JP3021852B2 - Image coding device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image encoding apparatus, and more particularly to an image encoding apparatus which combines a subband encoder and a vector quantizer used for digital video information communication and recording of digital video information, for example. The present invention relates to an encoding device.

[0002]

2. Description of the Related Art Subband coding is described, for example, in the Journal of the Institute of Television Engineers of Japan, Vol. 44, no. 2 (19
90), pages 154 to 156. Also,
The quantization method by vector quantization will be briefly described below. Vector quantization, L-number input signal to one set of M in the L-dimensional Euclidean space R ^L of the input vector _{x = (x i0, x i1} , x i2, ..., x iL-1), i =
, M−1, and N input vectors x are prepared in advance as N reproduced vectors y = (y _j0 , y _j1 , y _j2 ,
.., Y _jL-1 ), j = 0, 1,..., N−1, and is mapped to a reproduction alphabet (code book) Y, and the index is output. On the receiving side, it is possible to decode the encoded index to obtain a reproduced vector including an error with respect to the input vector. Note that the reproduction alphabet is also prepared in advance on the receiving side. Mapping Q from input vector to playback alphabet
(X) is expressed as Equation 1 by dividing the L-dimensional Euclidean space R ^L into N subspaces [S _j ], j = 0, 1,..., N−1.

[0003]

(Equation 1)

[0004] A distortion measure d (x, y) by quantization when an input vector x is mapped to a reproduction vector y is defined as in the following equation (2).

[0005]

(Equation 2)

At this time, the average distortion error per sample in the vector quantization is given by Equation 3.

[0007]

(Equation 3)

Generally, in vector quantization, a reproduction vector is determined so that the average distortion error is minimized as much as possible. The method of combining the sub-band coding and the vector quantization is described on page 7-26 of the 7th volume of the IEICE 1990 Spring National Convention. That is, a first band in which both the horizontal component and the vertical component of the image have passed through the low-pass filter, the high-pass filter passes through the horizontal component, and the low-pass filter passes through the vertical component. The second band, the low-pass filter is passed through the horizontal component, and the high-pass filter is passed through the vertical component. The third band and both the horizontal component and the vertical component pass through the high-pass filter. After division into four bands of the fourth band, vector quantization of each band is performed. Further, by passing the components of the first band through a filter having the same configuration, the first band can be divided into two dimensions and four parts. In this case, the band is divided into seven parts, and a high-quality image with a good SN ratio is obtained.

[0009]

As described above, when each of the band-divided data is vector-quantized, since the correlation characteristics differ depending on the direction of each band, a separate codebook is required for each band. There is a drawback that the configuration is complicated because it is necessary to prepare.

[0010] Therefore, a main object of the present invention is to provide an image coding apparatus capable of simplifying the configuration by sharing a code book.

[0011]

According to a first aspect of the present invention, in a horizontal and vertical frequency domain of an image signal, at least two filters are provided by a horizontal high-pass filter and a horizontal low-pass filter, and a vertical high-pass filter and a vertical low-pass filter, respectively.
Sub-band encoding means for generating at least four band data by performing dimensional frequency division; first band data and a horizontal low-pass filter which have passed through a horizontal high-pass filter and a vertical low-pass filter among at least four band data; Vectorization means for vectorizing one of the second band data that has passed through the vertical high-pass filter with respect to the other so that the vector configuration order of pixels in the time domain in the horizontal direction and the vertical direction is reversed. A plurality of vector quantization coding means for performing vector quantization coding on the first and second band data in accordance with a vector output from the means, a first codebook shared by the plurality of vector quantization coding means, A plurality of vector quantization decoding means for performing vector quantization decoding of an output of the vector quantization coding means, The second codebook shared by the decoding means, and one of two vectors representing the first band data and the second band data output from the plurality of vector quantization decoding means are horizontally arranged with respect to the other. An image encoding apparatus includes an inverse vectorization unit that performs inverse vectorization so that the vector configuration order of pixels in a time domain in a direction and in a vertical direction is reversed.

