CN102801427A - Encoding and decoding methods and systems of lattice vector quantization of variable rate of source signal - Google Patents

Encoding and decoding methods and systems of lattice vector quantization of variable rate of source signal Download PDF

Info

Publication number
CN102801427A
CN102801427A CN 201210279991 CN201210279991A CN102801427A CN 102801427 A CN102801427 A CN 102801427A CN 201210279991 CN201210279991 CN 201210279991 CN 201210279991 A CN201210279991 A CN 201210279991A CN 102801427 A CN102801427 A CN 102801427A
Authority
CN
Grant status
Application
Patent type
Prior art keywords
quantization
vector
lattice
lattice vector
obtain
Prior art date
Application number
CN 201210279991
Other languages
Chinese (zh)
Other versions
CN102801427B (en )
Inventor
张勇
Original Assignee
深圳广晟信源技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date

Links

Abstract

The invention relates to encoding and decoding methods and systems of lattice vector quantization of a variable rate of a source signal. The encoding method comprises the following steps of: S1, converting an input source signal into a frequency domain, so as to obtain a spectrum coefficient and control information; S2, grouping and carrying out bit distribution on the spectrum coefficient, so as to obtain bit distribution information; S3, carrying out lattice vector quantization on the spectrum coefficient based on the bit distribution information; and S4, packaging a quantized index, the bit distribution information and the control information into a bit stream. Compared with the traditional variable rate vector quantizer which can be used for storing a plurality of vector codebooks, the method provided by the invention has no need of storing the vector codebook. Furthermore, compared with the traditional vector quantization, the calculation complexity is greatly reduced due to a high-speed algorithm, namely the encoding method has the advantages of low calculation complexity and realization of the variable rate quantization.

Description

源信号变速率格矢量量化的编解码方法和系统 Variable rate trellis source signal vector quantization encoding and decoding methods and systems

技术领域 FIELD

[0001] 本发明涉及源信号编码领域,更具体地,涉及源信号变速率格矢量量化的编解码方法和系统。 [0001] The present invention relates to a signal encoding a source, and more particularly, relates to a method and system for encoding and decoding a source signal of the variable-rate lattice vector quantization.

背景技术 Background technique

[0002] 现有的数字源信号的编码通常采用的是变换编码,其将待编码的信号分成为帧的采样块,并采用诸如离散傅立叶变换、离散余弦变换等线性正交变换对每帧信号进行处理,求取变换系数,然后对变化系数进行量化,以进ー步提高压缩效果。 [0002] Existing digital source signals are usually encoded using transform coding is that the signal to be encoded is divided into sampled blocks of the frame, and linear orthogonal transform such as discrete Fourier transform, discrete cosine transform or the like for each frame signal processed transform coefficients is obtained, then variation coefficients are quantized, to further improve the intake ー compression.

[0003] 在量化方法中常用的ー种方法是矢量量化方法,在其中,将几个采样系数组在ー起形成ー个矢量,且以ー个码本项对每个矢量进行近似(量化)。 [0003] In the conventional quantization method ー method is vector quantization method, in which are formed several sample sets of coefficients in vectors ー ー from, and at a ー codebook entry for each vector is approximated (quantized) . 为量化输入矢量所选择的码本项通常是根据“距离最小”准则得出的码本中最近的邻点。 Quantizing the input vector is selected codebook entry is typically the nearest neighbor is "minimum distance" criterion derived in accordance with codebook. 在码本集合中增加更多的码本会增加比特率和复杂性,但会降低量化的平均失真。 Adding more codes present in the codebook set to increase in bit rate and complexity but reduces the average distortion of quantization.

[0004] 另ー方面,为了适应源的不断变化的特征,通常使用自适应比特分配。 [0004] Another aspect ー, to adapt to the changing characteristics of a source, adaptive bit allocation is generally used. 通过自适应比特分配,可使用不同的码本尺寸来量化源矢量。 Adaptive bit allocation, different codebook sizes to quantize the source vector. 在变换编码中,在不超过量化所有系数的可用比特数的最大值情况下,分配给源矢量的比特数通常取决于该矢量相对于同一帧中其他矢量的能量。 In transform coding, in the case where the maximum value does not exceed the number of available bits for all quantized coefficients, the number of bits allocated to a source vector typically depends on the vector with respect to the other in the same frame energy vector. 图I和图2详细描述了常见的变速率量化编、解码器的量化框图。 FIGS. I and 2 described in detail with common variable rate quantization encoding, a block diagram of the decoder quantizer. 图I和图2中示出的变速率量化编码器和解码器使用多个码本,它们通常具有不同的比特率,以量化源矢量X。 FIGS. I and shown in variable rate quantizer 2 encoder and decoder use multiple codebooks, they generally have different bit rates to quantize a source vector X. 通常通过对信号进行变换并获取所有的变换系数或其子集,来获得源矢量。 By generally transformed signal and acquires all or a subset of the transform coefficients to obtain the source vector.

[0005] 图I中示出了常见的变速率量化编码器,其关键部件是用Q表示的量化器,该量化器用于选择一个码本号n和一个码矢索引i来表征源矢量X的量化值I。 [0005] Figure I shows a common variable rate quantizer encoder, the key component is a quantizer represented by Q, the quantization for selecting a codebook number n, and a codevector index i to characterize the source vector X, quantized value I. 码本号n指明编码器选择的码本,而索引i表示在该特定码本中选择的码矢量。 Codebook number n specifies the codebook selected by the encoder while the index i denotes the code vector selected in this particular codebook.

[0006] 通常,将适当的无损编码技术分别应用于块En和Ei中的n和i (即,图I中的En和Ei),以便在将它们复合在复用器MUN中以存储或通过通信信道传输之前,減少被编码的码本号nE和索引iE的平均比特率。 [0006] Generally, an appropriate lossless coding technique is applied to blocks En and Ei, respectively, where n and i (i.e., in FIG I Ei and En), so that they recombine to store a multiplexer or by MUN prior to the communication channel transmission, by reducing the average bit rate codebook number nE and index iE encoded.

