JPH08172630A - Moving image coder - Google Patents

Abstract

PURPOSE: To improve the image quality and to reduce the code quantity by enhancing the accuracy of motion compensation in the coder for a moving image. CONSTITUTION: Since a motion vector detected by a motion detection circuit 44 of a coding circuit 24 from a subband signal of an LL2 component is used for a reference position to detect a motion with respect to a subband signal of other component, the motion vector is given to a reference calculation motion detection circuit 50 of coding circuits 12, 14,16, 18, 20, 22. Processing such as motion compensation, inter subband signal difference, quantization of a predicted error signal and variable length coding is conducted by using the motion vector detected from each component and the code after the processing is stored in a buffer 34.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture image coding apparatus, and more particularly to a moving picture coding apparatus for performing sub-band coding of a digital moving picture.

[0002]

2. Description of the Related Art Conventionally, subband coding has been known as one of compression methods for image signals, particularly HDTV image signals. In subband coding, an image signal is divided into a plurality of frequency bands by horizontal and vertical band division filters, and an optimal encoding process is performed for each divided frequency band.

FIG. 6 is an explanatory diagram for dividing an input image signal into four as an example of sub-band division. As a band division filter, a QMF (quadrature mirror filter) is known to remove aliasing distortion. In FIG. 6, QM
F400, 408, and 412 are bandpass filters having high-pass characteristics (shown by HP), and QMFs 402, 410, and 41.
4 is a band-pass filter having a low-pass characteristic (shown by LP),
The downsampling (decimation) circuits 404 and 406 are horizontal downsampling, and the downsampling circuits 416, 418, 420 and 422 are vertical downsampling. That is, the image signal is QMF40.
It is divided into two bands by horizontal filtering by 0 and 402. Further down-sampling circuit 40
4, 406, horizontal downsampling is performed. After this, QMF 408, 410, 412, 41
Vertical filtering by 4 divides it into 4 bands. Further down-sampling circuits 416, 41
Downsampling in the vertical direction is performed by 8, 420 and 422. As a result, the input image signal is LL1:
Low-frequency component (representing the lowest frequency band) in both horizontal and vertical directions LH1: Low-frequency component only in horizontal direction HL1: Low-frequency component only in vertical direction HH1: Four band components of high-frequency component in both horizontal and vertical directions (hereinafter, subbands) It is divided into). The number of pixels in each subband is ½ of the input image signal in both the horizontal and vertical directions.

As another example of the above sub-band division, FIG. 7 is an explanatory diagram for dividing an input image signal into seven. In FIG. 7, QMF 500, 508, 512,
524, 532, and 536 are high-pass characteristics (illustrated by HP)
Band division filters of QMF502, 510, 514,
526, 534, and 538 are low-pass characteristics (shown by LP)
Band division filters, downsampling (decimation) circuits 504, 506, 528, and 530 are horizontal downsampling and downsampling circuits 516 and 51.
8,520,522,540,542,544,546
Is vertical downsampling. After the input image signal is divided into four in the same manner as in the operation of FIG. 3, the LL1 component is repeatedly filtered in the horizontal and vertical directions and down-sampled to obtain seven sub-bands (HH1, HL1, LH1, HH2, HL2, LH2,
LL2: Numerical values 1 and 2 represent the number of layers). The lowest frequency band here is LL2. In addition, each subband of layer 2 (HH2, HL2, LH2, LL2)
The number of pixels is 1 in the input image signal in both the horizontal and vertical directions.
It is / 4.

As such a moving picture coding apparatus, for example, there is one in which the picture coding apparatus described in Japanese Patent Laid-Open No. 16887/1990 is applied to subband coding. FIG.
FIG. 4 is an explanatory diagram of a conventional moving picture coding apparatus for band division subband coding. In FIG. 8, 610 and 612
Is a horizontal band division filter, 614, 616, 61.
8, 620 are vertical band division filters, 622, 6
Reference numerals 24, 626, and 628 are encoding circuits having the same configuration, 630 is a frame memory that stores a subband signal of the previous frame, and 632 is a subband signal of the current frame and a subband signal of the previous frame of the frame memory 630. A motion detection circuit for detecting a motion vector between
4 is a motion compensation circuit that performs motion compensation based on the motion vector detected by the motion detection circuit 632 and generates a prediction signal,
636 is a subtractor, 638 is a discrete orthogonal transform circuit that orthogonally transforms the difference between the prediction signal and the subband signal of the current frame, 640 is a quantizer that quantizes the transform coefficient, and 642 is the quantized signal and motion. A variable-length coding circuit that performs variable-length coding of a vector such as Huffman code, 644 is a buffer that temporarily stores and transmits the coded signal, and 646 is an inverse quantizer that dequantizes the quantized signal. , 648 is a discrete inverse orthogonal transform circuit for inverse orthogonal transforming the dequantized signal, and 650 is an adder.

