KR101713860B1 - Method and apparatus for quantization encoding and de-quantization decoding using trellis - Google Patents

Abstract

The present invention relates to a quantization coding and inverse quantization decoding method and apparatus using trellis, and unlike the TCQ index, information on a path is not encoded or decoded, Quantization coding and inverse quantization decoding are performed using a new index by classifying the quantization levels assigned to the co-sets so that the co-sets can be selected and assigning indexes.

Description

[0001] The present invention relates to a method and apparatus for quantization encoding and dequantization decoding using trellis,

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to quantization encoding and inverse quantization decoding, and more particularly, to a method and apparatus for quantizing or inverse quantizing using trellis.

Trellis quantization coding is a kind of vector quantization scheme, and a vector codebook required for coding is constituted of a scalar codebook corresponding to each element constituting a vector, and a convolutional encoder convolution coder to represent the trellis structure. A trellis path for optimal coding is searched using a viterbi algorithm, and the trellis quantization coding scheme has a much smaller characteristic than the unstructured vector quantization.

And to provide a method and apparatus for quantizing or dequantizing using a trellis.

According to an aspect of the present invention, there is provided a quantization encoding method comprising: detecting an index corresponding to an input in a Trellis Coded Quantization (TCQ) codebook; And entropy-encoding the detected index. In the inverse quantization, the TCQ codebook is entropy-coded in a predetermined state provided in a trellis with an index only, the quantization level to which the coset is allocated is classified and the index is allocated so that the coset can be selected.

According to an aspect of the present invention, there is provided an inverse quantization decoding method comprising: entropy decoding an index; Detecting co-sets included in the restored index in a TCQ codebook; Detecting in the trellis a coset that coincides with the cosets corresponding to the branch connecting the current state and the next states among the detected cosets; And detecting a quantization level corresponding to the recovered index and the detected coset in a TCQ codebook.

According to an aspect of the present invention, there is provided a method of configuring a quantizer, the method comprising: setting co-ssets to be used in a trellis and a TCQ codebook; Grouping the set co-ssets into co-sets that are incompatible with co-ssets assigned to a branch connected to a predetermined state; And classifying and indexing the quantization levels provided in the TCQ codebook using the grouped co-ssets.

According to an aspect of the present invention, there is provided an apparatus for encoding a quantized image, the apparatus comprising: an index detector for detecting an index corresponding to an input in a TCQ codebook; And an entropy coding unit for entropy-coding the detected index. In the inverse quantization, the TCQ codebook includes a quantization level to which a co-set is assigned so that a co-set corresponding to a specific branch can be selected in a predetermined state provided in the trellis, And an index is assigned.

According to an aspect of the present invention, there is provided an apparatus for decoding an inverse quantized signal, comprising: an entropy decoding unit for entropy decoding and restoring an index; A coset detector for detecting co-sets included in the restored index in a TCQ codebook; A path detector for detecting, in a trellis, a coset coinciding with coaches corresponding to a branch connecting the current state and the next states among the detected coaches; And a quantization level detector for detecting, in the TCQ codebook, a quantization level corresponding to the recovered index and the detected coset.

According to an aspect of the present invention, there is provided a quantizer configuration apparatus comprising: a coarse setting unit configured to set coils to be used in a trellis and a TCQ codebook; A coset configuration unit for grouping the set cosets into cosets that are not compatible with the cosets assigned to the branches connected to the predetermined state; And an indexing unit for classifying and indexing the quantization levels provided in the TCQ codebook using the grouped co-ssets.