[0007] 图2示出了变速率量化解码器。 [0007] FIG 2 illustrates a variable rate decoder quantizer. 该解码器的输入端提供了用于将ニ进制码nE和iE分离解复用器DEMUX ;该解码器还包括无损解码模块(即,Dn和Di),在其中解码nE和iE为码本号n和索引i ;该解码器还包括接收码本号n和索引i并进行逆量化的逆量化器(用Q—1表示),其使用码本号和索引来恢复源矢量X的量化值I。 Input of the decoder provides for Ni binary codes nE and iE The DEMUX demultiplexer separated; the lossless decoder further comprises decoding modules (i.e., Dn and Di), wherein the decoding codebook nE and iE number n and index i; the decoder further comprises receiving a codebook number n and index i and the inverse quantizer inverse quantization (using Q-1 shown) which uses a codebook number and the index to recover the quantized value of the source vectors of X I. 不同的n值通常产生不同的比特分配从而产生不同的比特率,每维所需比特数(即,码本比特率)的定义为:分配给源矢量的比特数与源矢量的维数的比值。 Different values ​​of n usually result in different bit allocation to generate a different bit rate, the number of bits required for each dimension (i.e., the codebook bit rate) is defined as: ratio of the number of allocated bits dimension source vector to the source vector .

[0008] 通常,码本的构建可以采用以下多种方法: [0008] Generally, the codebook may be constructed using a variety of the following methods:

[0009] ー种流行的方法是根据源的分布,采用训练算法(如k均值算法)来优化码本项。 [0009] ー species popular method is based on the distribution source using the training algorithm (e.g., k-means algorithm) to optimize the codebook entry. 该方法得到非结构化码本,其对于待量化的每个源矢量通常必须进行存储和穷举搜索。 The method to obtain unstructured codebook, which for each source vector to be quantized must often be stored and exhaustive search. 因此,该方法的缺点是内存需求大,且计算复杂,它随码本比特率的增加而成指数增长。 Thus, the disadvantage of this method is the large memory requirements and computational complexity, which increase exponentially with the codebook bit rate of growth. 如果变速率方法基于上述非结构化的码本,则内存需求大和计算复杂的缺陷会进一步加大,因为通常需要为每个可能的位分配特定的码本。 If variable rate based on the method of unstructured codebook, calculation of the memory requirements Japanese complex defects further increase, because usually need to allocate a specific code to each possible bit present. [0010] 另ー种方法是使用格矢量量化器,其降低了捜索复杂度,并且在许多情况下,可以有效地減少存储需求。 [0010] Another way is to use ー lattice vector quantizer which reduces the complexity Dissatisfied cable, and in many cases, can effectively reduce storage requirements. 格矢量量化是ー种代数型矢量量化器,它的特点是在多维信号空间中,构造ー种有规律的网络,网络中的点称为格点,并以格点进行矢量量化,把信号空间划分为胞腔。 Lattice vector quantization is ー species algebraic vector quantizer, it is characterized in the multidimensional signal space, configured ー planted with regular network, the network is called grid point, and the lattice point vector quantization, the signal space It is divided into cell cavity. 由于网络是有规律的,故格点和胞腔也是有规律的。 Since the network is regular, and the cell chamber so that the lattice points are regular. 格矢量量化器的主要优点是容易构造码书,且可以进行高维量化。 The main advantage of lattice vector quantizer codebook is easy to construct, and can be high dimensional quantization. 图3示出了ニ维空间中的例子,其中基本矢量是Vl和v2,该例子中使用的格是基本的六角形点阵,用A2表示,该图中用十字标识的所有点可如下获得: Figure 3 shows an example of Ni-dimensional space, wherein the basic vector is Vl and V2, the cells used in the examples are substantially hexagonal lattice, represented by A2, all cross points identified by the drawing may be obtained as follows :

[0011] y=klvl+k2v2 (1) [0011] y = klvl + k2v2 (1)

[0012] 其中,y是空间格点,且kl和k2可以是任何整数。 [0012] where, y is a spatial grid points, and kl and k2 can be any integer. 注意到图3只是表不空间格点的ー个子集,因为该空间格点本身可无穷扩展。 Noting FIG. 3 but not ー subsets table space grid because the grid space itself may be extended infinitely.

[0013] 当选择某一空间格点来构造量化码本时,通常选择格点的某一子集来获得具有给定(有限)比特数的码本,使用格点的好处是在确定码本内的所有格点的源矢量X的最近邻点时,存在快速码本搜索算法,并且与其他非结构化的码本相比,可以极大减少复杂性。 [0013] When selecting a spatial lattice quantization codebooks constructed, typically selected subset of lattice points to obtain a codebook with a given (finite) number of bits of this code, the benefits of using grid points is determined when the nearest neighbor of the source vector X possessive point in the presence of fast codebook search algorithms, unstructured and present compared to other codes, it can greatly reduce the complexity. 此夕卜,使用格点无需存储码本,因为码本可以从生成矩阵中获得。 This Bu Xi, grid need not be stored using the present code, since this code may be obtained from the generator matrix.

[0014] 格矢量量化中经常使用的格点是D8格。 [0014] Lattice lattice vector quantization is often used in cell D8. D8是由8维整数格的Z8格点V= (Vl,-,V8) Z8 D8 is an integer of 8-dimensional lattice points of the lattice V = (Vl, -, V8)

组成,且满足 Composition, and satisfy

Figure CN102801427AD00051

=偶数,即: = Even, i.e.:

;=1 ; = 1

[0015] [0015]

Figure CN102801427AD00052

[0016] D8格中任意8维格点y可以通过如下的方法生成: [0016] D8 lattice point y may be any 8-dimensional lattice generated by the following method:

[0017] Y=[k1 k2 …k8]GD8 (3) [0017] Y = [k1 k2 ... k8] GD8 (3)

[0018] 其中k” k2,. . .,k8是有符号的整数,Gd8是生成矩阵,定义为: [0018] where k "k2 ,., k8 are signed integers, GD8 generator matrix is ​​defined as.:

[0019] [0019]

Figure CN102801427AD00053

[0020] 容易验证Gd8是生成矩阵的逆矩阵GS为: [0020] The generator matrix is ​​easy to verify Gd8 inverse matrix GS is:

Figure CN102801427AD00061

[0022] 该逆矩阵在获取D8格点y的坐标时非常有用。 [0022] The inverse matrix is ​​useful in obtaining the coordinate D8 lattice point y.

