JP3371773B2 - Signal processing method and signal reproducing device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a DCT for video signals.
The present invention relates to a signal processing method and a signal reproducing device for decoding a video signal coded by band compression accompanied by block processing such as (discrete cosine transform).

[0002]

2. Description of the Related Art In recent years, transform coding such as DCT has become the main stream compression method. These are the information of the original image 1 /
This is a method of highly efficient compression of about 10 to 1/100.

However, as the compression rate becomes higher, noise peculiar to digital compression, that is, block noise and mosquito noise are generated, and the decoded image quality is deteriorated.

Recently, various proposals have been made as methods for reducing the noise that causes the deterioration of the decoded image quality. Among them, the noise reduction method that utilizes the fact that the time-axis correlation of images is strong is a method of storing a plurality of past images in the frame memory or field memory and configuring a filter in the time-axis direction. It effectively reduces only the noise component that has little correlation with
Widely known.

FIG. 4 shows a block diagram of conventional noise reduction in the time axis direction. In FIG. 4, the compressed data is decoded by the decoder 5 and then output as a signal g. A signal h obtained by delaying the signal g by one frame in the frame memory 6 is subtracted from the signal g to obtain a difference signal i.
Is subtracted from the signal g by weighting the signal j with the coefficient K, which is the coefficient circuit K7, to obtain the signal k in which the noise component of the signal g is reduced. In this case, one frame memory for the number of pixels is always required. Although there is a feedback type noise reduction, one frame memory is always required.

[0006]

As described above, in the conventional noise reduction in the time axis direction, in addition to the decoding process of the compressed data, only for the noise reduction,
A frame or field memory is required, which causes an increase in circuit scale and an increase in cost, which is not practical.

We have found that such noise due to block coding is easily detected on a block-by-block basis, that is, it is effective by averaging the DC and DC values of the block in the time axis direction. It was also found that it is particularly effective to suppress the fluctuation of the DC value of the B frame.

An object of the present invention is to propose a signal processing method and a signal reproducing apparatus which make use of this and enable noise reduction in the time axis direction with almost no increase in the number of memory circuits.

[0009]

DISCLOSURE OF THE INVENTION The present invention comprises three types of intraframe or field coding (I), forward frame or interfield predictive coding (P), bidirectional frame or interfield predictive coding (B). Is a signal processing method at the time of decoding the data encoded by the encoding method using the above encoding method and band-compression encoding with a plurality of adjacent pixels as one block, and the DC component of the block. Is configured to perform the filtering processing in the time axis direction only.

As a result, noise reduction in the time axis direction can be achieved with almost no increase in the number of memory circuits.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The invention according to claim 1 of the present invention is the intra-frame or intra-field coding (I), the forward frame or inter-field predictive coding (P), the bi-directional frame or inter-field predictive coding. (B) is a signal processing method which uses three types of encoding methods and which is performed when decoding data encoded by an encoding method in which a plurality of adjacent pixels are used as one block and is band compression encoded. , Is configured so that only the DC component of the block is filtered in the time axis direction, and a memory for noise reduction other than the memory necessary for decoding compressed data is not significantly increased, and calculation is performed. This has the effect of obtaining a noise reduction effect in the time axis direction with a significantly reduced amount.

According to a second aspect of the present invention, only the B frame or field is subjected to a filtering process in the time axis direction using the information of the preceding or following I or P frame or field, and then output. It has the effect of obtaining a noise reduction effect in the time axis direction, with almost no need for a memory for noise reduction.

According to the third aspect of the present invention, the value of the target frame or field is K, the target values of the preceding and following frames or fields are A1, A2, and K
And A1 and A2, S1 = | K-A1 |, S2 = |
K-A2 |, when at least one of S1 and S2 is smaller than a predetermined value T1, it is configured to perform a filtering process in the time axis direction, and the influence of noise reduction on the original signal. The effect that can be made small.

According to a fourth aspect of the present invention, the value of the target frame or field is K, the target values of the preceding and following frames or fields are A1, A2, and K
And A1 and A2, S1 = | K-A1 |, S2 = |
K-A2 |, when at least one of S1 and S2 is larger than the predetermined value T1, the filtering process in the time axis direction is not performed, and noise reduction reduces the original signal. The effect is that the influence can be reduced.

According to a fifth aspect of the present invention, when the motion vector amount of the target pixel or block is MV, the filtering process in the time axis direction is performed only when MV is smaller than the predetermined amount T2. It is configured, and has an effect of obtaining an effect of reducing image bleeding and afterimage in the time axis direction.

The inventions according to claims 6 to 10 are:
Intra-frame or field coding (I), forward frame
Frame or inter-field predictive coding (P), bidirectional frame
3 types of code, which are the frame or inter-field predictive coding (B)
Encoding method, and the neighboring pixels are
It is encoded by the band compression encoded encoding method as the lock.
A signal reproducing device for decoding encoded data,
A DC value storage means for storing the DC component of the block;
From the input data of the DC value storage means, the DC value storage means
Subtracting means for subtracting the output data of the
A coefficient means for multiplying the force by a coefficient and an input of the DC value storage means
Force data and a calculation means for calculating the output of the coefficient means
And that only the DC component of the block has a time axis direction filter.
A signal reproducing device characterized by applying a filter ,
It has an effect that it is possible to provide a device that obtains a noise reduction effect in the time axis direction that requires substantially no noise reduction memory and that significantly reduces the amount of calculation.

