JP2003032685A - Noise reduction method and image processing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a video signal with high quality as a noise-reduced video signal even if an MPEG-decoded video signal is a signal obtained by decoding any of I-picture, P-picture or B-picture. SOLUTION: An MPEG-coded video signal Se is decoded by an MPEG decoder 2, and the video signal after being decoded is delayed by a delay memory 6 to be supplied to a noise reduction section 3. A cyclic coefficient Kr and an output coefficient Ko are changed depending on which picture I, P or B picture is decoded into a video signal Sd after being decoded to be inputted to the section 3 and decoded video signal before one frame or before one field. When the section 3 is formed as a non-cyclic one, a coefficient K which is to be multiplied on a signal subtracted from the decoded video signal Sd as a noise is set at Ki<Kp<Kb. Ki, Kp and Kb are coefficients with respect to I, P and B pictures respectively.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention
Ving Picture Experts Grou
The present invention relates to a method and an apparatus for decoding a video signal encoded by the p) method and reducing noise of the decoded video signal.

[0002]

2. Description of the Related Art Images such as movies and televisions are recorded on a DVD (D
When recorded on a recording medium such as digital versatile disc) or transmitted by digital broadcasting or the like, the video signal is encoded by the MPEG method, and at the time of reproduction or on the receiving side, the encoded video signal is decoded, Furthermore, noise of the decoded video signal is reduced as needed.

In this case, noise reduction (noise re
As the method of subtraction, a non-recursive method and a recursive method are considered, as in the case of a general video signal noise reduction method.

In the non-recursive noise reduction method, a minute signal of a video signal after decoding is extracted, the signal obtained by multiplying the extracted minute signal by a constant coefficient is regarded as noise, and the decoded image is decoded. Subtract from signal.

In the cyclic noise reduction method, the periodicity of the frame or field of the video signal is used to, for example, after decoding at that time, as shown in FIG. 1 of Japanese Patent Laid-Open No. 11-215403. Of the video signal and the video signal read out from the cyclic memory one frame before or one field before the decoded video signal is calculated, and a limiter extracts a component of a predetermined level or less from the difference signal. The signal obtained by multiplying the extracted components by a certain coefficient is regarded as noise, and subtracted from the decoded video signal at that time, and the subtracted video signal is written to the cyclic memory and noise Output as a reduced video signal.

[0006]

In the MPEG system, the video signal is I (Intra-coded) whether it is MPEG1 which is an encoding standard for storage media or MPEG2 which is an extended encoding standard for broadcasting and communication. , P
(Predictive-coded), B (Bidi
directionally predictive-co
As one of the three types of pictures (ded), one I picture and a plurality of P pictures and B pictures make a GOP.
It is compression-encoded so as to form an encoding unit called (Group Of Pictures).

An I picture is an image coded within a frame or field without using motion-compensated prediction between frames or fields, and is an image for which the original image can be decoded only with that information. The image has the largest amount of data and is the closest to the original image as compared with the picture and the B picture.

A P-picture is an image coded by using motion-compensated prediction in one direction between frames or fields, that is, by predicting from the closest past I-picture or P-picture, and is more data than I-picture. Although the amount is small, the amount of data is larger than that of the B picture, which is the image next to the original image next to the I picture.

B-pictures are predicted using inter-frame or inter-field bidirectional motion-compensated prediction, which is unique to MPEG, ie, from the closest past I-picture or P-picture and the next I-picture or P-picture. And is an image that has been encoded with a smaller amount of data than the I picture and P picture, and has the most different portions from the original image.

However, in the conventional method for reducing the noise of the MPEG-decoded video signal, as described above, the difference between the three types of pictures of I, P, and B in the MPEG system is not considered and the minute The coefficient by which the signal or the differential signal is multiplied is made constant, and the non-cyclic or cyclic noise reduction characteristic is made constant.

In the noise reduction method disclosed in the above-mentioned Japanese Patent Laid-Open No. 11-215403, the cyclic noise reduction characteristic is changed, but this is performed by the scan converter between the MPEG decoder and the noise reduction unit. A non-interlaced video signal with a frequency of 24 Hz has a field frequency of 60H.
It corresponds to conversion into an interlaced video signal of z, and does not change the noise reduction characteristic in consideration of the difference between the three types of pictures of I, P, and B.