发明内容 SUMMARY

[0023] 为了解决传统的变速率矢量量化器因随码本个数的増加而导致存储空间增加、且在量化过程中因需要对码本进行全搜索以获得最好的量化矢量而使其捜索的运算复杂度高等缺陷,本发明特给出了以下技术方案。 [0023] In order to solve the conventional variable rate due to enlargement of the vector quantizer codebook with added lead to an increase the number of storage space, and the full search of the codebook in the quantization process by the need to obtain the best vector quantization index and it Dissatisfied the high complexity of computing defects, the present invention gives the following technical solutions Mint.

[0024] 本发明解决其技术问题采用的第一技术方案是,构造ー种源信号变速率格矢量量化的编码方法,包括: [0024] The first aspect to solve the technical problems of the present invention is configured ー provenance signal ramp rate lattice vector quantization encoding method, comprising:

[0025] SI,将输入源信号从时域变换到频域以获得谱系数和控制信息; [0025] SI, the input source signal from the time domain to the frequency domain to obtain spectral coefficients and control information;

[0026] S2,对所述谱系数进行分组和比特分配以获得比特分配信息; [0026] S2, the spectral coefficients are grouped and bit allocation to obtain bit allocation information;

[0027] S3,基于所述比特分配信息,格矢量量化所述谱系数; [0027] S3, the bit allocation information, lattice vector quantization of the spectral coefficients;

[0028] S4,将量化索引、所述比特分配信息、所述控制信息打包成编码比特流. [0028] S4, the quantization index, the bit allocation information, the control information is packed into the encoded bit stream.

[0029] 在本发明所述的源信号变速率格矢量量化的编码方法中,所述步骤S3进ー步包括: [0029] In the variable-rate source signal encoding method for lattice vector quantization of the present invention, the step S3 into ー step comprises:

[0030] S31,对于所述谱系数,计算偏移矢量; [0030] S31, the spectral coefficients to calculate the offset vector;

[0031] S32,对所述偏移矢量进行缩放,得到缩放矢量; [0031] S32, the offset vector is scaled to obtain the scaling vector;

[0032] S33,在D8格空间中搜索与所述缩放矢量最临近的格点; [0032] S33, the search space lattice D8 scaled vectors nearest to the grid point;

[0033] S34,计算所述最临近的格点坐标; [0033] S34, calculates the nearest lattice point coordinates;

[0034] S35,利用所述坐标计算D8格矢量; [0034] S35, the coordinate calculation using D8 lattice vector;

[0035] S36,比较所述D8格矢量与所述最临近的格点是否一致,如果一致,则量化结束,输出所述坐标;如果不一致,则对所述缩放矢量执行逼近量化。 [0035] S36, the comparison is consistent with D8 lattice vector of the nearest grid point, if yes, quantized, the coordinate output; if not, the execution of the scaling vector quantization approximation.

[0036] 在本发明所述的源信号变速率格矢量量化的编码方法中,所述步骤S36中的逼近量化进ー步包括: [0036] In the variable-rate source signal encoding method for lattice vector quantization of the present invention, the approximation step S36 the quantization step ー feed comprising:

[0037] S361,将所述缩放矢量再次缩放,得到再次缩放矢量,运用步骤S33-S35计算得到第二最临近的格点、所述第二最临近的格点坐标,和第二D8格矢量; [0037] S361, re-scaling the scaled vectors to obtain the second nearest lattice point coordinates, D8 lattice vector and the second vector is scaled again, using the steps S33-S35 calculated by the second most adjacent lattice points, ;

[0038] S362,比较所述第二D8格矢量与所述第二最临近的格点是否相等,如果不相等,则重复步骤S361,直至所述第二D8格矢量与所述第二最临近的格点相等。 [0038] S362, comparing the second lattice vector D8 is equal to the second nearest grid point, if not equal, step S361 is repeated, until the second D8 lattice vector and the second nearest equal to the grid.

[0039] 在本发明所述的源信号变速率格矢量量化的编码方法中,所述步骤S36中的逼近量化进ー步包括: [0039] In the variable-rate source signal encoding method for lattice vector quantization of the present invention, the approximation step S36 the quantization step ー feed comprising:

[0040] S363,运用步骤S33-S35计算得到第三D8格矢量、第三最临近的格点和第三最临近的格点坐标; [0040] S363, steps S33-S35 using the calculated third D8 lattice vector, the third nearest grid point and the third nearest grid point coordinates;

[0041] S364,比较所述第三D8格矢量与所述第三最临近的格点,如果两者不相等,则量化结束,输出所述第三最临近的格点坐标及量化比特数;如果两者相等,则重复步骤S363直至两者不相等,最后输出所述第三最临近的格点坐标及量化比特数。 [0041] S364, comparing the third D8 lattice vector and the third nearest grid point, if the two are not equal, then the quantized output of the third nearest grid point coordinates and the number of quantization bits; If both are equal, step S363 is repeated until the two are not equal, the final output of the third nearest grid point coordinates and the number of quantization bits.

[0042] 在本发明所述的源信号变速率格矢量量化的编码方法中,在所述步骤S31中,所述偏移矢量满足: [0042] In the variable-rate source signal encoding method for lattice vector quantization according to the present invention, in the step S31, the vector satisfies the offset:

[0043] [0043]

Figure CN102801427AD00071

[0044] 其中,歹卩表示偏移矢量,yp表示所述谱系数的子矢量,a= (2_6 2_6…2_6)。 [0044] wherein, bad Jie represents the offset vector, yp represents the number of sub-vectors of the spectrum, a = (2_6 2_6 ... 2_6).

