed by the filter means to the strength of the nonlinear processing in the nonlinear processing means. 13. The signal processing device according to claim 12, wherein: 14. The signal processing apparatus according to claim 13, wherein said filter means adapts characteristics to the accuracy of noise detection in said noise level detection means. 15. The noise level detecting means, comprising: an average level detecting means for detecting an average noise level based on a difference value between motion vectors detected for each block; and an average level detected by the average level detecting means. A sticking determination means for determining the presence or absence of signal sticking within one video unit from the noise level of the image signal; and smoothing a temporal variation of an average noise level from the average level detecting means. 13. The signal processing apparatus according to claim 12, further comprising: a filter unit configured to perform a filter process for masking an input when the determination is made. 16. The sticking determination unit, wherein the noise level detecting unit determines whether or not there is signal sticking for each block based on a minimum value of a difference between motion vectors detected for each block; Means for detecting an average noise level based on a motion vector detected for each block for a block determined to have no sticking by the means; and a time-dependent average noise level detected by the average level detecting means. 13. The signal processing apparatus according to claim 12, further comprising: a filter unit configured to perform a filter process for smoothing a fluctuation. 17. A motion vector is detected from an input video signal, motion correction is performed on the input video signal based on the detected motion vector, and a difference between the motion-corrected signal and the input video signal is detected. Obtaining a signal, detecting a noise level of the input video signal based on the detection result of the motion vector, performing a nonlinear process on the difference signal in accordance with the detected noise level, A signal processing method comprising combining the applied signal and the input video signal. 18. A motion vector between video units constituting the input video signal is detected, and a moving object is extracted from the input video signal one video unit before based on the detected motion vector. 18. The signal processing method according to claim 17, wherein the motion vector is corrected using the motion vector. 19. The signal processing method according to claim 18, wherein the video unit is a frame or a field. 20. The signal processing method according to claim 17, wherein the input video signal is divided into blocks, and a motion vector is detected for each block. 21. The noise level detection, comprising: detecting an average noise level based on a difference value of a motion vector detected for each block; and detecting a signal within one video unit from the detected average noise level. 21. A filter process for judging the presence or absence of sticking, smoothing a temporal variation of the average noise level, and performing a filtering process for masking an input when sticking is determined to be present by the sticking determination. Signal processing method. 22. In the noise level detection, the presence or absence of signal sticking is determined for each block based on a minimum value of a difference between motion vectors detected for each block, and the sticking determination is determined to be no sticking. 21. The method according to claim 20, wherein an average noise level of the detected block is detected based on a motion vector detected for each block, and a temporal variation of the detected average noise level is smoothed. Signal processing method. 23. A motion vector is detected from an input video signal, a motion correction is performed on the input video signal based on the detected motion vector, and a difference signal between the motion-corrected signal and the input video signal is provided. Converting the difference signal into a frequency band, detecting a noise level of the input video signal based on the detection result of the motion vector, and detecting the noise level of the signal converted into the frequency band in the noise level detection. A signal processing method comprising: performing a non-linear process having a strength corresponding to a detected noise level for each frequency band; and synthesizing the signal subjected to the non-linear process and the input video signal. 24. A motion vector between video units constituting the input video signal is detected, and a moving object is extracted from the input video signal one video unit before based on the detected motion vector. 24. The signal processing method according to claim 23, wherein the motion correction is performed using the motion vector. 25. The signal processing method according to claim 24, wherein the video unit is a frame or a field. 26. The signal processing method according to claim 23, wherein the input video signal is divided into blocks, and a motion vector is detected for each block. 27. The noise level detection includes detecting an average noise level based on a difference value of a motion vector detected for each block, and detecting a signal within one video unit from the detected average noise level. 27. A filter process for judging the presence or absence of sticking, smoothing a temporal variation of the average noise level, and performing a filter process for masking an input when sticking is determined to be present by the sticking determination. Signal processing method. 28. The noise level detection, wherein presence or absence of signal sticking is determined for each block based on a minimum value of a difference between motion vectors detected for each block, and no sticking is determined by the sticking determination. 27. The method according to claim 26, further comprising: detecting an average noise level of the detected block based on a motion vector detected for each block; and smoothing a temporal variation of the detected average noise level. Signal processing method.