FIG. 1 shows an embodiment of the present invention.
~ It demonstrates using FIG. (Embodiment 1) FIG. 1 is a DC component correction block diagram of a DCT block inside a decoder.

In FIG. 1, the compressed data is decoded by the DC value decoder 1 and then output as a DC component signal a. From the signal a, the signal a is converted into the DC value by the frame memory 2
The signal b delayed by 1 frame is subtracted to obtain the difference signal c, and the difference signal c is weighted by the coefficient K to subtract the signal d from the signal a, thereby reducing the fluctuation of the signal a which is the DC component. This is a configuration for obtaining the signal e. The AC component is the same as that of the conventional decoder and is not a target for correction.

Here, FIG. 1 is similar to the conventional example as a block diagram, but the memory capacity is significantly different. That is, in the conventional frame memory, all pixels are targeted. For example, in the case of 720 × 480 pixels, 8b
Since it has two elements Y and C in it, it is 720 × 480 × 8 ×
2 = required about 5.5 megabits of memory,
In the present invention, since it is 1/64 in the case of an 8 × 8 pixel block, about 86 kilobits are sufficient. Further, the amount of calculation is similarly reduced to 1/64. This has a great effect on making IC.

Next, a second embodiment will be described in which the DC value frame memory shown in FIG. 1 is unnecessary. Decodes data encoded by three types of encoding methods: intra-frame or intra-field encoding (I), forward frame or inter-field predictive encoding (P), bidirectional frame or inter-field predictive encoding (B) The decoder refers to the preceding or following I or P frame as a reference image when decoding the B frame. That is, the two frames before and after exist in the memory in the decoder. The present invention is to reduce the noise of only the B frame by using the DC value of the data of the two frames. In this case, the memory for noise reduction is unnecessary. Further, this corresponds to that the block fluctuation of the B frame is easily detected as noise.

FIG. 2 is a pixel-level schematic diagram showing changes of a target value, for example, a DC value of a block, for each frame from 1 frame to 13 frames. Let the values of each frame be a to m.

Here, 1 and 13 frames are intraframe coding (I), 4, 7 and 10 frames are forward interframe predictive coding (P), and other frames are bidirectional interframe predictive coding (B). ). The frames subject to noise reduction may be all frames as shown in the first embodiment, but in the second embodiment, the case of only B frames will be described.

When decoding the second frame and the third frame, the first, fourth and fourth frames are present in the frame memory in the decoder as reference images.

Therefore, the decoded value b of the two frames is corrected using the information of the reference values a and d. The correction method is averaging, and a filter will be applied.

When the filter coefficient is, for example, 1: 1: 1, the 2 frame value b is corrected to (a + b + d) / 3 and the 3 frame value c is corrected to (a + c + d) / 3.

When the filter coefficient is, for example, 1: 2: 1, the 2 frame value b is corrected to (a + 2b + d) / 4 and the 3 frame value c is corrected to (a + 2c + d) / 4.

When the filter coefficient is, for example, 1: 2: 3, the 2 frame value b is corrected to (2a + 3b + c) / 6 and the 3 frame value c is corrected to (a + 3c + 2d) / 6.

When the filter coefficient is, for example, 1: 2: 4, the 2 frame value b is corrected to (2a + 4b + c) / 7 and the 3 frame value c is corrected to (a + 4c + 2d) / 7.

Here, various filter coefficients can be considered. The values of a and d are not corrected. In this way, a case where the filter coefficient is set to 1: 2: 3 as an example of the corrected result is shown by a black circle in FIG.

Here, for a frame with a large level change, it can be determined that the change is not noise, so it is not a target for correction. In FIG. 2, this corresponds to 8 and 9 frames.

In the case of 8 frames, | g−h | <T1, |
Let h−j |> T1. Here, T1 is a threshold that is arbitrarily set. At this time, as the correction of 8 frames, information of 7 frames, which is a frame having a difference smaller than T1, may be used for correction. In this case, the 8-frame value h is, for example, (g + 2h) / 3.

Further, for 8 frames, it is not necessary to make any correction in consideration of | h−j |> T1.

As a result, image bleeding in the time axis direction and afterimage can be suppressed. Also, the P frame and the B frame have a motion vector MV.

Since the noise reduction in the time axis direction is effective when the motion is small, it is possible to control the above correction by observing the magnitude of MV.

FIG. 3 is a schematic diagram showing the relationship between an example of the motion vector MV and the correction target frame at that time. M
When V is larger than the predetermined value T2, all the above-mentioned corrections are prohibited in that section, so that in FIG. 3, 8 to 12 frames are the correction prohibited section.