Next, the operation of the image coding apparatus shown in FIG. 8 will be described. The input image signal is a horizontal band division filter 61.
0, 612, vertical band division filters 614, 61
It is divided into four subband signals by 6, 618 and 620. These sub-band signals are input to independent encoding circuits 622, 624, 626 and 628, respectively. For the sub-band signal of the current frame, a motion vector is detected between the sub-band signal of the previous frame in the frame memory 630 and the motion detection circuit 632. Using the motion vector, motion compensation circuit 634 independently performs motion compensation prediction on each subband signal,
The difference between the signal predicted by the subtractor 636 and the subband signal of the current frame is calculated, and the prediction error signal is orthogonally transformed by the discrete orthogonal transformation circuit 638. The transform coefficient is quantized and variable-length coded for each subband signal by the quantizer 640 and the variable-length coding circuit 642, and then the buffer 64.
4 is output. The motion vector detected by the motion detection circuit 632 is also coded by the variable length coding circuit 642 and output to the buffer 644. On the other hand, for the signal quantized by the quantizer 640, the inverse quantizer 646, the discrete inverse orthogonal transform circuit 648, the adder 650,
The reverse processing is performed by the motion compensation circuit 634 to decode the signal. This is stored in the frame memory 630 and becomes the subband signal of the previous frame.

On the other hand, Japanese Unexamined Patent Publication No. 4-322593 discloses a moving picture coding apparatus as shown in FIG.

In FIG. 9, reference numerals 610 and 612 denote horizontal band division filters, and 614, 616, 618 and 620.
Is a vertical band division filter, 700, 702, 70
4, 706 are coding circuits, and the coding circuits 700, 7
02 has the same configuration as the encoding circuit 704. In the encoding circuit 704, the output side of the band division filter 618 is connected to the plus input side of the subtractor 636. The subtraction output side of the subtractor 636 is connected to the input side of the discrete orthogonal transform circuit 638, and the output side of the discrete orthogonal transform circuit 638 is connected to the input side of the quantizer 640, as in the conventional example of FIG. , To the variable length coding circuit 708.
The output side of the variable length encoding circuit 708 is connected to the input side of the buffer 644. The other output side of the quantizer 640 is connected to the input side of the dequantizer 646, the output side of the dequantizer 646 is connected to the input side of the discrete inverse orthogonal transform circuit 648, and the discrete inverse The output side of the orthogonal transformation circuit 648 is connected to the plus input side of the adder 650.
The addition output side of the adder 650 is connected to the input side of the frame memory 710, and the output side of the frame memory 710 is connected to the other input side of the motion compensation circuit 634. The output side of the motion compensation circuit 634 is the subtractor 6
It is connected to the negative input side of 36 and the positive input side of the adder 650.

Next, the encoding circuit 706 has substantially the same configuration as the encoding circuit 704, and the same reference numerals are used for the same or corresponding components. However, the motion detection circuit 7
12 are provided, and the output sides of the band division filter 620 and the frame memory 630 are connected to each other. Then, one output side of the motion detection circuit 712 is connected to the input side of the motion compensation circuit 634 of the encoding circuits 700, 702, 704, and the other output side thereof is connected to the variable length encoding circuit 6.
It is connected to the input side of 42.

Of the above parts, the band division filter 61
0, 612, 614, 616, 618 and 620 are the same as those shown in FIG. Further, the encoding circuit 706 is
The configuration is similar to that of the encoding circuit 628 shown in FIG. 8, and its operation is also similar. However, the encoding circuit 70
0, 702, and 704 are not provided with the motion detection circuits individually, and the motion detection circuit 712 in the encoding circuit 706 is provided.
The motion vector detected by is used.

Next, the operation of the moving picture coding apparatus shown in FIG. 9 will be described. The input image signal is divided into two sub-band signals of a low frequency component and a high frequency component in the horizontal direction by horizontal band division filters 610 and 612. And
These two sub-band signals are passed through the vertical band-splitting filters 614, 616, 618, 620 to 4
It is divided into two frequency components (HH1, HL1, LH1, LL1). Of these, the subband signal of the LL1 component is input to the motion detection circuit 712 of the encoding circuit 706. The sub-band signal of the decoded image of the previous frame stored in the frame memory 630 is also input to the motion detection circuit 712, and the motion vector is detected from the sub-band signal between these frames. The detected motion vector is input to the motion compensation circuit 634 on the one hand. In the motion compensation circuit 634, the frame memory 6
In the decoded signal of the previous frame stored in 30,
A signal at a position shifted by the motion vector value is cut out as a prediction signal. Next, the prediction signal is input to the subtractor 636 together with the subband signal of the current frame output from the band division filter 620. The subtractor 636 obtains the difference between both subband signals,
The result is divided into blocks each including a plurality of pixels and supplied to the discrete orthogonal transform circuit 638. The discrete orthogonal transform circuit 638 performs an orthogonal transform process such as DCT (discrete cosine transform) on the input signal, and outputs transform coefficients to the quantizer 640. Next, in the quantizer 640,
The input transform coefficient is quantized, and the quantized transform coefficient value is supplied to the variable length coding circuit 642 together with the motion vector value detected by the motion detection circuit 712. In the variable length coding circuit 642, variable length coding is performed on each input signal to obtain a coded LL1 component subband signal, which is stored in the buffer 644. Further, the transform coefficient value quantized by the quantizer 640 is also supplied to the inverse quantizer 646 to be subjected to inverse quantization processing. Then, the transform coefficient after the inverse quantization is input to the discrete inverse orthogonal transform circuit 648, where the inverse orthogonal transform processing such as IDCT (discrete cosine inverse transform) is performed, and the decoding difference of the subband signal of the LL1 component is obtained. can get. The subband signal of the LL1 component is restored by adding the prediction signal output from the motion compensation circuit 634 in the adder 650 to the decoding difference. The restored subband signal is stored in the frame memory 630 as a subband signal of the previous frame, and is used for motion detection of the image of the next frame.