1 is a block diagram illustrating an embodiment of a convolutional encoder corresponding to an 8-state trellis using a 4-co-set.
FIG. 2 conceptually illustrates an embodiment of an 8-state trellis using a 4-coset.
FIG. 3 is a conceptual diagram illustrating an embodiment of a 4-coset TCQ codebook.
FIG. 4 is a flowchart illustrating an embodiment of a method of constructing a quantizer according to a new indexing method according to the present invention.
FIG. 5 is a conceptual diagram of an embodiment of an 8-co-set TCQ codebook.
FIG. 6 is a block diagram of an embodiment of a convolutional encoder of a 16-state trellis code using a 4-co-set.
7 is a conceptual diagram of an embodiment of a 16-state trellis using an 8-coset.
8 is a block diagram of an embodiment of an 8-state trellis code convolutional encoder using an 8-co-set.
FIG. 9 is a conceptual diagram illustrating an embodiment of 8-state trellis.
10 is a block diagram of an embodiment of a convolutional encoder of a 16-state trellis code using an 8-co-set.
11 is a conceptual diagram illustrating one embodiment of a 16-state trellis using a 4-co-set.
FIG. 12 conceptually shows an embodiment of a 4-co-set TCQ codebook to which a quantization level is not assigned to '0'.
FIG. 13 conceptually shows an embodiment of an 8-co-set TCQ codebook to which a quantization level is not assigned to '0'.
14 is a conceptual diagram illustrating an embodiment of a 4-co-set TCQ codebook in which quantization level 0 is set as a dead-zone.
FIG. 15 is a flowchart illustrating an embodiment of a quantization encoding method according to the present invention.
16 is a flowchart illustrating an inverse quantization decoding method according to an embodiment of the present invention.
17 is a block diagram of an apparatus for constructing a quantizer according to a new index scheme according to an embodiment of the present invention.
FIG. 18 is a block diagram of an embodiment of a quantization encoding apparatus according to the present invention.
FIG. 19 is a block diagram of an embodiment of an inverse quantization decoding apparatus according to the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of a method and apparatus for quantization encoding and dequantization decoding according to the present invention will be described in detail with reference to the accompanying drawings.

The embodiment of the quantization encoding method and apparatus according to the present invention can perform quantization encoding using a convolution encoder, a trellis, and a trellis coded quantization codebook.

The convolutional encoder receives a path and selects any one of the branches from a predetermined state provided at a predetermined stage and branches connected to the next stage. 1 is a block diagram showing an embodiment of a convolutional encoder corresponding to an 8-state trellis using a 4-coset. Referring to FIG. 1, '0' or '1', which is a code indicating a path, is input through an input terminal x (n), and 'D' denotes a code indicating z0, z1 and z2 And outputs a code indicating a coset which is an identifier assigned to each branch via output terminals y1 (n) and y2 (n). For example, a coset corresponding to the codes output through the output terminals y1 (n) and y2 (n) in Fig. 1 can be implemented as shown in Table 1 below.

y1 (n) y2 (n)  Coset 0 0 D0 0 One D1 One 0 D2 One One D3

This convolutional encoder is provided corresponding to the path of the state provided in the trellis. The trellis is provided in a plurality of states and each state provided in a predetermined stage is connected to at least one state among a plurality of states provided in a next stage by a branch, It can be identified as a corset with branches.

For example, FIG. 2 conceptually shows an embodiment of 8-state trellis using a 4-coset corresponding to the convolutional encoder shown in FIG. In FIG. 2, the state is composed of eight states corresponding to 0 to 7, and in FIG. 2, the coset is composed of four cosets corresponding to D0, D1, D2 and D3, and is allocated to correspond to each branch. When a coset provided at the tip of two co-sits allocated to each state is detected in FIG. 2, the path of x (n) of the convolutional encoder of FIG. 1 becomes '0' A branch provided at the top of the branch is selected as a path, and when a corset provided at the rear end is detected, the path becomes '1' and a branch provided at the lower end of the two branches is selected as a path. If the state is 4 and the coset D0 is detected, the branch provided at the lower end corresponds to the coset provided at the subsequent stage, and the code 1 representing the path is selected.

1, 2 and Table 1, it is assumed that the initial state is '0', and codes '1', '1', and '1', which represent paths, are input to the convolutional encoder shown in FIG. (N) and y2 (n) by the convolutional encoder shown in FIG. 1 when '1' and '0' are input sequentially through the input terminal x Table 2 shows the states determined by the trellis shown in Fig. 2 according to the path input to the coset, x (n) corresponding to this code in Fig.

x (n) z2 z1 z0 State y1 y2 Coset One 0 0 0 0 One 0 D2 One 0 0 One One 0 0 D0 One 0 One One 3 0 One D1 One One One One 7 One One D3 0 One One One 7 0 One D1

Assuming that z0, z1, and z2 are all assigned '0' as initial values and 1 is input to x (n) as a code representing a path, the convolutional encoder shown in FIG. D2 'is selected as the corresponding coset in the table of Table 1 by outputting' 0 'from y2 (n), and in the state 0 provided in the trellis shown in FIG. 2, the branch is' D2' Quot; 1 " is selected as the next state. Then, when 1 is input to x (n), 'x' is shifted to z1, '1' is shifted to z0, and '0' 0 'is shifted to z2 so that' 0 'is output from y1 (n) and' 0 'is output from y2 (n) by the convolutional encoder shown in FIG. 'D0' is selected and 'D0' is selected as the branch and '3' is selected as the next state in the state 1 provided in the trellis shown in FIG. Similarly, if '1' is input to x (n), '0' is output from y1 (n) and '1' is output from y2 (n) Is selected. Next, when '1' is input to x (n), '1' is output from y1 (n) and '1' is output from y2 (n) to select coset 'D3' do. Finally, if '0' is input to x (n), '0' is output from y1 (n) and '1' is output from y2 (n).