Thus, in the conventional noise reduction method,
Since the noise reduction characteristic is made constant without considering the difference between the three types of pictures of I, P, and B, for example, when the above coefficient is increased in the non-cyclic noise reduction unit, the decoded image When the signal is a B picture having the smallest data amount and the most different portion from the original image, the noise is sufficiently removed, and the S / N (signal pair Although a video signal having a high noise ratio) is obtained, when the decoded video signal has a large amount of data and an I picture closest to the original image is decoded, non-noise components are removed as noise. And lose fidelity to the original image,
The video signal after noise reduction becomes a blurred image.

On the contrary, in the case of reducing the above coefficient, when the decoded video signal is a decoded I picture, the video signal after noise reduction is faithful to the original image and has no blur. However, when the video signal after decoding is a B picture decoded, noise is not sufficiently removed, and the video signal after noise reduction is S / S.
N decreases.

Therefore, according to the present invention, when a video signal encoded by the MPEG system is decoded and noise of the decoded video signal is reduced, the decoded video signal is an I picture, a P picture and a B picture. Regardless of which of the pictures is decoded, it is possible to obtain a high-quality video signal with sufficiently improved S / N and no blur due to unnecessary noise removal, as a video signal after noise reduction. It was made possible.

[0015]

According to the noise reduction method of the present invention, an I picture that is a coded image within a frame or a field, a P picture that is a one-way predictive coded image between frames or fields, and a frame. A video signal encoded by forming a group of B pictures which are bidirectional predictive coded images between fields or fields to reduce the noise of the decoded video signal and to decode the decoded video signal. The noise reduction characteristics of the decoded video signal are controlled to be changed depending on which of the I picture, the P picture and the B picture is decoded.

The video processing apparatus according to the present invention includes an I picture, which is a coded image within a frame or a field, a P picture, which is a unidirectional predictive coded image between frames or fields, and both frames or fields. A decoder that forms a group of B pictures that are directionally predictive encoded images and decodes the encoded video signal, a noise reduction unit that reduces noise in the decoded video signal, and a decoded video signal A detection control unit for changing and controlling the noise reduction characteristic of the noise reduction unit according to whether the I picture, P picture, or B picture is decoded.

[0017]

DESCRIPTION OF THE PREFERRED EMBODIMENTS [First Embodiment ... FIGS. 1 to 5]
FIG. 1 shows a first embodiment of a video processing apparatus of the present invention, which is a case where noise of an MPEG decoded video signal is reduced by a non-recursive noise reduction section.

The input terminal 1 is a video signal Se encoded by the MPEG system, which is reproduced from a recording medium such as a DVD or obtained by receiving digital broadcasting.
Is supplied.

As shown in FIG. 12, the MPEG format image data has a hierarchical structure of a sequence layer, a GOP layer and a picture layer. In the sequence layer, many sequences are continuous, a sequence header code SHC is arranged at the beginning of each sequence, a plurality of GOPs are continuously arranged after that, and a sequence end code SEC is arranged at the end of each sequence. Will be placed.

In the GOP layer, a group start code GSC forming a GOP header is provided at the beginning of each GOP.
Are arranged, and thereafter, a group of one I picture and a plurality of P pictures and B pictures is arranged.
In the picture layer, a picture start code PSC and a picture coding type PCT forming a picture header are arranged at the beginning of each picture, and encoded data is arranged after that.

The picture start code PSC indicates whether each picture in the picture layer is an I, P, or B picture.
It can be found by reading the picture coding type PCT that follows. For example, when the picture coding type PCT is “001”, it is an I picture, when it is “010”, it is a P picture, and when it is “011”, it is a B picture. The I, P, and B pictures are as described above.

In the example of FIG. 1, the coded video signal Se of the above format supplied to the input terminal 1 is supplied.
Is decoded by the MPEG video decoder 2, and the decoded video signal Sd is supplied to the noise reduction unit 3.