[0045] 在本发明所述的源信号变速率格矢量量化的编码方法中,在所述步骤S32中,所述缩放矢量满足: [0046] [0045] In the variable-rate source signal encoding method for lattice vector quantization according to the present invention, in the step S32, the scaling vector satisfies: [0046]

Figure CN102801427AD00072

[0047] 其中,表示所述缩放矢量,P (p)=2E(p)/6表示缩放因子,R (P)表示每个所述谱系数的子矢量分配的量化比特数。 [0047] where represents the scaling vector, P (p) = 2E (p) / 6 represents a scaling factor, R (P) represents a quantization bit number of the sub-vector for each spectrum allocation.

[0048] 在本发明所述的源信号变速率格矢量量化的编码方法中,R(P)满足: [0048] In the variable-rate source signal encoding method for lattice vector quantization according to the present invention, R (P) is satisfied:

[0049] [0049]

Figure CN102801427AD00073

[0050] 其中,所述谱系数的个数为N,将所述N个谱系数分成L个8維子矢量,V表示一帧源信号总的量化编码比特数,Q表示帧源信号经过比特分配算法后剰余比特数为Q。 [0050] wherein the number of spectral coefficients is N, the N L spectral coefficients into 8-dimensional sub-vectors, V represents the total number of quantization bits encoding a source signal, Q represents a source signal after the frame bit allocation after several algorithms for Surplus remainder bits Q.

[0051] 本发明解决其技术问题采用的第二技术方案是,构造ー种源信号变速率格矢量量化的编码系统,包括: [0051] The second aspect to solve the technical problems of the present invention is configured ー provenance signal ramp rate lattice vector quantization coding system, comprising:

[0052] 正交变换模块,用于将输入源信号从时域变换到频域以获得谱系数和控制信息; [0052] The orthogonal transformation module for converting an input source signals from the time domain to the frequency domain to obtain spectral coefficients and control information;

[0053] 谱系数分组和比特分配模块、用于对所述谱系数进行分组和比特分配以获得比特分配信息; [0053] The spectral coefficients packets and bit allocation means for grouping the spectral coefficients and the bit allocation to obtain bit allocation information;

[0054] 格矢量量化模块,用于基于所述比特分配信息,格矢量量化所述谱系数; [0054] The lattice vector quantization module, based on the bit allocation information, lattice vector quantization of the spectral coefficients;

[0055] 编码比特流模块,用于将量化索引、所述比特分配信息、所述控制信息打包成编码比特流。 [0055] The encoded bit stream module, the quantization index for the bit allocation information, the control information is packed into the encoded bit stream.

[0056] 本发明解决其技术问题采用的第三技术方案是,构造ー种源信号变速率格矢量量化的解码方法,包括: [0056] The third aspect of solving the technical problems of the present invention is configured ー provenance signal ramp rate lattice vector quantization decoding method, comprising:

[0057] SI,接收编码比特流进行解码以获得解码比特流; [0057] SI, reception encoded bit stream to obtain decoded bit stream;

[0058] S2,对所述解码比特流进行比特分配和量化索引解码; [0058] S2, the decoded bitstream bit allocation and quantization index decoding;

[0059] S3,基于解码的量化索引进行逆格矢量量化得到重建量化矢量; [0059] S3, based on the decoded quantization index lattice vector quantization to obtain inverse quantized reconstruction vectors;

[0060] S4,基于所述控制信息对所述重建量化矢量进行逆正交变换得到重建信号。 [0060] S4, to obtain a reconstructed signal based on the control information of the reconstructed vector quantization inverse orthogonal transform.

[0061] 本发明解决其技术问题采用的第四技术方案是,构造ー种源信号变速率格矢量量化的解码模块,包括: [0061] The fourth aspect of solving the technical problems of the present invention is configured ー provenance signal ramp rate lattice vector quantization decoding module, comprising:

[0062] 编码比特流解码模块,用于接收编码比特流进行解码以获得解码比特流; [0062] The encoded bit stream decoding module for receiving the encoded bit stream to obtain decoded bit stream;

[0063] 比特分配和量化索引解码模块,用于对所述解码比特流进行比特分配和量化索引解码;[0064] 逆格矢量量化模块,用于基于解码的量化索引进行逆格矢量量化得到重建量化矢量; [0063] The bit allocation and quantization index decoding means for decoding the bit stream for decoding the bit allocation and quantization index; [0064] inverse lattice vector quantization means for inverse lattice vector quantization based on the quantization indices decoded rebuilt vector quantization;

[0065] 逆正交变换模块,用于基于所述控制信息对所述重建量化矢量进行逆正交变换得 [0065] The inverse orthogonal transformation module based on the control information to obtain the reconstructed vector quantization inverse orthogonal transform

到重建信号。 To reconstruct the signal.

[0066] 相比于传统的变速率矢量量化器存储多个矢量码本,本发明方法无需存储矢量码本;此外,存在快速算法,其运算复杂度较传统矢量量化大幅度降低,即具有低运算复杂度的优点;第三,还具有可以实现变比速率量化的优点。 [0066] Compared to the conventional variable rate vector quantizer stores a plurality of vector codebook, the method of the present invention does not require vector codebook memory; moreover, the presence of fast algorithm, computational complexity than the conventional vector quantization greatly reduced, i.e., having low computation complexity advantage; third, there is an advantage can be realized than the rate of change of the quantization.

[0067] 本领域技术人员应该意识到,前述概括仅仅是为了提供本发明的特定方面的简单描述。 [0067] Those skilled in the art should appreciate that the foregoing general description for simplicity only certain aspects of the present invention is provided. 通过结合附图并參照权利要求和以下优选实施例的详细描述,能够获得对本发明的更完全的理解。 Conjunction with the drawings and the following detailed description and claims with reference to preferred embodiments, it is possible to obtain a more complete understanding of the present invention.