A second invention provides a first horizontal high-pass filter, a first horizontal low-pass filter, a first vertical high-pass filter, and a first vertical low-pass filter in the horizontal and vertical frequency domains of an image signal, respectively. A first sub-band encoding unit that performs at least two-dimensional frequency division by a band filter to generate at least four band data;
And a second vertical low-pass filter and a second vertical high-pass filter and a second vertical low-pass filter, wherein the first horizontal low-pass filter and the first A second sub-band encoder for dividing the band data passed through the vertical low-pass filter into four band data, wherein the first sub-band encoder includes a first horizontal high-pass filter and a first A set of band data that has passed through the vertical high-pass filter and band data that has passed through the second horizontal high-pass filter and the second vertical high-pass filter in the second sub-band encoding means is referred to as first band data. And band data that has passed through the first horizontal high-pass filter and the first vertical low-pass filter in the first sub-band encoder and the second horizontal band in the second sub-band encoder. A set of band data that has passed through the bandpass filter, the second horizontal high-pass filter, and the second vertical low-pass filter is referred to as a second band data set, and the first subband encoding unit performs the first horizontal band Of the band data passed through the low-pass filter and the first vertical high-pass filter and the band data passed through the second horizontal low-pass filter and the second vertical high-pass filter in the second sub-band encoding means Third pair
Vector quantization coding means for performing vector quantization coding on two band data included in one of the first, second and third sets, and two vector quantizations This is a pixel encoding device including a code book shared by encoding means.

A third aspect of the present invention is a pixel encoding apparatus combining the first and second aspects of the present invention.

[0014]

According to the first aspect of the invention, since the image data is substantially isotropic when viewed two-dimensionally, the band data that has passed through the horizontal high-pass filter and the vertical low-pass filter in the sub-band encoding means is combined with the horizontal low-pass filter. For the band data that has passed through the bandpass filter and the vertical high-pass filter, the vector quantization means converts the input vector input to the vector quantization coding means for performing vector quantization coding on each band data on the coding side. In the configuration stage, for either one of the band data, by configuring the vector configuration order of the pixels in the time domain so that the horizontal direction and the vertical direction are opposite to each other, each input vector has almost the same statistical property. Will have. As a result, the same codebook can be shared by the vector quantization encoding means for each band data. Accordingly, on the decoding side, the same codebook can be shared by the vector quantization decoding means for each band data. Then, at the time of vectorization on the encoding side, inverse vectorization is performed so that the vector configuration order of pixels in the time domain is reversed in the horizontal and vertical directions.

In the second invention, in the sub-band encoding means, the band data that has passed through the horizontal low-pass filter and the vertical low-pass filter at least once can be regarded as a reduced original image or original band data. Therefore, it can be considered that the original image or the original band data has almost the same contents. In other words, if the filter finally passes through the filter having the same configuration in the horizontal region and passes through the filter having the same configuration in the vertical region, it passes through the combination of the horizontal low-pass filter and the vertical low-pass filter several times in the preceding stage. Can be regarded as having almost the same contents, and when they are vectorized, they have almost the same statistical properties. Therefore, the same codebook can be shared when performing vector quantization encoding or decoding.

According to a third aspect of the present invention, as in the first aspect, vector quantization is performed on band data passed through the horizontal high-pass filter and vertical low-pass filter and band data passed through the horizontal low-pass filter and the vertical high-pass filter. If the codebook can be shared by the encoding and coding means, and if the codebook finally passes through the filter having the same configuration in the horizontal region and passes through the filter having the same configuration in the vertical region as in the second invention, Regardless of how many times the horizontal low-pass filter and the vertical low-pass filter are passed in the previous stage, the codebook can be shared by the vector quantization coding means for vector-quantizing each of them. Accordingly, the same codebook can be shared for vector quantization encoding or decoding for all band data that has finally passed the combination of the high-pass filter and the low-pass filter.

[0017]

According to the present invention, since the code book can be shared, the configuration can be simplified. The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description of embodiments with reference to the drawings.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Referring to FIGS. 1 to 4, an image encoding apparatus 10 of this embodiment includes a subband encoder 12, a vector quantization encoder 14, a vector quantization decoder 16, and a subband decoder. And a chemical converter 18. The image data is first input from the terminal 20 of the subband encoder 12 and is input to a horizontal high-pass filter 22 and a horizontal low-pass filter 24.
Outputs from the horizontal high-pass filter 22 and the horizontal low-pass filter 24 are input to 2: 1 sub-samplers 26a and 26b in the horizontal direction, respectively. The output from the sub-sampler 26a is guided to a vertical high-pass filter 28a and a vertical low-pass filter 30a, and the respective outputs are converted to vertical 2: 1 sub-samplers 32a and 32a.
The signal is output to terminals 34a and 34b via b. The output from the sub-sampler 26b is input to the vertical high-pass filter 28b and the vertical low-pass filter 30b, and is output to the terminals 34c and 34d via the vertical 2: 1 sub-samplers 32c and 32d. Accordingly, the band data of the horizontal high band vertical high band component is output to the terminal 34a, and the band data of the horizontal high band vertical low band component is output to the terminal 34b. Also, the terminal 34c has
The band data of the horizontal low band vertical high band component is output, and the terminal 3
The band data of the horizontal low-frequency component and the vertical low-frequency component is output to 4d.