Coasets provided in the trellis as shown in FIG. 2 are allocated to respective quantization levels provided in a Trellis Coded Quantization codebook. The TCQ codebook can determine a quantization level corresponding to each input, and a predetermined code set provided in the trellis corresponding to each quantization level is allocated. In the TCQ codebook, consecutive quantization levels are categorized in a predetermined unit, and an index is assigned to each class.

For example, FIG. 3 conceptually illustrates an embodiment of a 4-co-set TCQ codebook. 3, a TCQ index '0' is provided with a quantization level '0' to which the cosets D0 and D3 are assigned, and a TCQ index '-1' is provided with a quantization level '-1.5 -3.5 'to which the coset D0 is allocated, a quantization level' -4.5 'to which the coset D3 is allocated, and a quantization level' -4.5 'to which the coset D1 is allocated, Quot; + 4.5 ", the quantization level '+3.5' to which the coset D3 is assigned, and the quantization level '+ 4.5' to which the coset D0 is allocated.

The TCQ codebook is set so that different co-sits can be assigned to the quantization levels included in one index except the index including the quantization level '0'. In other words, the TCQ codebook classifies the quantization levels by configuring the quantization levels corresponding to the total number of cosets in one index. For example, when the total number of branches of the coset is four, the number of quantization levels provided in one index is also set to four, and a different coset is assigned to each quantization level. By assigning different co-sets to the quantization levels provided in one index, each quantization level can be identified in one index. The TCQ index corresponding to the input is detected in the TCQ codebook constructed in this manner, the path corresponding to the quantization level is detected and quantized in the trellis, and entropy coding is performed.

However, since the entropy encoding of the TCQ index and the path constituted by this method is inefficient in terms of entropy since the path is arbitrarily changed, it is necessary to consider the following new index method.

FIG. 4 is a flowchart illustrating an embodiment of a method of constructing a quantizer according to a new indexing method according to the present invention.

First, in step 400, the number of total co-ssets to be used in the trellis and the TCQ codebook is determined, and the co-sets are set correspondingly. In operation 400, the total number of cosets may be determined as 2 < n > (where n is an integer exceeding 2). For example, if the total number of cosets is determined as 4, four co-sits corresponding to D0, D1, D2 and D3 are set in the trellis and TCQ codebook as shown in Figs. 2 and 3, 8, the eight codets corresponding to D0, D1, D2, D3, D4, D5, D6 and D7 are set in the trellis and TCQ codebook.

The union coset is configured using the coset set in operation 400 (operation 410). Herein, the union coset constitutes co-ssets in which coaches assigned to a branch connected to a node of a certain state can not be compatible. In other words, the coaches assigned to the branch connected to the node of the predetermined state are not included in the same union cassette. If an embodiment in which the union coset is constituted by two co-sets as a unit is constituted by a table, it can be carried out as shown in Table 3 provided below.

Where N is an integer greater than 2 and K is the total number of cosets. D0-D2 and D1-D3 in the 4-co-set TCQ codebook and {D0-D4, D1-D5} and {D2-D6, D3- D7} can be used to construct a union coset.

In operation 420, quantization levels provided in the TCQ codebook are classified into predetermined units by using the union coset configured in operation 410, and indexed. In step 420, the quantization levels included in one index are indexed so that only the co-sets included in the other union co-ssets are allocated. For example, the quantization levels corresponding to D0 and D2 included in the same union cosset MUST not be included in any index in the 4-co-set TCQ codebook in which the union cosset is composed of {D0-D2, D1-D3} D3 is the same.

The first indexing method and the second indexing method, which will be described below, are an embodiment of the indexing method indexed in operation 420.

In the first index scheme, an index corresponding to a positive integer is allocated to quantization levels greater than '0', and an index corresponding to a negative integer is assigned to quantization levels lower than '0' And the indexes are symmetrically allocated with respect to '0' so that the absolute values of the quantization levels larger than '0' and the quantization levels assigned to '0' are the same but assigned differently from each other .