On the other hand, HH1 and H other than the above LL1 components
For each of the L1 and LH1 subband signals, the motion vector of the LL1 component is used as it is as a motion vector used for motion compensation prediction. That is, the motion vector detected by the motion detection circuit 712 is the motion compensation circuit 634 of the encoding circuits 700, 702, 704.
Respectively. In the motion compensation circuit 634, in the decoded signal of the previous frame stored in the frame memory 710, the signal at the position shifted by the motion vector value is cut out as a prediction signal. And the LL described above
Similar to the processing for one component, the processing of the difference between subband signals, the discrete orthogonal transformation, the quantization of the transform coefficient, and the variable-length coding are performed, and the processed code is stored in the buffer 644. Note that the variable-length coding circuit 708 does not code the motion vector value, unlike the variable-length coding circuit 642.

[0013]

However, the above configuration has the following problems.

(1) When encoding is performed by the conventional technique shown in FIG. 8, the movements of the subband signal components other than the LL1 component do not necessarily match the natural visual movements, and L
There is a high possibility that a motion vector different from the L1 component will be detected. For this reason, the deterioration of the image quality becomes remarkable, such that the low-frequency component and the high-frequency component of the decoded image have different motions and unnatural motions may occur.

(2) Further, a huge amount of calculation is required for detecting the motion vector for each of the divided subband signals. In addition, the amount of motion vector to be transmitted also increases.

(3) When encoding is performed by the conventional technique shown in FIG. 9, the movement of the subband signal of the LL1 component and the movement of the subband signal other than the LL1 component do not necessarily match, so that in the case of FIG. Although it is not as good as the decoded image of, the image quality is deteriorated.

(4) Furthermore, since the motion differs for each sub-band signal, if the motion vector of the sub-band signal of the LL1 component is used as it is, between the prediction signal output from the motion compensation circuit and the sub-band signal of the current frame. The absolute value of the prediction error signal, which is the difference of, increases, and the code amount increases accordingly.

The present invention has been made in view of the above problems of the prior art.
An object of the present invention is to provide a moving picture coding apparatus that improves the accuracy of motion compensation to reduce visual image quality deterioration and also reduces the code amount.

[0019]

In order to solve the above problems, a moving picture coding apparatus of the present invention comprises means for dividing a digitized image signal into a plurality of bands and subband signals for each band. Means for performing motion compensation using the detected motion vector to generate a prediction signal, means for quantizing a prediction error signal that is the difference between the prediction signal and the subband signal, and this quantized signal In a moving picture coding device having a means for coding the motion vector, a motion detecting means for detecting a motion vector for a sub-band signal in the lowest frequency band, and a corresponding band and a minimum for a sub-band signal other than the lowest frequency band. A search range of the motion detecting means with reference to a position shifted by a value obtained by multiplying a ratio N (N: a positive integer) of the size with the frequency band by the detected motion vector. A reference calculating motion detecting means for detecting a motion vector by searching a remote small range, is characterized by comprising an encoding means for encoding the sub-band signals of each band by using the motion vector.

[0020]

According to the present invention, the motion vector detected for the subband signal in the lowest frequency band is
It is also used as a reference position for motion compensation of the obtained subband signal and motion detection for a subband signal other than the lowest frequency band. For this reason, coding using motion vectors having no correlation is not performed in each subband.

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the moving picture coding apparatus of the present invention will be described below with reference to the drawings.