附图说明 BRIEF DESCRIPTION

[0068] 下面将结合附图及实施例对本发明作进ー步说明,附图中: [0068] The accompanying drawings and the following embodiments of the present invention will be further described embodiment into ー drawings in which:

[0069] 图I是传统的变速率量化编码器的量化框图; [0069] Figure I is a block diagram of a conventional variable rate quantizer quantizer encoder;

[0070] 图2是传统的变速率量化解码器的量化框图; [0070] FIG. 2 is a block diagram illustrating a quantization decoder quantizer conventional variable rate;

[0071] 图3是表不某空间格点的ー个子集; [0071] FIG. 3 is a table space is not a subset of lattice points ー;

[0072] 图4A示出了根据本发明一个实施例的编码器的框图; [0072] FIG 4A shows a block diagram of an embodiment of the encoder according to the embodiment of the present invention;

[0073] 图4B示出了根据本发明一个实施例的解码器的框图; [0073] FIG. 4B illustrates a block diagram of one embodiment of a decoder according to the embodiment of the present invention;

[0074] 图5示出了根据本发明一个实施例的量化方法的流程图;以及 [0074] FIG. 5 shows a flowchart of a quantization method according to an embodiment of the present invention; and

[0075] 图6示出了根据本发明一个实施例的逆量化方法的流程图。 [0075] FIG. 6 shows a flowchart of an inverse quantization method according to an embodiment of the present invention, an embodiment.

具体实施方式 detailed description

[0076] 本发明的主要目的是提供一种用于基于空间D8格点的源信号变速率量化技术(例如,包括量化方法和量化系统等),该量化技术与现有技术相比,能够实现变速率量化,且无需存储空间和运算复杂度低。 [0076] The main object of the present invention is to provide a method for variable rate signal based on a source spatial D8 lattice quantization techniques (e.g., quantization systems including quantization method and the like), the quantization techniques of the prior art can be achieved compared variable rate quantizer, and no storage space and low computational complexity. 该量化技术能够应用于各种变速率编码系统和可分级编码系统。 The quantization techniques can be applied to various variable-rate coding system and scalable encoding system. 为了叙述简便却不致混淆,本发明实施例可能省略了对本领域技术人员所公知的内容,例如比特分配算法、谱系数分组算法、格矢量量化的索引编解码算法等。 For simple description, but not to obscure embodiments of the present invention may omit the skilled person known content, such as bit allocation algorithm, spectral coefficients grouping algorithm, lattice vector quantization index encoding and decoding algorithms.

[0077] 在根据本发明的一个优选实施例中,源信号变速率格矢量量化的编码器主要包括:正交变换模块101、谱系数分组与比特分配模块102和格矢量量化模块103和编码比特流模块104。 [0077] In accordance with a preferred embodiment of the present invention, the signal source variable rate lattice vector quantization encoder including: an orthogonal transformation module 101, a packet with bit allocation spectral coefficients and lattice vector quantization module 102 and coding module 103 bits flow module 104. 而源信号变速率格矢量量化的解码器为源信号变速率格矢量量化的编码器的逆系统,主要包括编码比特流解码模块204、比特分配和量化索引解码模块201、逆格矢量量化模块202和正交变换模块203。 While the source signal ramp rate lattice vector quantization decoder is a source signal ramp rate lattice vector quantization of the inverse system encoder, including coded bit stream decoding module 204, a bit allocation and quantization index decoding module 201, an inverse lattice vector quantization module 202 and orthogonal transform module 203. 图4示意性示出了整个编解码器的框图。 FIG 4 schematically shows a block diagram of the whole codec.

[0078] 具体而言,在图4A中,根据本发明的一个实施例,在编码端,在模块101中,利用诸如离散余弦变换(DCT)、改进的离散余弦变换(MDCT)等的正交变换之一将输入的原始信号从时域变换到频域,得到谱系数和控制信息;接下来,在模块102中,对谱系数进行分组以及比特分配,获得比特分配信息;然后,在模块103中,基于比特分配信息,对谱系数进行格矢量量化;最后,将控制信息、比特分配信息以及量化索引在模块104打包成编码比特流,输入信道或存储。 Orthogonal [0078] Specifically, in FIG. 4A, in accordance with one embodiment of the present invention, the encoding side, in module 101, such as by using a discrete cosine transform (DCT), modified discrete cosine transform (MDCT), etc. one is to transform the original input signal from the time domain to the frequency domain, and control information to obtain spectral coefficients; Next, at block 102, the grouping of spectral coefficients and a bit allocation to obtain bit allocation information; then, at block 103 based on the bit allocation information for lattice vector quantization of spectral coefficients; Finally, the control information, bit allocation information and the quantization index packed into the encoded bit stream, the input channel 104 or stored in the module.

[0079] 图4B中,根据本发明的一个实施例,在解码端,在模块204中,将接收到的编码比特流进行解码,并结合模块201进行比特分配和量化索引解码;在模块202中,根据模块201中解码的量化索引进行逆格矢量量化得到重建的量化矢量;最后,在模块203中,重建的量化矢量在控制信息的控制下进行逆正交变换,得到重建信号。 [0079] FIG. 4B, according to one embodiment of the invention, the decoding side, in the module 204, the received coded bit stream is decoded, and binding module 201 bit allocation and quantization index decoding; in module 202 , inverse lattice vector quantization index according to the decoded quantization module 201 to obtain reconstructed quantized vector; Finally, at block 203, the reconstructed quantized inverse orthogonal transformation vector under the control of control information to obtain the reconstructed signal.

[0080] 在图5中,详细示出了本发明的ー个更具体的实现过程。 [0080] In FIG. 5, shown in detail ー a more specific implementation of the present invention. 假定ー帧信号经过诸如离散傅立叶变换、离散余弦变换等线性正交变换之一处理后得到的谱系数个数为N,将上述N个谱系数分成L个8維子矢量(S卩,满足8 XL = N),假定每ー个8維子矢量分配的量化比特数为R(P)比持/維(P=0,1,...,L-1),ー帧信号总的量化编码比特数为¥,经过比特分配算法后剰余比特数为Q,编码端量化步骤如下: After the frame signal is assumed ー such as discrete number of spectral coefficients after the Fourier transform, one linear orthogonal transform, discrete cosine transform processing such as obtained is N, the said N L spectral coefficients into 8-dimensional sub-vectors (S Jie, 8 satisfy XL = N), assuming a number of quantization bits ー per 8-dimensional sub-vector assignment is R (P) than the support / dimension (P = 0,1, ..., L-1), the frame number signal ー coded bit quantization of total ¥ is, for Surplus remainder bits through bit number Q after allocation algorithm, the coding of the quantizer steps:

[0081] 在步骤301中,根据总编码码率和所选择的量化比特分配算法确定每ー个8維子矢量分配的量化比特数为R (P)比特/维(P=0,1,…,Ll),R(p)应满足如下的限制: [0081] In step 301, the total allocation algorithm according to the selected coding rate and determines the quantization bit number of quantization bits ー per 8-dimensional sub-vector assignment is R (P) bits / dimension (P = 0,1, ..., Ll), R (p) should meet the following constraints:

[0082] [0082]

Figure CN102801427AD00091

[0083] 接下来,在步骤302中,对某一任意的8维矢量ypKypa yp,2*“yP,8), [0083] Next, in step 302, an 8-dimensional vector of any one ypKypa yp, 2 * "yP, 8),

p=0, • • • LI,将其减去某一偏移矢量a= (2_6 2_6〜2_6),得到偏移后的矢 p = 0, • • • LI, which subtracts an offset vector a = (2_6 2_6~2_6), obtained after the offset vector

量t : The amount of t:

[0084] [0084]

Figure CN102801427AD00092

[0085] 在步骤303中,对上述步骤得出的偏移矢量歹卩进行缩放得到缩放矢量:,其中缩放因子(P)=2K(p)/6,则ろ= Yp/ P{P) [0085] In step 303, the above-described disparity vector obtained in step to give bad Jie zoom scaling vector: wherein the scaling factor (P) = 2K (p) / 6, the ro = Yp / P {P)

[0086] 在步骤304中,在D8格空间中搜索与缩放矢量! [0086] In step 304, the search space lattice D8 scaled vector! ^最临近格点V,其满足: ^ Nearest grid point V, which satisfies:

[0087] [0087]

Figure CN102801427AD00093

[0088] 在步骤305中,计算D8格格点V在R(p)比特/维的Voronoi扩展截断坐标k=(kiV..k8),其中0 彡Ici 彡2e(p)-1, i=l,2,…,8,k 的计算为: [0088] In step 305, the point V is calculated formatting D8 R (p) bits / dimension coordinate truncated Voronoi extension k = (kiV..k8), wherein 0 Ici San San 2e (p) -1, i = l , 2, ..., 8, k is calculated as:

[0089] [0089]

Figure CN102801427AD00094

[0090] 其中G=是D8格的逆生成矩阵,见式(5)。 [0090] where G = D8 lattice is inverse generator matrix, see formula (5).

[0091] 在步骤306中,根据给定坐标Ic=Gc1 V" k8)计算x=kGD8和Z=I^1X (其中Gd8是D8格的生成矩阵,见式4),并在D8格空间中搜索与缩放矢量z最临近格点\,然后计算D8格 [0091] In step 306, based on a given coordinate Ic = Gc1 V "k8) computing x = kGD8 and Z = I ^ 1X (where Gd8 is D8 lattice generator matrix, see Formula 4), and D8 lattice space Search and scalable vector z nearest grid point \, and then calculate the lattice D8

矢量c : Vector c:

[0092] c=x~r A (10) [0092] c = x ~ r A (10)

[0093] 在步骤307中,比较格矢量c和V,如果c和V 一致,则哗标k= Qc1 k2... k8)就是缩放矢量;^的最佳坐标,量化结束。 [0093] In step 307, the comparison and lattice vector V c, V c, and if the same, then the standard Wow k = Qc1 k2 ... k8) is scaled vectors; best coordinates ^ and quantized. 如果c和V不一致,那么矢量:^为局外点,此时需要逐次逼近的来量化。 If c and inconsistent V, then the vector: ^ as outliers, this time need to quantify the successive approximation.

[0094] 接下来,运用上述步骤303〜306的方法,逐次逼近。 [0094] Next, using the above-described method steps 303~306, successive approximation. 在步骤308中,将矢量:缩 In step 308, the vector: Condensation

放2,即t ニ;^/2在步骤309、310中,在D8格空间中搜索与缩放矢量:^最临近格点U,计算u的坐标j,即: 2 release, i.e., Ni t; ^ / 2 in step 309, 310, the search space lattice D8 scaled vector: ^ nearest lattice point U, calculates the coordinates u j, namely:

Figure CN102801427AD00101

[0096] 在步骤311、312中,利用坐标j计算得到D8格矢量c',比较格矢量ど和U,如果C1和u不相等,贝U重复步骤308到步骤312,直至c'和u相等。 [0096] In step 311, 312, is calculated using the coordinate j D8 lattice vector c ', and U-do compare lattice vector, and if u is not equal to C1, U shell step 308 to step 312 is repeated until the c' and u are equal .

[0097] 在步骤313中,计算缩放矢量W = ^/2OT ,其中m=3、4、5或6 ;在步骤314中,计算+W。 [0097] In step 313, it calculates a scaling vector W = ^ / 2OT, where m = 3,4,5 or 6; in step 314, calculates + W.

[0098] 在步骤315中,在D8格空间中搜索与缩放矢量\最临近格点u',计算u'的坐标j',即: [0098] In step 315, the search space lattice D8 scaled vector \ nearest lattice point u ', calculates u' coordinate j ', namely:

[0099] Jf = (UfG^18)Sr r = 2R(P) (⑵ [0099] Jf = (UfG ^ 18) Sr r = 2R (P) (⑵

[0100] 在步骤316中,利用坐标j'计算得到D8格矢量c",比较格矢量c"和u',如果c"和u'相等则k=j',并且重复步骤314到步骤316。如果c"和u'不相等,则停止循环。 [0100] In step 316, using the coordinate j 'calculated D8 lattice vector c ", compare lattice vector c" and u', if c "and u 'equal to the k = j', and repeats step 314 to step 316. If c "and u 'are not equal, then stop the loop.

[0101] 在步骤318中,将每ー个8維子矢量分配的量化比特数为R(P)比持/維(p=0, I. . . . , LI)和D8格坐标kp编码后传到解码端。 [0101] In step 318, each ー number of quantization bits of 8-dimensional sub-vector assignment is R (P) than the support / dimension (p = 0, I...., LI) and D8 Cor coordinate kp coding pass to the decoding side.