For example, there is a first index shown at the bottom of the 4-coset TCQ codebook shown in FIG. The first index is indexed with +1, +2, ..., + 8 with quantization levels greater than 0 with respect to '0' and quantization levels lower than '0' are indexed with -1, -2, ... ., And -8, and the indexes are symmetrically assigned to each other with the same absolute value but different signs from each other. Also, no index includes the quantization levels corresponding to D0 and D2 included in the same union cosset, and D1 and D3 are not provided in the same index. Accordingly, the first index includes quantization levels corresponding to one-half of the TCQ index as shown in FIG. 3 at one index.

The 4-co-located TCQ codebook can be indexed by the first indexing method using Equation 1 described below using the TCQ index scheme described above.

Here, n is the first index and t is the TCQ index.

Next, the second indexing method will be described. In the second index, an integer corresponding to a negative number is not allocated to an index, but only integers corresponding to '0' and positive integers are allocated to indexes, And sequentially allocates an index corresponding to a large positive integer as the quantization level gradually increases to allocate an index so that the largest index is included in the index including the largest quantization levels. The second indexing scheme is compared with the first indexing scheme. The first indexing scheme and the scheme for classifying the quantization levels are the same, and the order of sequentially indexing from the smallest quantization level to the largest quantization level is the same. However, in assigning an index, a first index is assigned a negative integer with an index smaller than '0' centered on an index including a quantization level 0, a positive integer is assigned with an index larger than '0' Quot; 0 " is allocated to the index including the smallest quantization levels and an index is assigned to the positive integer sequentially up to the index including the largest quantization levels.

For example, the 4-co-set TCQ codebook can be indexed by the second indexing method using Equation 2 described below using the TCQ index scheme described above.

Where n is the second index and t is the TCQ index.

If the TCQ index, the first index, and the second index are performed in the 8-state trellis shown in FIG. 2 and the 4-co-located TCQ codebook having the two zero levels shown in FIG. 3, Can be coped with. Table 4 shows the indexes for union coset C0 composed of cosets D0 and D2, and Table 5 is indexes for union coset C1 composed of cosets D1 and D3.

If the TCQ index, the first index, and the second index are performed in an 8-co-set TCQ codebook having no zero level, they can be matched as shown in Tables 6 to 9 below. Table 6 shows the indexes for the union cassette A0 composed of the cosets D0 and D4, and Table 7 is the indices for the union cassette A1 composed of the coats D1 and D5, and Table 8 shows the indexes for the union cassette A2 composed of the coats D2 and D6 And Table 9 are indices for the union cassette A3 consisting of the coats D3 and D7.

In indexing by the first index method and the second index method, the index including the quantization level '0' is indexed differently from the index including the quantization levels except for '0'. Quot; two zero level "in which two co-sets are assigned to the quantization level '0', a 'no zero level' in which no co-set is allocated, and a 'dead-zone' provided in the dead zone.

First, a 4-co-set TCQ codebook is shown in FIG. 3 as an example of a TCQ codebook in which two co-sets are assigned to a quantization level '0', and an 8-co-set TCQ codebook is shown in FIG. Here, the 4-coset TCQ codebook shown in FIG. 3 uses a convolutional encoder of an 8-state trellis code using the 4-coset shown in FIG. 1 and a 4-state trellis shown in FIG. 2 And can be performed using a convolutional encoder of a 16-state trellis code using the 4-co-set shown in FIG. 6 and a 16-state trellis shown in FIG. The 8-co-located TCQ codebook shown in FIG. 5 is obtained by using the 8-state trellis code convolutional encoder using the 8-co-set shown in FIG. 8 and the 8-state trellis shown in FIG. And can be performed by using a convolutional encoder of a 16-state trellis code using the 8-co-set shown in FIG. 10 and a 16-state trellis shown in FIG.

Second, a 4-co-set TCQ codebook is shown in FIG. 12 as an example of a TCQ codebook to which a quantization level is not assigned to '0', and an 8-co-set TCQ codebook is shown in FIG. Here, the 4-coset TCQ codebook shown in FIG. 12 is obtained by using the convolutional encoder of the 8-state trellis code using the 4-coset shown in FIG. 1 and the 4-state trellis shown in FIG. 2 And can be performed using a convolutional encoder of a 16-state trellis code using the 4-co-set shown in FIG. 6 and a 16-state trellis shown in FIG. The 8-co-located TCQ codebook shown in FIG. 13 is obtained by using the convolutional encoder of the 8-state trellis code using the 8-co-set shown in FIG. 8 and the 8-state trellis shown in FIG. 9 And can be performed by using a convolutional encoder of a 16-state trellis code using the 8-co-set shown in FIG. 10 and a 16-state trellis shown in FIG.