In the noise reduction unit 3, the limiter 11 extracts the minute signal of the decoded video signal Sd, and the multiplier 12 converts the minute signal into the detected minute signal, which will be described later.
Is multiplied by the coefficient K, and the subtractor 13 subtracts the output of the multiplier 12 as noise from the decoded video signal Sd, and the output of the subtractor 13 is output as the noise-reduced video signal. Take out to terminal 5.

In the detection control unit 4, the decoded video signal Sd input to the noise reduction unit 3 from the above-mentioned picture coding type PCT in the coded video signal Se supplied to the input terminal 1 is I, It is detected which of P and B pictures is decoded, and when the decoded video signal Sd is a decoded I picture, a small coefficient Ki is generated as the coefficient K to obtain the decoded video. Signal S
If d is a P picture decoded, the coefficient K
As the coefficient Kp, a coefficient Kp larger than the coefficient Ki is generated. When the decoded video signal Sd is a B picture decoded, a coefficient Kb larger than the coefficient Kp is generated as the coefficient K. That is, Ki <Kp <Kb.

Therefore, as compared with the P picture and the B picture, the amount of data is the largest and the I which is the closest to the original picture is used.
For the video signal decoded from the picture, the coefficient Ki
Is small and the signal subtracted as noise from the decoded video signal Sd is small, so that a video signal that is faithful to the original image and has no blur can be obtained as the video signal after noise reduction.

Further, with respect to the video signal decoded from the B picture, which has the smallest amount of data and the largest difference from the original image, as compared with the I picture and the P picture,
Since the coefficient Kb is large and the signal subtracted as noise from the decoded video signal Sd becomes large, the noise is sufficiently removed, and a video signal with high S / N is obtained as a video signal after noise reduction.

For a video signal decoded from a P-picture, which has a smaller data amount than the I-picture and a larger data amount than the B-picture, is the second closest to the original picture after the I-picture, the coefficient Kp is between the coefficient Ki and the coefficient Kb. Since the signal subtracted as noise from the decoded video signal Sd has an intermediate size, the video signal after noise reduction is close to the original image and has a high S / N. Is obtained.

The coefficient Ki for the I picture is set to a sufficiently small value close to zero. As a specific example, when the level of the portion of the decoded video signal Sd input to the noise reduction unit 3 excluding the synchronous portion is 100% white and 700 mVp-p, and the output of the limiter 11 is 100 mVp-p, the I-picture is output. When the coefficient Ki is set to 0.005, the coefficient Kp to the P picture is set to 0.01, and the coefficient Kb to the B picture is set to 0.05, the fidelity to the original image is maximized for the I picture and S for the B picture is set to the maximum. / N is the highest, and the overall image quality is the highest.

Further, it is more preferable not only to change the coefficient K according to the picture type but also to change the coefficient K according to the transfer rate of the encoded video signal Se.

The transfer rate of the coded video signal Se is a simple image, where the compression rate is high, the rate is low, and the image is complex, and the compression rate is low, the rate is high. However, since the transfer rate is not described in the MPEG format, the data amount per unit time of the encoded video signal Se is measured and detected.

FIG. 2 shows an example of this case. In this example, the detection control unit 4 detects whether the decoded video signal Sd is a decoded picture of I, P, or B as described above, and at the same time, the code supplied to the input terminal 1 is detected. The transfer rate of the encoded video signal Se is detected.

As the transfer rate of the encoded video signal Se, it is desirable to detect the average rate of a certain length of time. Therefore, the video signal after decoding from the MPEG video decoder 2 is delayed by the delay memory 6 for the time required for the detection control unit 4 to detect the transfer rate,
The video signal Sd is supplied to the noise reduction unit 3.

The detection control unit 4 generates, as the coefficient K, a function of the transfer rate of the encoded video signal Se and the original picture type of the decoded video signal Sd input to the noise reduction unit 3. Then, it is sent to the noise reduction unit 3.

Specifically, when reproducing a video signal encoded by MPEG2 from a DVD, the transfer rate thereof changes from a minimum of about 1.5 Mbps to a maximum of about 9.8 Mbps.

In this case, the coefficient Ki for the I picture,
The coefficient Kp for the P picture and the coefficient Kb for the B picture are linearly changed as shown in FIG. 3 so that the coefficient Kp for the P picture and the coefficient Kb for the B picture are maintained so as to become smaller as the transfer rate becomes higher. As shown in FIG. 4, it is changed stepwise, or as shown in FIG.