(Embodiment 1) FIG. 1 is a block diagram showing a moving picture coding apparatus according to a first embodiment. In FIG. 1, reference numeral 10 denotes a band division circuit, which has the same configuration as that of FIG.
12, 14, 16, 18, 20, 22, and 24 are coding circuits, and the coding circuits 12, 14, 16, 18, and 20 have the same configuration as the coding circuit 22. The subband signals of the seven frequency components in FIG. 7 are input to each encoding circuit. That is, each subband (HH1, H
L1, LH1, HH2, HL2, LH2, LL2)
Encoding circuits 12, 14, 16, 18, 20, 2 respectively
2 and 24 are input. In the encoding circuit 22, the subband signal LH2 is connected to the input side of the reference calculation motion detection circuit 26 and the plus input side of the subtractor 28. The subtraction output side of the subtractor 28 is connected to the input side of the quantizer 30, and is supplied to the variable length coding circuit 32. The output side of the variable length coding circuit 32 is connected to the input side of the buffer 34. The output side of the reference calculation motion detection circuit 26 is connected to the input side of the motion compensation circuit 42 and the variable length coding circuit 3.
2 is connected to the other input side. The other output side of the quantizer 30 is connected to the dequantizer 36, and the output side of the dequantizer 36 is connected to the plus input side of the adder 38. The addition output side of the adder 38 is connected to the input side of the frame memory 40, and
The output side of is connected to the other input side of the reference calculation motion detection circuit 26 and the other input side of the motion compensation circuit 42. The output side of the motion compensation circuit 42 is connected to the minus input side of the subtractor 28 and the plus input side of the adder 38.

Next, the encoding circuit 24 is almost the same as the encoding circuit 2.
2 has the same configuration, and the same reference numerals are used for the same or corresponding components. A difference from the coding circuit 22 is a motion detection circuit 44, to which the subband signal LL2, which is the lowest frequency band of seven divisions, and the output side of the frame memory 40 are connected. Then, one output side of the motion detection circuit 44 has one of the coding circuits 12, 14, 16, 18, 2
0 and 22 are connected to the input side of the reference calculation motion detection circuit 26.

The operation of the moving picture coding apparatus configured as described above will be described below with reference to FIGS. 1, 2 and 3. First, FIG. 2 is a diagram for explaining the operation of the reference calculation motion detection circuit 26 in the subband LH2. 200
Is the LL2 component signal of the current frame, 210 is the LH2 component signal of the current frame, 220 is the block position for detecting the motion vector, 230 is the LL2 component signal of the previous frame, 24
0 is the LH2 component signal of the previous frame, 250 is the boundary indicating the search range of the motion detection circuit 44, 260 is the boundary indicating the search range of the reference calculation motion detection circuit 26, and 270 is the motion vector 291 detected by the motion detection circuit 44. The block position of the prediction signal shifted by the (symbol V) value, 280 is the reference block position shifted by the motion vector V, 290 is the motion vector 292 (symbol ∇V) value detected by the reference calculation motion detection circuit 26, and is 280. The block position of the prediction signal shifted from the reference block position, 293 is LH2
It is the motion vector of the block 220 in the component (symbol V ′). Next, FIG. 3 is a diagram for explaining the operation of the reference calculation motion detection circuit 26 in the subband LH1. The same reference numerals are used for the same or corresponding components as in FIG. 300 is the LH1 component signal of the current frame, 310 is the block position of the LH1 component for detecting the motion vector, 3
20 is the LH1 component signal of the previous frame, 330 is the boundary indicating the search range of the reference calculation motion detection circuit 26, 340 is the motion vector 391 that is twice the motion vector 291 (symbol 2).
V) is a reference block position shifted by a value, and 350 is a motion vector 39 detected by the reference calculation motion detection circuit 26.
The block position of the prediction signal shifted from the reference block position of 340 by 2 (symbol ∇V) value, and 393 is the motion vector (symbol V ′) of the block 310 in the LH1 component. The input image signal is processed by the band division circuit 10 into seven subbands (HH1, HL1, LH1, HH2, HL).
2, LH2, LL2). Of these, L
The L2 component subband signal is input to the motion detection circuit 44 of the encoding circuit 24. The subband signal of the decoded image of the previous frame stored in the frame memory 40 is also input to the motion detection circuit 44, and the motion vector is detected in block units from the subband signals between these frames. The detected motion vector is input to the motion compensation circuit 42 on the one hand. Motion compensation circuit 4
In No. 2, in the decoded signal of the previous frame stored in the frame memory 40, the block signal at the position shifted by the motion vector value is cut out as a prediction signal. Next, the prediction signal is input to the subtractor 28 together with the subband signal of the current frame output from the band division circuit 10. The subtractor 28 obtains the difference between the two subband signals, and the result is input to the quantizer 30. In the quantizer 30, the input prediction error signal is quantized, and the quantized value is supplied to the variable length coding circuit 32 together with the motion vector value detected by the motion detection circuit 44. In the variable length coding circuit 32,
Variable-length coding is performed on each input signal to obtain a coded LL2 component subband signal, and the buffer 3
Stored in 4. The value quantized by the quantizer 30 is also supplied to the inverse quantizer 36 to be subjected to the inverse quantization process, and the decoding difference of the subband signal of the LL2 component is obtained. The sub-band signal of the LL2 component is restored by adding the prediction signal output from the motion compensation circuit 42 in the adder 38 to the decoding difference. The restored subband signal is stored in the frame memory 40 as a subband signal of the previous frame and is used for motion detection of the image of the next frame.