Third, a 4-co-set TCQ codebook is shown in FIG. 14 as an embodiment of a TCQ codebook in which quantization level 0 is set as a dead-zone. Here, the 4-co-set TCQ codebook shown in FIG. 14 is obtained by using the convolutional encoder of the 8-state trellis code using the 4-co-set shown in FIG. 1 and the 4-state trellis shown in FIG. And can be performed using a convolutional encoder of a 16-state trellis code using the 4-co-set shown in FIG. 6 and a 16-state trellis shown in FIG.

FIG. 15 is a flowchart illustrating an embodiment of a quantization encoding method according to the present invention.

First, the quantization level of the input value is detected in the TCQ codebook (operation 1500). For example, assuming that the convolutional encoder shown in FIG. 1, the trellis shown in FIG. 2, and the TCQ codebook shown in FIG. 12 are used for quantization into a first index, (0.5, -5.5, 0.5, 1.5, -0.5, 5.5, 7.5, -1.5) corresponding to each input are input to the input values when the input values are 0.1, 1.3, -0.9, 5.8, And detects it in the illustrated TCQ codebook.

The index including the quantization level detected in operation 1500 is detected in the TCQ codebook (operation 1510). The index detected in operation 1510 is the first index or the second index. In step 1500, the quantization levels (0.5, -5.5, 0.5, 1.5, -0.5, 5.5, 7.5, -1.5) detected in step 1500 are included 1 indexes (0, -3, 0, +1, 0, +3, +4, +4, -1) are detected in the TCQ codebook shown in FIG. These results are expressed in bit-planes as shown in Table 14 described below.

The indexes detected in operation 1510 are entropy-encoded (operation 1520). The TCQ index and the information indicating the path together with the TCQ index are also entropy-encoded. However, in step 1520, the information indicating the path is not subjected to entropy encoding separately, and only the first index or the second index is entropy- And performs encoding.

16 is a flowchart illustrating an inverse quantization decoding method according to an embodiment of the present invention.

First, in operation 1600, the bit streams transmitted from the encoding end are demultiplexed and entropy decoding is performed to restore the indices. The indices restored in operation 1600 are the first index or the second index described above.

The COs included in the index restored in operation 1600 are detected in the TCQ codebook (operation 1610).

In operation 1620, a coset coinciding with a coset allocated to a branch connecting a current state and a next state among the cosets detected in operation 1610 is detected. The branch corresponding to the coset detected in operation 1620 is determined as a path, and the node connected to the branch becomes the next state. Unlike the TCQ index, in step 1620, information on a path is not separately input from the encoding end. However, the reason why the path can be determined in the trellis is that, in indexing the quantization levels in the first index scheme and the second index scheme, The first index or the second index is assigned by setting the union co-set so that a specific branch can be selected. In other words, in the case of the first index method or the second index method, co-sets to which coaches assigned to a branch connected to a node of a predetermined state are not compatible with each other are configured, Are not included in the same union cassette, it is possible to detect the path without receiving information on the path separately from the decoding end.

In operation 1630, the quantization level corresponding to the coset detected in operation 1620 among the quantization levels included in the index reconstructed in operation 1610 is detected.

The flowchart shown in Fig. 16 will be described with reference to a concrete example. Here, the convolutional encoder corresponding to the 8-state trellis using the 4-coset shown in FIG. 1, the 8-state trellis shown in FIG. 2, the 4-coset TCQ codebook shown in FIG. 12, 1 index to perform inverse quantization decoding. Assume that the initial state of the 8-state trellis of FIG. 2 is set to '0' and the indexes restored in operation 1600 are (0, -3, 0, +1).

In step 1610, D0 and D3, which are included in the first index 0, are detected in the 4-co-located TCQ codebook shown in FIG. 13, and in step 1620, Among the cosets D0 and D3 detected in operation 1610, the coset D0, which is a coset coinciding with the cosets D0 and D2 corresponding to the branch which can be selected in the initial state 0 of the 8-state trellis shown in FIG. 2, In step 1630, inverse quantization is performed by detecting a quantization level of 0.5 corresponding to the coset D0 at the first index 0. In addition, since the coset D0 among the cosets D0 and D2 connected to the initial state 0 is selected in operation 1620, the state corresponding to D0 in the state 0 is the branch provided at the upper end, and the next state becomes zero.