In any case, when the transfer rate is low, the coefficient K is large, and the decoded video signal S appears as noise.
When the signal subtracted from d becomes large and the transfer rate is high, the coefficient K is small and the signal subtracted from the video signal Sd after decoding as noise becomes small, so that the video signal after noise reduction has a higher image quality. The video signal of is obtained.

[Second Embodiment ... FIGS. 6 to 9] FIG.
The 2nd Embodiment of the video processing apparatus of this invention is shown, Comprising: It is a case where the noise of the MPEG decoded video signal is reduced by the cyclic noise reduction part.

In this example, as shown in FIG.
A coded video signal Se in a format such as 2
Is decoded by the MPEG video decoder 2, and the decoded video signal is delayed by the delay memory 6 to obtain the video signal Sd.
Is supplied to the noise reduction unit 3.

In the noise reduction unit 3, the subtracter 31 calculates the difference signal between the decoded video signal Sd and the decoded video signal read one frame before or one field before read from the cyclic memory 37. , The limiter 32 extracts the component of the difference signal below a predetermined level, and the multiplier 33
Then, the extracted component is multiplied by the cyclic coefficient Kr sent from the detection control unit 4 described later, and the subtracter 35 subtracts the output of the multiplier 33 as noise from the decoded video signal Sd, and subtracts The output of the device 35 is written in the cyclic memory 37, and the output coefficient K sent from the detection control unit 4 described later is added to the component extracted from the limiter 32 in the multiplier 34.
and the subtractor 36 subtracts the output of the multiplier 34 as noise from the decoded video signal Sd, and the subtractor 3
The output of 6 is output terminal 5 as a video signal after noise reduction.
Take it out.

In the detection control unit 4, the decoded video signal Sd input to the noise reduction unit 3 from the above-mentioned picture coding type PCT in the encoded video signal Se supplied to the input terminal 1 and one frame. As shown in FIG. 7, it is detected which of I, P and B pictures the decoded video signal of the previous or one field before is decoded, and the above cyclic coefficient Kr and output coefficient Ko are obtained. Generate different coefficients.

In FIG. 7, as the decoded video signal Sd input to the noise reduction unit 3, a P picture is displayed in a period (certain frame or field) F0 and a B picture is displayed in the next period F1.
A picture, a P picture in the next period F2, an I picture in the next period F3, a B picture in the next period F4, a P picture in the next period F5, and a B picture in the next period F6. When a signal obtained by decoding a B picture in the period F7 and a P picture in the next period F8 is obtained, the cyclic coefficient Kr and the output coefficient Ko are "0".
Is a coefficient of zero (a signal that is reduced as noise is zero), "small" is a minimum coefficient, "large" is a maximum coefficient, and "medium" is a coefficient between the minimum and maximum. Means

In the period F0, the input video signal Sd is I
A signal obtained by decoding a P picture closest to the original image next to the picture is obtained, so both the output coefficient Ko and the cyclic coefficient Kr are set to "medium".

In the period F1, since a signal obtained by decoding the B picture with the highest degree of ambiguity is obtained as the input video signal Sd, the output coefficient Ko is set to "large" to greatly improve the S / N. However, since a signal obtained by decoding a B picture with a high degree of ambiguity cannot be positively used for cyclic use, the cyclic coefficient Kr is set to “small”.

In the period F2, P is used as the input video signal Sd.
Since a signal obtained by decoding a picture is obtained, both the output coefficient Ko and the cyclic coefficient Kr are set to "medium", as in the period F0.

In the period F3, a signal obtained by decoding the I picture closest to the original image is obtained as the input video signal Sd. Therefore, the output coefficient Ko is set to "0" and the video signal close to the original video signal is output. 5, the cyclic coefficient Kr is also set to “0”, and a video signal close to the original video signal is written in the cyclic memory 37.

In the period F4, a signal obtained by decoding a highly ambiguous B picture is obtained as the input video signal Sd, and a signal obtained by decoding an I picture close to the original image is read from the cyclic memory 37. Therefore, the output coefficient Ko is set to “large” to improve the S / N ratio and the cyclic coefficient Kr is increased.
Is set to “large” and the video signal with improved image quality is written in the patrol memory 37.