Of these outputs, band data output to terminals 34b and 34c are input to input terminals 36a and 36b of vector quantization encoder 14 shown in FIG. 1, and input to vectorizers 38a and 38b. Vectorized. Either of the vectorizers 38a and 38b forms a vector such that the vector configuration order of the pixels in the time domain is opposite to each other in the horizontal and vertical directions. This is shown in FIG.

The band data vectorized by the vectorizer 38a is input to a vector quantization encoder 40a. Then, the band data vectorized by the vectorizer 38b is input to the vector quantization encoder 40b. In the vector quantization encoders 40a and 40b,
Each band data is vector-quantized coded, and one band data is configured such that the vector configuration order of the pixels in the time domain at the stage of vectorization is opposite to each other in the horizontal direction and the vertical direction. Therefore, each band data has almost the same statistical properties. Therefore, the same codebook 42a can be shared during vector quantization coding. Then, from the vector quantization encoders 40a and 40b, the codebook 4 is output to output terminals 44a and 44b, respectively.
The index data indicating the vector of each band data extracted from 2 is output.

The respective index data output from output terminals 44a and 44b are input to input terminals 46a and 46b of vector quantization decoder 16 shown in FIG. The index data input from the input terminal 46a is input to the vector quantization decoder 48, and the index data input from the input terminal 46b is input to the vector quantization decoder 48b. The vector quantization decoders 48a and 48b perform vector quantization decoding based on the input index data. At this time, the same codebook 42b can be shared. Vector quantization decoders 48a and 48
The output of b is input to inverse vectorizers 50a and 50b, respectively. The inverse vectorizers 50a and 50b are arranged such that the configuration order of the inverse vectorization of the pixels in one of the time domains is horizontal and vertical so as to correspond to the operation of the vectorizers 38a and 38b of the vector quantization encoder 14. The inverse vectorization is performed by performing the reverse operation of the vectorizers 38a and 38b so that the directions are reversed. Then, the band data of each of the output terminals 52a and 52b is output from the inverse vectorizers 50a and 50b.

The respective band data output from the output terminals 52a and 52b of the vector quantization decoding device 16 are supplied to the input terminal 5 of the subband decoding device 18 shown in FIG.
4b and 54c. In addition, the input terminals 54a and 54d of the sub-band decoding device 18 receive band data output from the output terminals 34a and 34d of the sub-band coding device 12 and passed through any coding device and decoding device, respectively. You. Input terminals 54a to 54
The band data input from d are respectively 1:
It is input to two interpolators 56a to 56d. Vertical 1:
The band data is input to the vertical high-pass filter 58a via the two interpolators, and the band data is input to the vertical low-pass filter 60a via the 1: 2 interpolator 56b in the vertical direction. Also,
The band data is input to the vertical high-pass filter 58b via the 1: 2 interpolator 56c in the vertical direction, and the band data is input to the vertical low-pass filter 6 via the 1: 2 interpolator 56b in the vertical direction.
0b. Then, the output of the vertical high-pass filter 58a and the output of the vertical low-pass filter 60a are combined and passed through a 1: 2 interpolator 62a in the horizontal direction.
4 is input. The output from the horizontal high-pass filter 58b and the output from the vertical low-pass filter 60b are combined, and input to the horizontal low-pass filter 66 via the horizontal 1: 2 interpolator 62b. And the horizontal high-pass filter 6
4 and the output of the horizontal low-pass filter 66 are synthesized,
A reproduced image signal is output to the output terminal 68.