In step 1610, D1 and D2, which are included in the first index-3, are detected in the 4-coset TCQ codebook shown in FIG. 13, and in step 1620, Among the cosets D1 and D2 detected in step 1610, a coset D2 which is a coset coinciding with the cosets D0 and D2 corresponding to the branch which can be selected in the state 0 of the 8-state trellis shown in FIG. 2 is selected, Quantization is performed by detecting the quantization level -5.5 corresponding to the coset D2 in the first index-3. In addition, since the coset D0 connected to the state 0 and the coset D2 selected in the middle of the D2 are selected in operation 1620, the state corresponding to D2 in the state 0 corresponds to the branch provided at the lower end.

In step 1610, COs D0 and D3 included in the first index 0 are detected in the 4-co-set TCQ codebook shown in FIG. 13, and in step 1620, Among the cosets D0 and D3 detected in the first stage of the 8-state trellis shown in FIG. 2, the co-set D0 corresponding to the branch which can be selected in the state 1 of the 8-state trellis shown in FIG. 1 from the index 0 by detecting the quantization level 0.5 corresponding to the coset D0. In addition, since the coset D0 connected to the state 1 and the coset D2 in the middle D2 connected to the state 1 are selected, the state corresponding to D2 in the state 1 is the branch provided at the lower end.

In step 1610, D1 and D2, which are included in the first index +1, are detected in the 4-co-set TCQ codebook shown in FIG. 13, and in step 1620, Among the cosets D1 and D2 detected in step 1610, the coset D1, which is a coset coinciding with the cosets D1 and D3 corresponding to the branch which can be selected in the state 3 of the 8-state trellis shown in FIG. 2, Quantization is performed by detecting the quantization level 1.5 corresponding to the coset D1 at the first index +1. In addition, since the coset D1 connected to the state 3 and the coset D1 located in the middle D3 connected to the state 3 are selected, the state corresponding to D1 in the state 3 is the branch provided at the lower end.

17 is a block diagram of an apparatus for constructing a quantizer according to a new indexing method according to the present invention. The apparatus constituting the quantizer includes a coset number determining unit 1700, a union coset generating unit 1710, And an indexing unit 1720.

The number-of-coats determining unit 1700 determines the total number of co-ssets to be used in the trellis and TCQ codebook, and sets the co-sets accordingly. In the number-of-coats determining unit 1700, the number of total co-sets can be determined as 2 < n > (where n is an integer exceeding 2). For example, if the total number of cosets is determined as 4, four co-sits corresponding to D0, D1, D2 and D3 are set in the trellis and TCQ codebook as shown in Figs. 2 and 3, 8, the eight codets corresponding to D0, D1, D2, D3, D4, D5, D6 and D7 are set in the trellis and TCQ codebook.

The union coset constituent unit 1710 constitutes a union coset using the coset set by the coset number determination unit 1700. [ Herein, the union coset constitutes co-ssets in which coaches assigned to a branch connected to a node of a certain state can not be compatible. In other words, the coaches assigned to the branch connected to the node of the predetermined state are not included in the same union cassette. If an embodiment in which the union coset is constituted by two co-sets as a unit is constituted by a table, it can be carried out as shown in Table 15 below.

Where N is an integer greater than 2 and K is the total number of cosets. D0-D2 and D1-D3 in the 4-co-set TCQ codebook and {D0-D4, D1-D5} and {D2-D6, D3- D7} can be used to construct a union coset.

The indexing unit 1720 classifies the quantization levels provided in the TCQ codebook into a predetermined unit and indexes the quantized levels using a union coset configured in the union coset generator 1710. In indexing by the indexing unit 1720, the quantization levels included in one index are indexed so that only co-sets included in different union co-ssets are allocated. For example, the quantization levels corresponding to D0 and D2 included in the same union cosset MUST not be included in any index in the 4-co-set TCQ codebook in which the union cosset is composed of {D0-D2, D1-D3} D3 is the same.

There are a first indexing method and a second indexing method which will be described below as an embodiment of an indexing method indexed by the indexing unit 1720.

In the first index scheme, an index corresponding to a positive integer is allocated to quantization levels greater than '0', and an index corresponding to a negative integer is assigned to quantization levels lower than '0' And the indexes are symmetrically allocated with respect to '0' so that the absolute values of the quantization levels larger than '0' and the quantization levels assigned to '0' are the same but assigned differently from each other .

For example, there is a first index shown at the bottom of the 4-coset TCQ codebook shown in FIG. The first index is indexed with +1, +2, ..., + 8 with quantization levels greater than 0 with respect to '0' and quantization levels lower than '0' are indexed with -1, -2, ... ., And -8, and the indexes are symmetrically assigned to each other with the same absolute value but different signs from each other. Also, no index includes the quantization levels corresponding to D0 and D2 included in the same union cosset, and D1 and D3 are not provided in the same index. Accordingly, the first index includes quantization levels corresponding to one-half of the TCQ index as shown in FIG. 3 at one index.