The correction prohibited section may be controlled on a frame-by-frame basis, on a 3-frame basis (reference frame basis) basis, or on a GOP (Group Of Picture) basis basis.
The control may be performed for a longer period. The embodiment shown in FIG. 3 shows an example in units of reference frames. In the case of GOP units, correction is prohibited for all GOPs in which even one frame has a MV larger than T2.

Although in the above-described embodiment, an example of frame correction is shown in FIGS. 2 and 3, the same applies to field correction.

The coefficient of the filter is not limited to the coefficient described in the embodiment, and it goes without saying that it can be arbitrarily set.

Further, in the above-described embodiment, the threshold values T1 and T2 are set to be constant regardless of the screen, but the threshold values are adaptively adjusted according to the screen, for example, the noise amount, the recording / reproducing mode, and the like. It is also possible to change the values of T1 and T2.

Further, the description of the present embodiment is exactly the same for the signal reproducing apparatus which outputs after performing the processing using the present invention.

[0041]

INDUSTRIAL APPLICABILITY As described above, according to the present invention, there are three types of intraframe or intrafield coding (I), forward frame or interfield predictive coding (P), bidirectional frame or interfield predictive coding (B). A signal processing method at the time of decoding data encoded by a band compression encoding method using a plurality of adjacent pixels as one block, which is a DC signal of the block. It is configured so that only the components are filtered in the time axis direction, which enables effective noise reduction in the time axis direction with almost no increase in the memory circuit. is there.

[Brief description of drawings]

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

FIG. 2 is a pixel level schematic diagram of the same.

FIG. 3 is a schematic diagram of the same

FIG. 4 is a block diagram for explaining a conventional example.

[Explanation of symbols]

1 DC value decoder 2 DC value frame memory 3 coefficient circuit K

Claims (10)

(57) [Claims] 1. Use of three types of coding schemes: intraframe or intrafield coding (I), forward frame or interfield predictive coding (P), bidirectional frame or interfield predictive coding (B), and A signal processing method at the time of decoding data encoded by a band compression encoding method using a plurality of adjacent pixels as one block, wherein a filter in the time axis direction is applied only to the DC component of the block. A signal processing method characterized by performing a process for applying a signal. 2. The method according to claim 1, wherein only the B frame or field is subjected to a filtering process in the time axis direction using the information of the preceding or following I or P frame or field, and then output. Signal processing method. 3. A value of a target frame or field is K, target values of preceding and following frames or fields are A1, A2, and a difference between K and A1, A2 is
When S1 = | K-A1 | and S2 = | K-A2 |, and when at least one of S1 and S2 is smaller than a predetermined value T1, processing for filtering in the time axis direction is performed. Item 3. The signal processing method according to any one of items 1 and 2. 4. A value of a target frame or field is K, target values of preceding and following frames or fields are A1, A2, and a difference between K and A1, A2 is
When S1 = | K−A1 | and S2 = | K−A2 |, when at least one of S1 and S2 is larger than a predetermined value T1, the filtering process in the time axis direction is not performed. The signal processing method according to claim 1. 5. When the motion vector amount of the target pixel or block is MV, the filtering process in the time axis direction is performed only when MV is smaller than a predetermined amount T2. 2. The signal processing method according to any one of 2. 6. Frame or field intra-coding
(I), forward frame or inter-field predictive coding
(P), bidirectional frame or inter-field predictive coding
(B) three types of encoding methods are used, and
Band compression coded with several pixels as one block
A signal reproducing device at the time of decoding data encoded by an encoding method , which stores a DC component of the block.
From the C value storage means and the input data of the DC value storage means
Subtraction means for subtraction output data of the DC value storing means
And coefficient means for multiplying the output of the subtraction means by a coefficient,
The input data of the DC value storage means and the output of the coefficient means are
And a calculation means for calculating the DC component of the block
Only the signal reproducing apparatus according to claim Rukoto over time axis direction of the filter. 7. B frame or field only, front and back
Using I or P frame or field information
Output after performing the processing of filtering in the time axis direction
7. The signal reproducing device according to claim 6, wherein 8. A frame or field of interest
The value is K, and it is the target of the previous or next frame or field.
The values to be set are A1 and A2, and the difference between K and A1 and A2 is
When S1 = | K-A1 | and S2 = | K-A2 |,
At least one of S1 and S2 is smaller than the predetermined value T1.
Specially, it performs the process of filtering the time and time axis.
The signal reproducing device according to claim 6, which is a characteristic of the signal reproducing device. 9. The frame or field of interest
The value is K, and it is the target of the previous or next frame or field.
The values to be set are A1 and A2, and the difference between K and A1 and A2 is
When S1 = | K-A1 | and S2 = | K-A2 |,
At least one of S1 and S2 is larger than the predetermined value T1.
When, do not perform processing to filter in the time axis direction
The signal reproducing device according to any one of claims 6 and 7, characterized in that
Place 10. The target pixel or block has
MV is a predetermined amount T2, where MV is the motion vector amount
When the value is smaller than the
The signal re-establishment according to any one of claims 6 and 7, characterized in that
Raw equipment.