On the other hand, HH1 and H other than the above LL2 components
For each of the L1, LH1, HH2, HL2, and LH2 subband signals, reference calculation is performed based on the position shifted by the value obtained by multiplying the motion vector of the LL2 component for each subband by the numerical values shown in Table 1 below. Motion detection circuit 26
Then, a motion vector is detected by searching a range smaller than the search range of the motion detection circuit 44. The method of determining the numerical values in Table 1 represents the ratio of the size of the corresponding subband to the LL2 component. That is, layer 2 (HH2, HL
(2, LH2, LL2) perform filtering in the horizontal direction and vertical direction and perform downsampling respectively, and further perform filtering in horizontal direction and vertical direction and perform downsampling respectively, so layer 1 (HH
1, HL1 and LH1) have only half the number of pixels in the horizontal and vertical directions. Therefore, each sub-band of layer 2 is twice as large as the LL2 component in both the horizontal and vertical directions. As an example, a motion vector detecting method for the LH2 component of layer 2 and the LH1 component of layer 1 will be shown. In FIG. 2, the motion detection circuit 44 of the encoding circuit 24 uses the LL2 component of the current frame and the frame memory 40.
Within the search range 250 from the block position 220 of the LL2 component of the previous frame stored in
To detect. Reference calculation of encoder 22 Motion detector 2
In 6, the range 260 smaller than the search range of 250 is searched for the motion vector (∇V) with the reference block position 280 shifted by the motion vector V value as a reference.
That is, the motion vector of the block 220 in LH2 is a value obtained from V ′ = V + ∇V, and the prediction signal is 29
The block position is 0. In FIG. 3, LH1 is LL
Since it is twice as large in both the horizontal and vertical directions as the two components, the motion vector 39 obtained by doubling the motion vector 291
Reference block position 34 shifted by 1 (symbol 2V) value
A range 330 smaller than the search range of 250 on the basis of 0
To find the motion vector (∇V). That is, L
The motion vector of the block 310 in H1 is V ′ = 2
It is a value calculated from V + ∇V, and the prediction signal is the block position of 350. Also, blocks 310, 3 of LH1
The sizes of 40 and 350 are twice as large as the block 220 of LL2 in both the horizontal and vertical directions. Other layers 2 (HH
2, HL2) and the layer 1 (HH1, HL1) subband components can also be subjected to the same processing as in FIG. 2 and FIG. 3 to obtain a motion vector to obtain a prediction signal.

[0026]

[Table 1]

Then, the encoding circuits 12, 14, 16, 1
The motion vector detected by the reference calculation motion detection circuit 26 of 8, 20, 22 is input to the motion compensation circuit 42 of each encoding circuit, and similarly, HH1, HL1, LH.
Predicted signals in each of the 1, HH2, HL2, and LH2 subbands are obtained. Then, similarly to the above-described processing for the LL2 component, the difference between subband signals, the quantization of the prediction error signal, and the variable-length coding processing are performed, and the processed code is stored in the buffer 34.

As described above, according to this embodiment, the motion vector detected in the subband signal of the LL2 component is also used as the reference position for motion detection with respect to the other subband signals, so that the motion compensation is performed. Accuracy is improved,
It is possible to reduce visual quality deterioration. Also, since the motion detection is performed for each subband signal, the absolute value of the prediction error signal, which is the difference between the prediction signal output from the motion compensation circuit and the subband signal of the current frame, becomes small, and the code amount It can also be reduced. Further, the motion vector values other than the LL2 component have a small search range, so that the code amount can be reduced.

(Embodiment 2) FIG. 4 is a block diagram showing a moving picture coding apparatus according to a second embodiment. In FIG. 4, reference numeral 10 denotes a band division circuit, which has the same configuration as that of FIG.
12, 14, 16, 18, 20, 22, and 24 are coding circuits, and the coding circuits 12, 14, 16, 18, and 20 have the same configuration as the coding circuit 22. In the encoding circuit 22, the reference calculation motion detection circuit 50 and the subtractor 2
8, quantizer 30, variable length coding circuit 32, buffer 3
4, inverse quantizer 36, adder 38, frame memory 4
0, the motion compensation circuit 42 is provided similarly to the encoding circuit 22 of the first embodiment. In the encoding circuit 24, the motion detection circuit 52, the subtractor 28, the quantizer 30, the variable length encoding circuit 32, the buffer 34, the dequantizer 36,
Adder 38, frame memory 58, motion compensation circuit 42
However, it is provided similarly to the encoding circuit 24 of the first embodiment. On the other hand, the output side of the LL2 component of the band division circuit 10 is
It is connected to the block determination circuit 54. The output side of the frame memory 58 in the encoding circuit 24 is also connected to the other input side of the block determination circuit 54. In the block determination circuit 54, the input signal is divided into blocks and read out, and the subband signal of the current frame and the frame memory 58 are read.
The sub-band signal of the decoded image of the previous frame read from is compared with the sub-band signal, and when there is no change, the identification code is supplied to the variable-length coding circuit 56 as an invalid block and otherwise as a valid block. Variable length coding circuit 5
In 6, variable length coding is performed on the identification code,
It is stored in the buffer 34. Also, the block determination circuit 5
The other output side of 4 is the motion detection circuit 52 of the encoding circuit 24.
Is connected to the other input side of the reference calculation motion detection circuit 50 of the encoding circuits 12, 14, 16, 18, 20, 22.