In the subband encoder 12 and the subband decoder 18, it is needless to say that the configuration order of each filter in the horizontal frequency domain and each filter in the vertical frequency domain may be reversed in the horizontal and vertical directions. Absent. Also,
In the case of vector quantization, especially when the number of dimensions of the vector is N ² (N
Is a natural number), the vector quantization encoding apparatus 14 can share the codebook 42a only by rotating it by 90 ° in one of the vectorization stages at the vectorizers 38a and 38b. The codebook 42a can also be shared by rotating one of the vector data extracted from the codebook 42a and input to the vector quantization encoders 40a and 40b by 90 °. Similarly, in the vector quantization decoding apparatus 16, when inverse vectorization is performed by the inverse vectorizers 50a and 50b, only one of them needs to be rotated by 90 °. The codebook 42b can be shared by rotating one of the vector data extracted from the codebook 42b and input to the vector quantization decoders 48a and 48b by 90 °.

Next, referring to FIG. 6, image coding apparatus 70 of another embodiment includes subband coding apparatuses 72a and 72a.
2b, a vector quantization encoder 74, a vector quantization encoder 76, and subband decoders 78a and 78b. The sub-band encoders 72a and 72b have the same configuration as the sub-band encoder 12 described above, and the sub-band decoders 78a and 78b have the same configuration as the sub-band decoder 18 described above.

The image data is input to an input terminal of the image encoding device 70.
And input to the sub-band encoder 72a.
Is done. The sub-band coding device 72a is
Since the configuration is the same as that of the
As shown in FIG. 6, a horizontal high-frequency
Band data (H _H
H_V), Pass through horizontal high-pass filter and vertical low-pass filter
Bandwidth data (H_HL_V), Horizontal low-pass filter and vertical
Band data (L_HH _V) And
And bands that have passed through the horizontal low-pass filter and the vertical low-pass filter
Area data (L_HL_V) 4 band data is output
You. Of these, L_HL_VIs a sub-band encoding device
72b. The subband encoder 72b is
The configuration is the same as that of the subband encoder 18 described above.
Thus, the horizontal height is further increased from the subband encoder 72b.
Band data passed through the high-pass filter and the vertical high-pass filter
(L_HL_VH_HH_V), Horizontal high-pass filter and vertical
Band data (L_HL_VH_HL
_V), And a horizontal low-pass filter and a vertical high-pass filter.
Passed band data (L_HL_VL_HH_V) And more
Bands passed through horizontal low-pass filter and vertical low-pass filter
Data (L_HL_VL_HL_V) 4 band data
Each is output. Therefore, the subband encoder 7
A total of 7 bands by the combination of 2a and 72b
Data is created.

Of these seven band data, L _H L _V L _H
The band data other than L _V is converted to the vector quantization coding device 74.
Is input to Vector quantization coding apparatus 74 as shown in FIG. 7, _{H H H V} and L _H L _{V H H H V} each vector quantization coding to the vector quantization coding apparatus 74a, _H H L _V and L _H L _{V H H} L _V vector quantization encoder for vector quantization coding, respectively 74
composed of b and L _H H _V and L _H L _V L _{H H V} a vector quantization coding apparatus 74c to vector quantization coding.

In the vector quantization coding device 74a,
Is H_HH_VAnd L_HL_VH_HH _VAre the input terminals
And input to the vectorizers 84a and 82b,
And 84b and are respectively vectorized.
The output of the vectorizer 84a is the vector quantization encoder 8
6a and the output of the vectorizer 84b is a vector
It is input to the quantization encoding 86b. Vector quantization code
In the converters 86a and 86b_HH_VAnd L_HL_VH
_HH_VAre respectively vector-quantized coded, and L_H
L_VH_HH_VIs H_HH_VCan be regarded as a reduced image of
H _HH_VHas the same vector components as
Can be considered. Therefore, the vector quantization code
Sharing the same codebook 88a for encoding
Can be. Then, the vector quantization encoder 86a and the
And each band extracted from the codebook 88a
The index indicating the area data vector is output terminal 90a.
And 90b.

In the vector quantization encoder 74b,
Is H_HL_VAnd L_HL_VH_HL _VAre the input terminals
And input to the vectorizers 94a and 92b.
And 94b to be vectorized. vector
The outputs of the transformers 94a and 94b are vector quantum
Input to the coding encoders 96a and 96b. vector
H at the quantization encoders 96a and 96b_HL_Vand
L_HL_VH_HL_VAre respectively vector-quantized coded
But L_HL_VH_HL_VIs H_HL_VIt is considered as a reduced version
Because I can do it, H_HL_VHas the same vector components as
Then it can be considered. Therefore, the vector quantum
Share the same codebook 98a during coded encoding
be able to. Then, the vector quantization encoder 96a
And 96b extracted from codebook 98a
Index indicating vector of each band data is output port
It is output to 100a and 100b, respectively.