In the period F5, P is used as the input video signal Sd.
Since a signal obtained by decoding a picture is obtained, the periods F0, F
Similar to 2, the output coefficient Ko and the cyclic coefficient Kr are both set to “medium”.

In the period F6, a signal obtained by decoding a highly ambiguous B picture is obtained as the input video signal Sd, and the signal obtained by decoding the P picture is read from the cyclic memory 37. Therefore, as in the case of the period F4, the output is performed. The coefficient Ko is set to “large” to improve the S / N ratio and the cyclic coefficient K is increased.
Also, r is set to “large”, and the video signal with the improved image quality is written in the cyclic memory 37.

In the period F7, a signal obtained by decoding a highly ambiguous B picture is obtained as the input video signal Sd, and the signal obtained by decoding the highly ambiguous B picture is also read from the cyclic memory 37, so that the output coefficient Ko is obtained. And the cyclic coefficient K
Both r are “small”.

In the period F8, P is used as the input video signal Sd.
Since a signal obtained by decoding a picture is obtained, the periods F0, F
Similarly to 2 and F5, both the output coefficient Ko and the cyclic coefficient Kr are set to “medium”.

FIG. 8 shows another example of the cyclic type. In this example, the noise reduction unit 3 uses the patrol memory 37.
Insert the switch 39 on the input side of the
9 is switched to the terminal X side, the output of the subtractor 35 is written in the cyclic memory 37, and the periods F1, F4 and F4 shown in FIG.
As in F6 and F7, the switch 39 is set to the terminal Y during the period in which the decoded picture signal Sd input to the noise reduction unit 3 is the decoded picture of the B picture with high ambiguity.
The configuration is such that the signal obtained by decoding the B-picture having a high degree of ambiguity is not written to the cyclic memory 37, and the signal read from the cyclic memory 37 at that time is cyclically written to the cyclic memory 37. To do.

According to this, noise is more reliably extracted and more effective even in the period next to the period in which the signal obtained by decoding the B picture is obtained as the decoded video signal Sd input to the noise reduction unit 3. The noise is reduced.

Instead of inserting the switch 39, writing to the patrol memory 37 and reading from the patrol memory 37 may be controlled so that the same result is obtained.

Also in the case of the cyclic type as shown in FIG. 6 or 8, not only the coefficient K (the output coefficient Ko and the cyclic coefficient Kr) is changed depending on the picture type, but also the transfer rate of the encoded video signal Se is changed. The coefficient K may be changed according to the above.

Further, in the example shown in FIGS. 6 and 8, the limiter 3 is used.
2 and subsequent cases are divided into a cyclic loop system composed of the multiplier 33 and the subtractor 35 and an output system composed of the multiplier 34 and the subtractor 36. In the noise reduction unit 3 of FIG. 6 or FIG. 34 and the subtractor 36 are omitted, and the subtractor 35
The output of the above may be taken out to the output terminal 5 at the same time as being written in the circulating memory 37. In this case, the cyclic coefficient simultaneously becomes the output coefficient.

The recursive noise reduction is intended to reduce noise in the decoded video signal on the assumption that the video signal has continuity. Therefore, when the motion of the video signal between the frames or between the fields is large, the afterimage becomes remarkable.

Therefore, the coefficient K depends on the picture type.
It is more preferable to detect not only the (output coefficient Ko and the cyclic coefficient Kr) but also the movement between frames or fields of the video signal and change the coefficient K according to the movement.

In this case, the motion vector between frames or fields of the video signal is detected, and the magnitude of the motion vector is as shown by the solid line, broken line or chain line in FIG.
When the magnitude of the motion vector is equal to or less than a certain value, the coefficient K becomes zero. When the magnitude of the motion vector is equal to or less than the value, the coefficient K (the minimum value and the maximum value of the output coefficient Ko and the cyclic coefficient Kr) is increased. , Intermediate value) becomes a fixed value, or changes in steps or continuously.

According to this, it is possible to reduce the afterimage when the motion between frames or fields of the video signal is large.