The 4-co-located TCQ codebook can be indexed by the first indexing method using Equation (3) described below using the TCQ index method described above.

Here, n is the first index and t is the TCQ index.

Next, the second indexing method will be described. In the second index, an integer corresponding to a negative number is not allocated to an index, but only integers corresponding to '0' and positive integers are allocated to indexes, And sequentially allocates an index corresponding to a large positive integer as the quantization level gradually increases to allocate an index so that the largest index is included in the index including the largest quantization levels. The second indexing scheme is compared with the first indexing scheme. The first indexing scheme and the scheme for classifying the quantization levels are the same, and the order of sequentially indexing from the smallest quantization level to the largest quantization level is the same. However, in assigning an index, a first index is assigned a negative integer with an index smaller than '0' centered on an index including a quantization level 0, a positive integer is assigned with an index larger than '0' Quot; 0 " is allocated to the index including the smallest quantization levels and an index is assigned to the positive integer sequentially up to the index including the largest quantization levels.

For example, the 4-co-located TCQ codebook can be indexed by the second indexing method using Equation 4 described below using the TCQ index scheme described above.

Where n is the second index and t is the TCQ index.

If the above-described TCQ index, first index and second index are performed in the 8-state trellis shown in FIG. 2 and the 4-co-set TCQ codebook provided in the two zero levels shown in FIG. 3, Can be coped with. Here, Table 16 is the indices for the union coset C0 consisting of the coats D0 and D2, and Table 17 is the indices for the union coset C1 consisting of the coats D1 and D3.

If the TCQ index, the first index, and the second index are performed in the 8-co-set TCQ codebook having no zero level, they can be matched as shown in Tables 18 to 21 below. Table 18 shows the indexes for the union cassette A0 composed of the cosets D0 and D4, Table 19 shows the indexes for the union cassette A1 composed of the coats D1 and D5, and Table 20 shows the indexes for the union cassette A2 composed of the coats D2 and D6 And Table 21 are the indices for the union cassette A3 consisting of the coats D3 and D7.

Lastly, if the TCQ index, the first index, and the second index are performed in the 8-co-set TCQ codebook provided with the two zero levels, it can be handled as shown in the following Tables 22 to 25. Table 23 shows the indexes for the union cassette A0 consisting of the coats D0 and D4 and Table 23 is the indexes for the union cassette A1 composed of the coats D1 and D5, And Table 25 are the indices for the union cassette A3 composed of the coats D3 and D7.

FIG. 18 is a block diagram of an embodiment of a quantization encoding apparatus according to the present invention. The quantization encoding apparatus includes a quantization unit 1800, an index detection unit 1810, and an entropy encoding unit 1820.

The quantization unit 1800 detects the quantization level of the value input through the input terminal IN in the TCQ codebook. Assuming that the convolutional encoder shown in FIG. 1, the trellis shown in FIG. 2, and the TCQ codebook shown in FIG. 12 are used for quantization into a first index, the quantization unit 1800 calculates (0.6, (0.5, -5.5, 0.5, 1.5, -0.5, 5.5, 7.5, -1.5) corresponding to each input when inputting the quantization level (-5.1, 0.1, 1.3, -0.9, 5.8, 12 in the TCQ codebook.

The index detector 1810 detects the index including the quantization level detected by the quantizer 1800 in the TCQ codebook. The index detected by the index detector 1810 is the first index or the second index described above. The index detector 1810 detects the quantization levels (0.5, -5.5, 0.5, 1.5, -0.5, 5.5, 7.5, -) detected by the quantization unit 1800 in the quantization unit 1800, (0, -3, 0, +1, 0, +3, +4, +4, -1) in the TCQ codebook shown in FIG. These results are expressed in bit-planes as shown in Table 26 described below.

The entropy encoding unit 1820 performs entropy encoding on the indexes detected by the index detection unit 1810 and outputs the indexes through an output terminal OUT. The entropy encoding unit 1820 performs the entropy encoding of the information indicating the path along with the TCQ index when the entropy encoding is performed on the TCQ index. However, in the entropy encoding unit 1820, the information indicating the path is not separately entropy- Index only entropy encoding is performed.

19 is a block diagram of an embodiment of an inverse quantization decoding apparatus according to the present invention and includes an entropy decoding unit 1900, a coset detecting unit 1910, a path detecting unit 1920, and a quantization level detecting unit 1930 .