In the vector quantization coding device 74c,
L_HH_VAnd L_HL_VL_HH_VAre input terminals 1
02a and 102b and input to the vectorizer 104
a and 104b and are vectorized. Baek
The outputs of the converters 104a and 104b are vector
Input to the tor quantization encoders 106a and 106b.
You. In the vector quantization encoders 106a and 106b
Is L_HH_VAnd L _HL_VL_HH_VAre each vector
L quantization coding,_HL_VL_HH_VIs L_HH_V
Can be regarded as a reduced version of_HH_VSimilar to
It can be considered to have vector components. But
Therefore, the same code block is used for vector quantization coding.
Can be shared.

Then, the vector quantization encoder 106a
From 106b and 106b, an index indicating the vector of each band data extracted from codebook 108a is output to output terminals 110a and 110b, respectively. The six indices output from the vector quantization encoding device 74 are input to the vector quantization decoding device 76 as shown in FIG. The vector quantization decoding device 76 is a vector quantization decoding device 76a that performs vector quantization decoding of the data that has been subjected to the vector quantization coding by the vector quantization coding 74a, and the vector quantization coding is performed by the vector quantization coding device 74b. A vector quantization decoding device 76b that performs vector quantization decoding on the quantized data and a vector quantization decoding device 76c that performs vector quantization decoding on the data that has been subjected to vector quantization encoding by the vector quantization encoding device 74c. Be composed.

In the vector quantization decoding device 76a, the output terminal 90a of the vector quantization coding device 74a is used.
And the index data output from 90b are input to input terminals 112a and 112b, respectively, and are input to vector quantization decoders 114a and 114b. In the vector quantization decoders 114a and 114b, vector quantization decoding is performed based on the input index data, and at this time, the same codebook 88b is shared. Then, vectors indicating band data are extracted from the codebook 88b based on the respective index data, and are respectively input from the vector quantization decoders 114a and 114b to the inverse vectorizers 116a and 116b.
Then, the band data de-vectorized by de-vectorizers 116a and 116b are output to output terminals 118a and 118a.
18b. Band data output from the output terminal 118a _{(H H H V} and L _H L _V H
_H H _V ) is input to the sub-band decoding device 78a, and the band data output from the output terminal 118b is input to the sub-band decoding device 78b.

In the vector quantization decoder 76b, the output terminal 100 of the vector quantization encoder 74b is used.
The index data output from a and 100b is input to input terminals 120a and 120b, and is input to vector quantization decoders 122a and 122b.
In the vector quantization decoders 122a and 122b,
Vector quantization decoding is performed based on the respectively input index data, and at this time, the same codebook 98b is shared. Then, vectors indicating band data are extracted from the codebook 98b based on the respective index data, and are respectively input from the vector quantization decoders 122a and 122b to the inverse vectorizers 124a and 124b. The band data de-vectorized by the de-vectorizers 124a and 124b are output to output terminals 126a and 126b.
b. The band data output from the output terminal 126a is input to the sub-band decoding device 78a, and the band data (H _H) output from the output terminal 126b.
L _V and L _H L _V H _H L _V ) are input to the sub-band decoding device 78b.

In the vector quantization decoder 76c, the output terminal 110 of the vector quantization encoder 74c is used.
a and 110b are input to input terminals 128a and 128b, respectively.
It is input to the vector quantization decoders 130a and 130b. Vector quantization decoders 130a and 130
In b, vector quantization decoding is performed based on the respectively input index data, and at this time, the same codebook 108b is shared. Then, based on the respective index data, the code book 108
b, vectors each indicating band data are extracted,
Each is a vector quantization decoder 130a and 13
0b is input to inverse vectorizers 132a and 132b. And inverse vectorizers 132a and 132
b is the output terminal 134a
And 134b. Output terminal 134
The band data output from the sub-band decoding device 7
Is input to 8a, the output terminal band data output from 134b (L _{H H V} and _{L H L V L H H V} ) is input to the sub-band decoding apparatus 78b.

The output from the sub-band encoder 72b is
L of the bandwidth data_HL_VL_HL _VIs any encoding
Sub-band decoding device 78b
Is input to In the sub-band decoding device 78b,
An operation reverse to that of the band encoding device 72b is performed. Sand
And four band data (L_HL_VH_HH_V, L_HL_VH_H
L_V, L_HL_VL_HH_VAnd L_HL_VL_HL_V) To 1
Combine into one. And the band data combined into one
(L_HL_V) Is further input to the sub-band decoding device 78a.
Power. In the sub-band decoding device 78a,
An operation reverse to that of the encoding device 72a is performed. That is, 4
Band data (H_HH_V, H_HL _V,
L_HH_VAnd L_HL_V) And output the playback image data.
Output to the output terminal 136.