[0102] 相对上述编码端量化的方法,图6示出了解码端逆量化的流程图,具体实施步骤如下: [0102] The method of encoding the opposite ends of the quantization, FIG. 6 shows a flowchart of an inverse quantization decoder, specific implementation steps are as follows:

[0103] 在步骤401、402中,从编码的码流中解码得到每ー个8維子矢量分配的量化比特数为R(P)比特/维(p=0,1,. . .,L-1)和D8格坐标kp。 [0103] In step 401, from the encoded bit stream obtained by decoding each of the number of quantization bits ー 8-dimensional sub-vector assignment is R (P) bits / dimension (p = 0,1 ,..., L- 1) and D8 grid coordinates kp.

[0104] 接下来,在步骤403中,对给定坐标kp=(kp,I kp,2...kp,8)计算x=kGD8,并在D8格空间中搜索与缩放矢量z最临近格点\,然后计算D8格矢量夕〃: [0104] Next, in step 403, a given coordinate kp = (kp, I kp, 2 ... kp, 8) calculate x = kGD8, and searches for the nearest lattice with lattice D8 scaled vector z in space point \, then calculates D8 lattice vector Xi 〃:

[0105] yp=xm (13) [0105] yp = xm (13)

[0106] 然后,在步骤404中,对矢量:^^进行逆缩放得到,缩放因子为P (p)=2e(p)/6 : [0106] Then, in step 404, vector: ^^ obtained inverse scaling, scaling factor P (p) = 2e (p) / 6:

[0107] Jp =ypP{P) (14) [0107] Jp = ypP {P) (14)

[0108] 最后,在步骤405中,将矢量歹卩加上偏移矢量a= (2_6 2'"2_6),得到重建矢量fp : [0108] Finally, in step 405, the bad Jie vector plus the offset vector a = (2_6 2 ' "2_6), to obtain reconstructed vector fp:

Figure CN102801427AD00102

[0110] 需要说明的是,本发明不局限于对频谱系数进行量化,还适用于语音编码中LPC系数的量化。 [0110] Incidentally, the present invention is not limited to quantized spectral coefficients, it is also applicable to speech coding quantized LPC coefficients. 此外,本发明能够应用于各种变速率编码系统和可分级编码系统,具有广泛的适用性。 Further, the present invention can be applied to various variable-rate coding system and scalable coding system has broad applicability.

Claims (10)