The entropy decoding unit 1900 demultiplexes the bit stream received from the encoding end through the input terminal IN and performs entropy decoding to restore the indexes. The indexes restored by the entropy decoding unit 1900 are the first index or the second index.

The cousse detecting unit 1910 detects co-sets included in the index reconstructed by the entropy decoding unit 1900 in the TCQ codebook.

The path detecting unit 1920 detects a coset coinciding with a coset assigned to a branch connecting the current state and the next states among the cosets detected by the coset detecting unit 1910. The branch corresponding to the coset detected by the path detecting unit 1920 is determined as a path, and the node connected to the branch becomes the next state. Unlike the TCQ index, the path detector 1920 does not separately receive information on the path from the encoder, but the reason for determining the path in the trellis is that in indexing the quantization levels for the first index scheme and the second index scheme, And a first index or a second index is allocated by setting a union coset so that a specific branch can be selected in a predetermined state. In other words, in the case of the first index method or the second index method, co-sets to which coaches assigned to a branch connected to a node of a predetermined state are not compatible with each other are configured, Are not included in the same union cassette, it is possible to detect the path without receiving information on the path separately from the decoding end.

The quantization level detector 1930 detects the quantization level corresponding to the coset detected by the path detector 1920 among the quantization levels included in the index reconstructed by the coset detector 1910, and outputs the quantized values through the output terminal OUT .

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims. . Accordingly, the true scope of the present invention should be determined by the appended claims.

Furthermore, the present invention can be embodied as a computer-readable code on a computer-readable recording medium (including all devices having an information processing function). A computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer system is stored. Examples of computer-readable recording devices include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like.

1900: entropy decoding unit 1910: coset detection unit
1920: path detecting unit 1930: quantization level detecting unit

Claims (14)

Detecting an index corresponding to an input in a Trellis Coded Quantization (TCQ) codebook; And
And entropy-encoding the detected index,
Wherein the TCQ codebook comprises a branch information of a trellis state for selecting a predetermined coset and a quantization level assigned to the coset, Information and path information. 2. The method of claim 1,
A positive integer is assigned to quantization levels greater than '0', a negative integer is assigned to quantization levels less than '0', and symbols are assigned symmetrically with respect to '0' / RTI > 3. The method of claim 2, wherein the index is indexed based on a trellis structure having two zero levels. A computer-readable recording medium having recorded thereon a program for executing the quantization method according to any one of claims 1 to 3. At least one process, the process comprising:
A Trellis Coded Quantization (TCQ) codebook detects an index corresponding to an input,
Entropy-encodes the detected index,
Wherein the TCQ codebook comprises a branch information of a trellis state for selecting a predetermined coset and a quantization level assigned to the coset, Information and path information. 6. The method of claim 5,
A positive integer is assigned to quantization levels greater than '0', a negative integer is assigned to quantization levels less than '0', and symbols are assigned symmetrically with respect to '0' / RTI > 7. The quantization apparatus of claim 6, wherein the index is indexed based on a trellis structure having two zero levels. Restoring an index by entropy decoding;
Detecting co-sets included in the restored index in a TCQ codebook;
Detecting in the trellis a coset that coincides with the cosets corresponding to the branch connecting the current state and the next states among the detected cosets; And
Detecting in the TCQ codebook a quantization level corresponding to the recovered index and the detected coset,
Wherein the index includes information on the quantization level and path information. 9. The method of claim 8,
A positive integer is assigned to quantization levels greater than '0', a negative integer is assigned to quantization levels less than '0', and symbols are assigned symmetrically with respect to '0' / RTI > 10. The inverse quantization method of claim 9, wherein the index is indexed based on a trellis structure having two zero levels. A computer-readable recording medium on which a program for executing the inverse quantization method according to any one of claims 8 to 10 is recorded. At least one processor, wherein the processor
Entropy decoding is performed to restore the index,
Detecting co-sets included in the restored index in a TCQ codebook,
Detecting a co-set in the trellis that coincides with the co-sits corresponding to the branch connecting the current state and the next states among the detected co-
A quantization level corresponding to the recovered index and the detected coset is detected in the TCQ codebook,
Wherein the index includes information on the quantization level and path information. 13. The method of claim 12,
A positive integer is assigned to quantization levels greater than '0', a negative integer is assigned to quantization levels less than '0', and symbols are assigned symmetrically with respect to '0' Inverse quantization unit. 14. The inverse quantization apparatus of claim 13, wherein the index is indexed based on a trellis structure having two zero levels.

Images