In this embodiment, the vector quantization coding is performed for three combinations of band data using three vector quantization coding devices and three vectors, but the vector quantization is performed for all of them. No need to do. However, by performing vector quantization coding on a combination of three types of band data, more efficient coding can be realized.

Next, referring to FIG. 9, an image coding apparatus 140 according to still another embodiment includes a subband coding apparatus 72a.
And 72b, vector quantization encoding devices 142a and 142b, vector quantization decoding devices 144a and 144b, and subband decoding devices 78a and 78b.
8b. Subband encoders 72a and 72b
Has the same configuration as the above-described sub-band encoder 12, and the sub-band decoders 78a and 78b have the same configuration as the above-described sub-band decoder 18.

The image data is input to the input terminal 146 of the image encoding device 140, and is input to the sub-band encoding device 72a.
Is input to As in the previous examples from the sub-band coding apparatus _{72a, H H H V, H} H L V, L H H V and L _H L 4 single band data of _V is output. this house,
L _H L _V is input to the previous embodiment similar further sub-band coding apparatus 72b. From the sub-band coding apparatus _{72b, L H L V H H} H V, L H L V H H L V, L H L V L
_H H _V and L _H L _V L _H L 4 single band data of _V are output. Therefore, a total of seven band data are created by the combination of the sub-band encoders 72a and 72b.

Of the seven band data, H_HL_V, L
_HH_V, L_HL_VH_HL_VAnd L _HL_VL_HH_VOf 4
Band data is supplied to the vector quantization encoder 142a.
Is entered. Vector quantization encoder 142a
As shown in FIG. 10, the input terminals 148a, 148
b, 148c and 148d respectively contain band data H _H
L_V, L_HH_V, L_HL_VH_HL_VAnd L_HL_VL_H
H_V, And each of the vectorizers 150a, 150
It is input to 50b, 150c and 150d. Vect
The converters 150a to 150d output the input band data.
Data is vectorized. At this time, H_HL_VAnd L_H
L_VH_HL_VAnd or L_HH_VAnd L_HL_vL_HH_VWhen
Vector of pixels in the time domain when vectorizing
The composition order is such that the horizontal and vertical directions are opposite to each other
Is vectorized. This is shown in FIG.

Then, the vectorizers 150a to 150b
Are output to the vector quantization encoders 152a to 152d, respectively. At this time,
For the _H H L _V and _{L H L V H H L V} , L H L V H
Since _{the H} L _V can be regarded as one H _H L _V is obtained by reducing the bandwidth data of H _H L _V, regarded as a band data having substantially the same components. Therefore, the same codebook can be shared for vector quantization coding. Similarly, it is possible to share the same codebook also and L _H H _V and _{L H L V L H H V} . further,
In the vectorization stage, H _H L _V and L _H L _V H _H L _V
Combinations or L _H H _V in combination or L _H of H _V and L _H L _V L _H H _V either vector forming order of pixel in the time domain in one of mutually horizontal and vertical inverse of a combination of the , So that the vector quantization encoder 152a
152d can share the same codebook 154a. Then, the vector quantization encoder 152a
To 152d output index data of a codebook 154a indicating vectors of the respective band data to output terminals 156a to 156d, respectively.

The index data output from the output terminals 156a to 156d of the vector quantization encoding device 142a is used for the vector quantization decoding device 144a shown in FIG.
And input to the vector quantization decoders 160a to 160d, respectively. In the vector quantization decoders 160 to 160d, vector quantization decoding is performed based on the input index data. At this time, the same codebook 154b is shared. Then, the vector data extracted from the codebook 154b based on each index data from the vector quantization decoders 160a to 160d are input to the inverse vectorizers 162a to 162d, and are inversely vectorized. At this time, the vectorizers 150a to 150d of the vector quantization encoder 142a
Inverse vectorization is performed corresponding to the vectorization performed in. That is, in the vector quantization coding apparatus 142a, and H _H L _V and L _H L _V H _H combination configuration sequence horizontal directions of the vectors of the pixel in the time domain of the L _V and the vertical direction is reversed In this case, the inverse vectorization is performed in the inverse vectorizers 162a and 162c such that the configuration order of the inverse vectorization of the pixels in the time domain is reversed in the horizontal direction and the vertical direction. Will be On the other hand, when the combination of L _H H _V and L _H L _V L _H H _V is vectorized in the same manner as described above in the vector quantization encoding device 142a, the inverse vectorizers 162b and 162d , The same inverse vectorization as described above is performed. Then, the band data subjected to vector quantization decoding from the inverse vectorizers 162a to 162d are output to output terminals 164a to 164d, respectively.