[Embodiment of Interpolation for MPEG1 ... FIGS. 10 and 11] The image data of MPEG1 is SIF.
(NTT (National Television), which is called "Source Input Format".
n System Committee range, the number of horizontal pixels is 352, the number of lines is 240, the frame frequency is 29.
It is a non-interlaced signal of 94Hz, PAL (Ph
In the case of “alternation by line”, it is a non-interlaced signal having 352 horizontal pixels, 288 lines, and a frame frequency of 25 Hz.

Therefore, when a video signal coded by MPEG1 is decoded and converted into an NTSC standard video signal or a PAL standard video signal for output, interpolation is performed in both horizontal and vertical directions. Average value interpolation is generally used as the interpolation method.

In this case, as for the horizontal direction, FIG.
As shown in (A), if the image data of each pixel in the horizontal direction before interpolation is Pa, Pb, Pc, Pd, ..., Pa ′ = (Pa + Pb) / 2 (1) , The average value of the image data of two adjacent pixels is set as the image data of the pixel to be located between the two pixels.

On the other hand, in the vertical direction, MP
Only the image data of one field exists in the decoded output of EG1. Therefore, as shown in FIG. 10B, the image data Va, Vb,
Let Vc, Vd, ... Be the image data of each line of the odd field, and for each corresponding pixel, Va '= (Va + Vb) / 2 (2) The average value of the image data of one line is set as the image data of the line of the even field which should be located between the two lines.

However, the image data Va ', Vb', Vc ', ... Of each line in the even field is blurred image data obtained by averaging the image data of two adjacent lines in the odd field. .

Therefore, in the above-mentioned embodiment, when the coded video signal Se supplied to the input terminal 1 is coded by MPEG1, the coded video signal Se is
Is decoded by the MPEG video decoder 2, the above-described interpolation is performed on the decoded video signal Sd, and noise is reduced by the noise reduction unit 3 for the interpolated video signal, the noise reduction effect becomes weak, and the code MPEG video signal Se
It is more effective in noise reduction to perform decoding by the video decoder 2, reduce noise in the decoded video signal Sd in the noise reduction unit 3, and then perform the above interpolation for the noise-reduced video signal.

FIG. 11 shows an example of a DVD reproducing apparatus which performs interpolation after noise reduction in this way. Disc (DV
In D) 41, a video signal and an audio signal encoded by MPEG1 are multiplexed, modulated and recorded.

From the disc 41, the optical pickup 42
The signal is read by. The read signal is taken out through the RF amplifier 43, demodulated by the demodulation unit 44, error-corrected by the error correction unit 45, written in the buffer 46, and read from the buffer 46. The multiplexed signal output from the buffer 46 is separated by the demultiplexer 47 into an encoded video signal and an encoded audio signal.

The coded video signal separated by the demultiplexer 47 is decoded by the MPEG video decoder 2, and the decoded video signal is noise-reduced by the noise reduction section 3 as described above, and further noise-reduced. The video signal is interpolated by the interpolation calculator 48 as described above, and the NTSC standard video signal or the PAL standard video signal is obtained at the video output terminal 49.

The encoded audio signal separated by the demultiplexer 47 is decoded by the MPEG audio decoder 51, and the decoded audio signal is obtained at the audio output terminal 52.

The controller 59 controls each section of the reproducing apparatus, and the detection control section 4 is also controlled by the controller 59 to send the above-mentioned coefficient to the noise reduction section 3.

[0071]

As described above, according to the present invention, M
Whether the I-picture, the P-picture or the B-picture is decoded as the PEG-decoded video signal, the S / N is sufficiently improved as the video signal after noise reduction, and unnecessary noise is generated. It is possible to obtain a high-quality video signal without blurring due to removal.

[Brief description of drawings]

FIG. 1 is a diagram showing an example of a video processing device of the present invention.

FIG. 2 is a diagram showing an example of a video processing device of the present invention.

FIG. 3 is a diagram showing an example of characteristics when a coefficient is changed according to a transfer rate.

FIG. 4 is a diagram showing an example of characteristics when a coefficient is changed according to a transfer rate.

FIG. 5 is a diagram showing an example of characteristics when a coefficient is changed according to a transfer rate.

FIG. 6 is a diagram showing an example of a video processing device of the present invention.