The operation of the moving picture coding apparatus configured as described above will be described below. The input image signal is
The band division circuit 10 allows seven subbands (HH1,
HL1, LH1, HH2, HL2, LH2, LL2). Of these, LL2, which is the lowest frequency band
The subband signal of the component and the subband signal of the decoded image of the previous frame stored in the frame memory 58 of the encoding circuit 24 are input to the block determination circuit 54. In the block determination circuit 54, the input signal is divided into blocks in units of 2 (pixels) × 2 (lines) and read out, and a sub-band signal of the current frame and a signal at the same position in the same block, which is blocked by the sub-band signal of the previous frame, are read. And the difference is determined to be equal to or less than a predetermined threshold. The identification code of the valid / invalid block is the motion detection circuit 52 of the encoding circuit 24 and the encoding circuit 1.
2, 14, 16, 18, 20, 22 are connected to the input side of the reference calculation motion detection circuit 50. It is also supplied to the variable length coding circuit 56. In the variable length coding circuit 56, the valid / invalid identification code is variable length coded and stored in the buffer 34. The size of each block in each subband means the subbands (HH2, HL2, L) of layer 2
H2) is 2 (pixels) × 2 (lines), which is the same as the LL2 component. Since the sub-bands of the layer 2 are filtered so that the number of pixels is further halved both in the horizontal and vertical directions with respect to the sub-band signal of the LL1 component, the sub-bands of the layer 1 (HH1, HL1, LH
In 1), the sub-band block of layer 2 is 2 (pixels)
When it is x2 (line), it is 4 (pixel) x 4 (line). Next, the LL2 component subband signal is input to the motion detection circuit 52 of the encoding circuit 24. Motion detection circuit 5
2, the valid / invalid identification code determined by the block determination circuit 54 is used, and if the information is invalid information, the subsequent coding process is not performed and the block (2 × 2) is a subband signal of the previous frame. It is used as it is as a subband signal of the current frame. On the other hand, in the case of valid information, a motion vector is detected between the subband signal of the current frame and the subband signal of the decoded image of the previous frame stored in the frame memory 58. The detected motion vector is input to the motion compensation circuit 42 on the one hand. In the motion compensation circuit 42, in the decoded signal of the previous frame stored in the frame memory 58, the signal at the position shifted by the motion vector value is cut out as a prediction signal. Next, the prediction signal is input to the subtractor 28 together with the subband signal of the current frame output from the band division circuit 10. The subtracter 28 obtains the difference between the two subband signals, and the result is supplied to the quantizer 30. In the quantizer 30, the input prediction error signal is quantized, and the quantized value is supplied to the variable length coding circuit 32 together with the motion vector value detected by the motion detection circuit 52. In the variable length coding circuit 32,
Variable-length coding is performed on each input signal to obtain a coded LL2 component subband signal, and the buffer 3
Stored in 4. The value quantized by the quantizer 30 is also supplied to the inverse quantizer 36 to be subjected to the inverse quantization process, and the decoding difference of the subband signal of the LL2 component is obtained. The sub-band signal of the LL2 component is restored by adding the prediction signal output from the motion compensation circuit 42 in the adder 38 to the decoding difference. The restored subband signal is stored in the frame memory 40 as a subband signal of the previous frame and is used for motion detection of the image of the next frame.

On the other hand, HH1 and H other than the above LL2 components
For each of the L1, LH1, HH2, HL2, and LH2 subband signals, the reference position for motion detection is calculated for each subband using Table 1 as in the first embodiment, and each reference calculation motion detection circuit 50 calculates. , A range smaller than the search range of the motion detection circuit 52 is searched for a motion vector. From the valid / invalid identification code of the block determination circuit 54, if valid information, motion compensation, inter-subband signal difference, and prediction error are performed in the same manner as the process for the LL2 component described above using the detected motion vector. Signal quantization,
The variable length coding process is performed, and the code after the process is stored in the buffer 34. On the other hand, in the case of invalid information, the subsequent encoding process is not performed, and the block uses the subband signal of the previous frame as it is as the subband signal of the input frame.

As described above, according to this embodiment, the motion vector detected in the subband signal of the LL2 component is also used as the reference position for motion detection with respect to the other subband signals, so that motion compensation is performed. Accuracy is improved,
Visual image quality deterioration can be reduced. Further, by using the identification code of the valid / invalid block of the block determination circuit, it is not necessary to perform the encoding process in the case of the invalid block, so that it is possible to shorten the code processing time and the code amount.