H out of the seven band data shown in FIG._HH
_VAnd L_HL_VH_HH_VIs a vector quantization coding apparatus 1
42b. Band data H_HH_VAnd L_HL
_VH _HH_VIs the vector quantization encoding apparatus 1 shown in FIG.
42b input terminals 166a and 166b,
It is input to vectorizers 168a and 168b. Be
In the converters 168a and 168b, each band data is
Vectorized, each output is a vector quantization code
Input to the converters 170a and 170b, respectively. Be
In the vector quantization encoders 170a and 170b,
Vector quantization code for each input band data
Is performed, but L_HL_VH_HH_VIs H_HH_VShrink
H_HH_VVector similar to
Can be considered as having the following components: Therefore,
In vector quantization coding, the same codebook 1 is used.
72a can be shared. And vector quantum
Codebooks from coding encoders 170a and 170b
172a indicates a vector of each band data extracted from 172a
Index is output to output terminals 174a and 174b
Is done.

The index data output from the output terminals 174a and 174b of the vector quantization encoding device 142b is used for the vector quantization decoding device 14 shown in FIG.
4b are input to input terminals 176a and 176b, and input to vector quantization decoders 178a and 178b. Vector quantization decoders 178a and 178b
In this case, vector quantization decoding is performed based on the respectively input index data, but at this time, the same codebook 172b is shared. Then, vectors indicating the band data are extracted from the codebook 172b based on the respective index data, and the vectors are respectively expressed by the vector quantization encoders 178a and 178b.
Are input to inverse vectorizers 180a and 180b, respectively. Then, the inverse vectorizers 180a and 1
The band data de-vectorized at 80b is output to output terminal 18
2a and 182b. Output terminal 1
Band data output from 82a is input to sub-band decoding device 78a, and band data output from output terminal 182b is input to sub-band decoding device 78b.

The output from the subband encoder 72b is
L of the bandwidth data_HL_VL_HL _VIs any encoding
Sub-band decoding device 78b
Is input to In the sub-band decoding device 78b,
An operation reverse to that of the band encoding device 72b is performed. Sand
The four band data are combined into one. And one
The combined band data is further converted to a sub-band decoding device 7.
8a. In the sub-band decoding device 78a,
An operation reverse to that of the band band encoding device 72a is performed. sand
That is, the band data divided into four is synthesized and the reproduced image is reproduced.
The image data is output to the output terminal 184.

As a result, vector quantization can be performed with a simple configuration, and a reproduced image with a good SN ratio and good image quality can be obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a vector quantization encoding device forming a part of an image encoding device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a vector quantization decoding device forming a part of the image coding device according to one embodiment of the present invention;

FIG. 3 is a block diagram illustrating a sub-band encoding device that forms a part of the image encoding device according to one embodiment of the present invention;

FIG. 4 is a block diagram showing a sub-band decoding device which forms a part of the image coding device according to one embodiment of the present invention;

FIG. 5 is an illustrative view showing a state in which the vector configuration order of pixels in a time domain when the band data is vectorized is such that the horizontal direction and the vertical direction are opposite to each other;

FIG. 6 is a block diagram showing another embodiment of the present invention.

FIG. 7 is a block diagram showing a configuration of the vector quantization encoding device shown in FIG.

8 is a block diagram illustrating a configuration of the vector quantization decoding device illustrated in FIG.

FIG. 9 is a block diagram showing still another embodiment of the present invention.

FIG. 10 is a block diagram showing a configuration of the vector quantization encoding device shown in FIG.

11 is a block diagram illustrating a configuration of the vector quantization decoding device illustrated in FIG.

12 is a block diagram illustrating a configuration of the vector quantization encoding device illustrated in FIG.

13 is a block diagram showing a configuration of the vector quantization decoding device shown in FIG.