7 is a diagram showing an example of coefficients of the video processing device of the example of FIG. 6;

FIG. 8 is a diagram showing an example of a video processing device of the present invention.

FIG. 9 is a diagram showing an example of characteristics when a coefficient is changed according to the magnitude of a motion vector.

FIG. 10 is a diagram for explaining interpolation in the case of MPEG1.

FIG. 11 is a diagram showing an example of a DVD reproducing device to which the present invention is applied.

FIG. 12 is a diagram showing a hierarchical structure of image data in MPEG format.

[Explanation of symbols]

Since the main parts are all described in the figure, they are omitted here.

   ─────────────────────────────────────────────────── ─── Continued front page    F-term (reference) 5C021 PA12 PA32 PA42 PA66 PA67                       PA79 RB07 RC01 SA24 XB11                       YA01 YC08                 5C059 KK01 MA00 PP05 PP06 PP07                       SS02 SS13 TA69 TB04 TC24                       TC37 UA02 UA05 UA11 UA33                 5J064 AA01 BB07 BC02 BC08 BC09                       BC17 BC24 BD01

Claims (6)

[Claims] 1. An I picture, which is a coded image within a frame or a field, a P picture, which is a unidirectional predictive coded image between frames or fields, and a bidirectional predictive coded image between frames or fields. A group of a certain B picture is formed to decode a coded video signal, noise of the decoded video signal is reduced, and the decoded video signal is an I picture, a P picture, or a B picture. A noise reduction method characterized by changing and controlling a noise reduction characteristic for a video signal after decoding, depending on whether it is decoded. 2. An I picture, which is an intra-frame or intra-field encoded image, a P picture, which is an inter-frame or inter-field unidirectional predictive encoded image, and a bi-directional, inter-frame or inter-field predictive encoded image. A group of a certain B picture is formed to decode a coded video signal, noise of the decoded video signal is reduced, and the decoded video signal is an I picture, a P picture, or a B picture. A noise reduction method characterized by changing and controlling noise reduction characteristics for a decoded video signal according to whether it is a decoded video signal or a transfer rate of an encoded video signal. 3. An I picture, which is an intra-frame or intra-field encoded image, a P picture, which is an inter-frame or inter-field unidirectional predictive encoded image, and a inter-frame or inter-field bi-directional predictive encoded image. A group of certain B pictures is formed to decode a coded video signal, noise of the decoded video signal is reduced by a cyclic noise reduction unit, and the decoded video signal is an I picture or P picture. A noise reduction method, characterized in that a coefficient for determining a noise reduction characteristic of the cyclic noise reduction unit is controlled to be changed according to which of a B picture and a B picture is decoded. 4. An I picture, which is a coded image within a frame or a field, a P picture, which is a unidirectional predictive coded image between frames or fields, and a bidirectional predictive coded image, between frames or fields. A decoder that forms a group of certain B pictures and decodes the coded video signal, a noise reduction unit that reduces noise of the decoded video signal, and a decoded video signal that is I picture, P picture, A video processing device, comprising: a detection control unit that changes and controls the noise reduction characteristics of the noise reduction unit according to which of the B pictures is decoded. 5. An I picture which is a coded image within a frame or a field, a P picture which is a unidirectional predictive coded image between frames or fields, and a bidirectional predictive coded image between frames or fields. A decoder that forms a group of certain B pictures and decodes the coded video signal, a noise reduction unit that reduces noise of the decoded video signal, and a decoded video signal that is I picture, P picture, A detection control unit for changing and controlling the noise reduction characteristic of the noise reduction unit according to which of the B pictures is decoded and the transfer rate of the encoded video signal. Video processing device. 6. An I picture which is a coded image in a frame or a field, a P picture which is a unidirectional predictive coded image between frames or a field, and a bidirectional predictive coded image between frames or fields. A decoder that forms a group of certain B pictures and decodes the encoded video signal, a cyclic noise reduction unit that reduces noise in the decoded video signal, and a decoded video signal that is an I picture, P A video processing apparatus, comprising: a detection control unit that changes and controls a coefficient that determines a noise reduction characteristic of the cyclic noise reduction unit depending on whether a picture or a B picture is decoded. .