(Third Embodiment) FIG. 3 is a block diagram showing the moving picture coding apparatus according to the third embodiment. In FIG. 3, 10 is a band division circuit, 12, 14, 16, 18, 20, 22,
24 is an encoding circuit, 28 is a subtractor, 30 is a quantizer, 3
2, 56 are variable length coding circuits, 34 is a buffer, 36 is an inverse quantizer, 38 is an adder, 42 is a motion compensation circuit, 50
Is a reference calculation motion detection circuit, 52 is a motion detection circuit, 60 is a block determination circuit, and 58 is a frame memory. However, the block determination circuit 60 is input from both the frame memory 58 of the encoding circuit 22 and the frame memory 58 of the encoding circuit 24.

The operation of the moving picture coding apparatus configured as described above will be described below. The input image signal is
The band division circuit 10 allows seven subbands (HH1,
HL1, LH1, HH2, HL2, LH2, LL2). Among these, the subband signal of the LL2 component and the subband signal of the LH2 component and each subband signal of the decoded image of the previous frame stored in the frame memory 58 of the encoding circuit 24 and the frame memory 58 of the encoding circuit 22. Is input to the block determination circuit 60. In the block determination circuit 60, signals of the same position of each input signal are
Blocks are read out in units of (pixels) × 2 (lines), and the difference between the subband signal of the current frame and the subband signal of the previous frame is calculated for 4 pixels of each component of LL2 and LH2 to obtain a total of 8 pixels. It is determined whether or not the difference value is less than or equal to a predetermined threshold. The identification codes of the valid / invalid blocks are the motion detection circuit 52 of the encoding circuit 24 and the encoding circuits 12, 14, 16, 18,
The reference calculation motion detection circuits 50 and 20 are connected to the input side. It is also supplied to the variable length coding circuit 56. In the variable length coding circuit 56, the valid / invalid identification code is variable length coded and stored in the buffer 34. Then L
The L2 component subband signal is input to the motion detection circuit 52 of the encoding circuit 24. In the motion detection circuit 52, the valid / invalid identification code determined by the block determination circuit 60 is used, and if the information is invalid information, the subsequent coding process is not performed and the block (2 × 2) is not a subframe of the previous frame. The band signal is used as it is as the subband signal of the current frame. On the other hand, in the case of valid information, a motion vector is detected between the subband signal of the current frame and the subband signal of the decoded image of the previous frame stored in the frame memory 58. The detected motion vector is input to the motion compensation circuit 42 on the one hand. In the motion compensation circuit 42, in the decoded signal of the previous frame stored in the frame memory 58, the signal at the position shifted by the motion vector value is cut out as a prediction signal. Next, the prediction signal is input to the subtractor 28 together with the subband signal of the current frame output from the band division circuit 10. The subtracter 28 obtains the difference between the two subband signals, and the result is supplied to the quantizer 30. In the quantizer 30, the input prediction error signal is quantized, and the quantized value is supplied to the variable length coding circuit 32 together with the motion vector value detected by the motion detection circuit 52. In the variable-length coding circuit 32, variable-length coding is performed on each input signal to obtain a coded LL2 component subband signal, which is stored in the buffer 34. Further, the value quantized by the quantizer 30 is also supplied to the inverse quantizer 36 to be subjected to the inverse quantization processing L
The decoding difference of the L2 component subband signal is obtained. The sub-band signal of the LL2 component is restored by adding the prediction signal output from the motion compensation circuit 42 in the adder 38 to the decoding difference. The restored subband signal is stored in the frame memory 40 as a subband signal of the previous frame and is used for motion detection of the image of the next frame.

On the other hand, HH1 and H other than the above LL2 components
For each of the L1, LH1, HH2, HL2, and LH2 subband signals, the reference position for motion detection is calculated for each subband using Table 1 as in the first embodiment, and each reference calculation motion detection circuit 50 A motion vector is obtained by searching a range smaller than the search range of the motion detection circuit 52.
Then, from the valid / invalid identification code of the block determination circuit 60, if valid information, motion compensation, difference between subband signals, and prediction error are performed in the same manner as the process for the LL2 component described above using the detected motion vector. Signal quantization and variable length coding processing are performed, and the processed code is stored in the buffer 34. On the other hand, in the case of invalid information, the subsequent encoding process is not performed, and the block uses the subband signal of the previous frame as it is as the subband signal of the input frame.

The present invention is not limited to the above-described embodiment. For example, (1) in the first embodiment, the band division circuit 10 performs the band division shown in FIG. In the case of FIG. 6, the LL2 component is further repeated to perform filtering in the horizontal and vertical directions, down-sampling is performed, and 10 band components (HH
1, HL1, LH1, HH2, HL2, LH2, HH
3, HL3, LH3, LL3). Further, it is obvious that this embodiment can be applied to other band division methods.

(2) In the second embodiment, the block determination circuit 54 is enabled / disabled by using the subband signal of the LL2 component.
Although the invalid block is determined, any other component may be used.

(3) In the third embodiment, the block decision circuit 60 decides the valid / invalid block by using the subband signals of the LL2 and LH2 components. However, even if a plurality of other components are used. Good.

(4) In this embodiment, the sub-band block of layer 2 is 2 × 2 and the sub-band block of layer 1 is 4 × 4, but other sizes may be used.