[Explanation of symbols]

10, 70, 140 ... image coding device 12, 72a, 72b ... subband coding device 14, 74, 142a, 142b ... vector quantization coding device 16, 76, 144a, 144b ... vector quantization decoding device 18 , 78a, 78b ... sub-band decoding devices 38a, 38b, 84a, 84b, 94a, 94b, 1
04a, 104b, 150a to 150d, 168a, 1
68b ... vectorizers 40a, 40b, 86a, 86b, 96a, 96b, 1
06a, 106b, 152a to 152d, 170a, 1
70b... Vector quantization encoders 42a, 42b, 86a, 88b, 98a, 98b, 1
08a, 108b, 154a, 154b, 172a, 1
72b Codebook 48a, 48b, 114a, 114b, 122a, 12
2b, 130a, 130b, 160a to 160d, 17
8a, 178b ... vector quantization decoders 50a, 50b, 116a, 116b, 124a, 12
4b, 132a, 132b, 162a to 162d, 18
0a, 180b ... inverse vectorizer

Continuation of the front page (56) References Takahiro Saito, Hiroshi Yasuda “Encoding method of still image” Journal of the Institute of Television Engineers of Japan, Vol. 44, no. 2. pp153-161 (1990) Kotaro Asai, Hirofumi Nishikawa, Atsumichi Murakami, "Image Coding Using Subband Vector Quantization" IEICE Spring 7-26 (1990) (58) Field (Int.Cl. ⁷ , DB name) H04N ^7/ 24-7/68 H04N 1/41 H03M 7/30 JICST file (JOIS)

Claims (4)

(57) [Claims] In the horizontal and vertical frequency domains of an image signal, at least two-dimensional frequency division is performed by a horizontal high-pass filter and a horizontal low-pass filter, and a vertical high-pass filter and a vertical low-pass filter, respectively.
Sub-band encoding means for generating one band data; of the at least four band data, first band data which has passed through the horizontal high-pass filter and the vertical low-pass filter, and the horizontal low-pass filter and the vertical high Vectorization means for vectorizing one of the second band data passed through the bandpass filter with respect to the other so that the vector configuration order of pixels in the time domain in the horizontal direction and the vertical direction is reversed. A plurality of vector quantization coding means for performing vector quantization coding on the first and second band data according to the vector output from the first, and a first shared by the plurality of vector quantization coding means.
A plurality of vector quantization decoding means for performing vector quantization decoding of the outputs of the plurality of vector quantization encoding means, and a second codebook shared by the plurality of vector quantization decoding means.
And the first band data and the second band data output from the plurality of vector quantization decoding units.
Image encoding means for inversely vectorizing one of the two vectors representing the band data of the other so that the vector configuration order of the pixels in the time domain in the horizontal direction and the vertical direction is reversed with respect to the other. apparatus. 2. In the horizontal and vertical frequency domains of the image signal, at least a first horizontal high-pass filter, a first horizontal low-pass filter, and a first vertical high-pass filter and a first vertical low-pass filter respectively. First sub-band encoding means for creating at least four band data by performing two-dimensional frequency division, and a second horizontal high-pass filter, a second horizontal low-pass filter, a second vertical high-pass filter and a second A second low-pass filter that further divides the band data that has passed through the first horizontal low-pass filter and the first vertical low-pass filter from the at least four band data into four band data. Wherein the first sub-band encoding means includes the first horizontal high-pass filter and the first vertical high-pass filter. A set of band data passed through the first band data and band data passed through the second horizontal high-pass filter and the second vertical high-pass filter in the second sub-band encoding means. And the band data that has passed through the first horizontal high-pass filter and the first vertical low-pass filter in the first sub-band encoding means and the second sub-band encoding means A second set of band data that has passed through the second horizontal high-pass filter, the second horizontal high-pass filter, and the second vertical low-pass filter, and the first sub-band code Band data having passed through the first horizontal low-pass filter and the first vertical high-pass filter in the encoding means, and the second horizontal low-pass filter and the second data in the second sub-band encoding means. Vertical high frequency A set of the band data which has passed through the filter as a third set of band data, the first, 2 included in one of the second and third set
A pixel coding apparatus comprising: two vector quantization coding means for performing vector quantization coding on one band data; and a code book shared by the two vector quantization coding means. 3. Vector quantization decoding means for performing vector quantization decoding of outputs of the two vector quantization coding means, and a codebook shared by the two vector quantization decoding means. An image coding apparatus according to claim 2. 4. An image coding apparatus including a combination of claim 1 and claim 2.