(5) In the above embodiment, the quantization and the variable length coding are performed after the motion compensation. However, the discrete orthogonal transform circuit is provided between the subtractor and the quantizer and between the inverse quantizer and the adder. Orthogonal transformation processing such as a discrete inverse orthogonal transformation circuit may be performed. Further, the discrete orthogonal transform, quantization, and variable length coding are not limited to these, and other methods may be used.

[0041]

As described above, according to the present invention, a motion vector is detected for a subband signal in the lowest frequency band, and the motion vector is used as a reference for motion detection for a subband signal other than the lowest frequency band. Since it is also used for position, since there is no coding with motion vectors that have no correlation between each subband signal, the accuracy of motion compensation is increased and visual image quality deterioration is reduced.
The code amount can also be reduced.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a moving picture coding apparatus according to a first embodiment of the present invention.

FIG. 2 is an operation explanatory diagram of a reference calculation motion detection circuit in layer 2.

FIG. 3 is an operation explanatory diagram of a reference calculation motion detection circuit in layer 1.

FIG. 4 is a configuration diagram of a moving picture coding apparatus according to a second embodiment of the present invention.

FIG. 5 is a configuration diagram of a moving picture coding apparatus according to a third embodiment of the present invention.

FIG. 6 is an explanatory diagram showing one method of band division.

FIG. 7 is an explanatory diagram showing one method of band division.

FIG. 8 is a block diagram of a conventional moving picture coding apparatus for band-split subband signals.

FIG. 9 is a schematic diagram of a conventional moving picture coding apparatus for band-divided subband signals.

[Explanation of symbols]

10 band division circuit 12, 14, 16, 18, 20, 22, 24, 622,
624, 626, 628, 700, 702, 704, 7
06 encoding circuit 26, 50 reference calculation motion detection circuit 28, 636 subtractor 30, 640 quantizer 32, 56, 642, 708 variable length encoding circuit 34, 644 buffer 36, 646 dequantizer 38, 650 addition 40, 58, 630, 710 Frame memory 42, 634 Motion compensation circuit 44, 52, 632, 712 Motion detection circuit 54, 60 Block determination circuit 610, 612 Horizontal band division filter 614, 616, 618, 620 Vertical direction Band division filter 400, 408, 412, 500, 508, 512, 5
24, 532, 536 High-pass band splitting filters 402, 410, 414, 502, 510, 514, 5
26, 534, 538 Low-pass band dividing filters 404, 406, 504, 506, 528, 530 Horizontal down-sampling circuits 416, 418, 420, 422, 516, 518, 5
20, 522, 540, 542, 544, 546 Vertical downsampling circuit 200 LL2 component signal of current frame 210 LH2 component signal of current frame 220, 310 Block position for detecting motion vector 230 LL2 component signal of previous frame 240 Previous LH2 component signal of frame 250, 260, 330 Boundary 270, 290, 350 indicating search range Block position of prediction signal, 280, 340 Reference block position, 291, 292, 293, 391, 392, 393 Motion vector value 300 Current frame LH1 component signal of 320 320 LH1 component signal of the previous frame

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical indication H03M 7/36 9382-5K

Claims (3)

[Claims] 1. A means for dividing a digitized image signal into a plurality of bands, a means for performing motion compensation using a detected motion vector for a subband signal of each band, and generating a prediction signal, and prediction. In a moving picture coding apparatus having means for quantizing a prediction error signal which is a difference between a signal and a subband signal, and means for coding the quantized signal and the motion vector, Motion detection means for detecting a motion vector for a subband signal, and a size ratio N (N: positive integer) between the applicable band and the lowest frequency band for subband signals other than the lowest frequency band
And a reference calculation motion detection means for detecting a motion vector by searching a range smaller than the search range of the motion detection means on the basis of a position shifted by a value multiplied by the detected motion vector, and the motion vector An encoding device for encoding a sub-band signal of each band using the encoding device. 2. A means for dividing a digitized image signal into a plurality of bands, and motion compensation using a motion vector detected only for an effective block for a sub-band signal of each band to obtain a prediction signal. Means for generating, means for quantizing a prediction error signal which is a difference between the prediction signal and the subband signal, and coding for the quantized signal, the motion vector and identification information of valid or invalid block. In a video encoding device having means, a block determining means for determining whether each subblock signal is valid or invalid by using an arbitrary subband signal, and a block determined to be a valid block for the subband signal in the lowest frequency band. The motion detection means for detecting the motion vector only for the sub-band signal other than the lowest frequency band A range smaller than the search range of the motion detecting means is searched with reference to a position shifted by a value obtained by multiplying the ratio N of the frequency band and the detected motion vector (N: a positive integer). A moving picture coding apparatus comprising: a reference calculation motion detecting means for detecting a motion vector; and a coding means for coding a subband signal of each band using the motion vector. 3. The moving picture coding apparatus according to claim 2, wherein the block judging means judges whether the block is valid or invalid by using subband signals of a plurality of bands.