  1. 1. ー种源信号变速率格矢量量化的编码方法,其特征在于,包括: Si,将输入源信号从时域变换到频域以获得谱系数和控制信息; S2,对所述谱系数进行分组和比特分配以获得比特分配信息; S3,基于所述比特分配信息,格矢量量化所述谱系数; S4,将量化索引、所述比特分配信息、所述控制信息打包成编码比特流。 1. quantization encoding method ー provenance variable rate lattice vector signal, characterized by comprising: Si, the input source signal from the time domain to the frequency domain to obtain spectral coefficients and control information; S2, the number of spectrum for packet and bit allocation to obtain bit allocation information; S3, based on the bit allocation information, lattice vector quantization of the spectral coefficients; S4, the quantization index, the bit allocation information, the control information is packed into the encoded bit stream.
  2. 2.根据权利要求I所述的源信号变速率格矢量量化的编码方法,其特征在于,所述步骤S3进ー步包括: S31,对于所述谱系数,计算偏移矢量; S32,对所述偏移矢量进行缩放,得到缩放矢量; S33,在D8格空间中搜索与所述缩放矢量最临近的格点; S34,计算所述最临近的格点坐标; S35,利用所述坐标计算D8格矢量; S36,比较所述D8格矢量与所述最临近的格点是否一致,如果一致,则量化结束,输出所述坐标;如果不一致,则对所述缩放矢量执行逼近量化。 2. I claim the variable rate source signal encoding method for lattice vector quantization, wherein said step includes the step S3 into ー: S31, for the spectral coefficients, calculating an offset vector; S32, on the scaling said offset vector to obtain a scalable vector; S33, in the search space lattice D8 scaled to the lattice points nearest the vector; S34, calculating the nearest lattice point coordinates; S35, by using the coordinate calculation D8 lattice vector; S36, comparing the D8 lattice vector is consistent with the nearest grid point, if yes, quantized, the coordinate output; if not, the execution of the scaling vector quantization approximation.
  3. 3.根据权利要求2所述的源信号变速率格矢量量化的编码方法,其特征在于,所述步骤S36中的逼近量化进ー步包括: S361,将所述缩放矢量再次缩放,得到再次缩放矢量,运用步骤S33-S35计算得到第二最临近的格点、所述第二最临近的格点坐标,和第二D8格矢量; S362,比较所述第二D8格矢量与所述第二最临近的格点是否相等,如果不相等,则重复步骤S361,直至所述第二D8格矢量与所述第二最临近的格点相等。 The signal source 2 becomes the rate lattice vector quantization encoding method as claimed in claim, wherein said approximating step S36 into ー quantization step comprises: S361, the re-scaling the scaling vector, the scaling again to give vector, using steps S33-S35 calculated by the second nearest grid point, the second nearest lattice point coordinates, D8 and the second lattice vector; S362, comparing the second and the second lattice vector D8 the nearest lattice point is equal, if not equal, step S361 is repeated, until the second D8 lattice vector and the second is equal to the nearest grid point.
  4. 4.根据权利要求3所述的源信号变速率格矢量量化的编码方法,其特征在于,所述步骤S36中的逼近量化进ー步包括: S363,运用步骤S33-S35计算得到第三D8格矢量、第三最临近的格点和第三最临近的格点坐标; S364,比较所述第三D8格矢量与所述第三最临近的格点,如果两者不相等,则量化结束,输出所述第三最临近的格点坐标及量化比特数;如果两者相等,则重复步骤S363直至两者不相等,最后输出所述第三最临近的格点坐标及量化比特数。 4. Source signal of the variable-rate lattice vector quantization of 3 encoding method according to claim, wherein said approximating step S36 into ー quantization step comprises: S363, is calculated using steps S33-S35 of the third cell D8 vector, the third nearest grid point and the third nearest grid point coordinates; S364, comparing the third D8 lattice vector and the third nearest grid point, if the two are not equal, the quantization end, coordinates and the number of quantization bits of the output of the third grid point nearest; and if both are equal, both the step S363 is repeated until it is not equal, the final output of the third nearest grid point coordinates and the number of quantization bits.
  5. 5.根据权利要求1-4中任ー权利要求所述的源信号变速率格矢量量化的编码方法,其特征在于,在所述步骤S31中,所述偏移矢量满足: 其中,表示偏移矢量,Yp表示所述谱系数的子矢量,a= (2_6 2_6…2_6)。 According to any of claims 1-4 according to claim ー source signal ramp rate lattice vector quantization encoding method as claimed in claim, wherein, in the step S31, the offset vector is satisfied: where represents the offset vector, Yp represents the number of sub-vectors of the spectrum, a = (2_6 2_6 ... 2_6).
  6. 6.根据权利要求5所述的源信号变速率格矢量量化的编码方法,其特征在于,在所述步骤S32中,所述缩放矢量满足: 其中,;^^表示所述缩放矢量,0 (P)=2K(p)/6表示缩放因子,R (P)表示每个所述谱系数的子矢量分配的量化比特数。 The coding method according to the source signal ramp rate lattice vector quantization of the preceding claims, characterized in that, in the step S32, the scaling vector is satisfied: wherein; ^^ represents the scaling vector 0 ( P) = 2K (p) / 6 represents a scaling factor, R (P) represents a quantization bit number of the sub-vector for each spectrum allocation.
  7. 7.根据权利要求6所述的源信号变速率格矢量量化的编码方法,其特征在干,R(p)满足: 7. The method of claim 6 encoding signal source variable rate lattice vector quantization of the preceding claims, characterized in that dry, R (p) is satisfied:
    Figure CN102801427AC00031
    其中,所述谱系数的个数为N,将所述N个谱系数分成L个8維子矢量,V表示ー帧源信号总的量化编码比特数,^表示帧源信号经过比特分配算法后剰余比特数为Q。 Wherein the number of spectral coefficients is N, the N L spectral coefficients into 8-dimensional sub-vectors, V represents the total number of coded bits quantization ー source signal frame, a rear frame ^ source signal through the bit allocation algorithm for Surplus I the number of bits is Q.
  8. 8. ー种源信号变速率格矢量量化的编码系统,其特征在于,包括: 正交变换模块,用于将输入源信号从时域变换到频域以获得谱系数和控制信息; 谱系数分组和比特分配模块、用于对所述谱系数进行分组和比特分配以获得比特分配信息; 格矢量量化模块,用于基于所述比特分配信息,格矢量量化所述谱系数; 编码比特流模块,用于将量化索引、所述比特分配信息、所述控制信息打包成编码比特流。 8. ー provenance variable rate lattice vector quantization signal encoding system, characterized by comprising: orthogonal transform block, for transforming the input source signal from the time domain to the frequency domain to obtain spectral coefficients and control information; grouping the spectral coefficients and a bit allocation module for grouping the spectral coefficients and the bit allocation to obtain bit allocation information; lattice vector quantization module, based on the bit allocation information, lattice vector quantization of the spectral coefficients; encoded bit stream module, quantization index for the bit allocation information, the control information is packed into the encoded bit stream.
  9. 9. ー种源信号变速率格矢量量化的解码方法,其特征在于,包括: SI,接收编码比特流进行解码以获得解码比特流; S2,对所述解码比特流进行比特分配和量化索引解码; S3,基于解码的量化索引进行逆格矢量量化得到重建量化矢量; S4,基于所述控制信息对所述重建量化矢量进行逆正交变换得到重建信号。 9. The variable rate signal ー provenance lattice vector quantization decoding method, comprising: SI, reception encoded bit stream to obtain decoded bit stream; S2, the decoded bitstream bit allocation and quantization indices decoded ; S3, inverse lattice vector quantization based on the quantization indices to obtain decoded vector quantized reconstruction; S4, based on the control information of the reconstructed vector quantization inverse orthogonal transform to obtain a reconstructed signal.
  10. 10. ー种源信号变速率格矢量量化的解码模块,其特征在于,包括: 编码比特流解码模块,用于接收编码比特流进行解码以获得解码比特流; 比特分配和量化索引解码模块,用于对所述解码比特流进行比特分配和量化索引解码; 逆格矢量量化模块,用于基于解码的量化索引进行逆格矢量量化得到重建量化矢量;逆正交变换模块,用于基于所述控制信息对所述重建量化矢量进行逆正交变换得到重建信号。 10. The variable rate signal ー provenance lattice vector quantization decoding module, characterized by comprising: an encoded bit stream decoding module for receiving an encoded bit stream to obtain decoded bit stream; bit allocation and the quantization index decoding module, with to the decoded bitstream bit allocation and quantization index decoding; inverse lattice vector quantization means for inverse lattice vector quantization based on the quantization indices to obtain decoded vector quantized reconstruction; inverse orthogonal transformation module, based on the control information on the reconstructed vector quantization inverse orthogonal transform to obtain a reconstructed signal.
CN 201210279991 2012-08-08 2012-08-08 Encoding and decoding methods and systems of lattice vector quantization of variable rate of source signal CN102801427B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201210279991 CN102801427B (en) 2012-08-08 2012-08-08 Encoding and decoding methods and systems of lattice vector quantization of variable rate of source signal

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 201210279991 CN102801427B (en) 2012-08-08 2012-08-08 Encoding and decoding methods and systems of lattice vector quantization of variable rate of source signal

Publications (2)

Publication Number Publication Date
CN102801427A true true CN102801427A (en) 2012-11-28
CN102801427B CN102801427B (en) 2015-05-06

Family

ID=47200412

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201210279991 CN102801427B (en) 2012-08-08 2012-08-08 Encoding and decoding methods and systems of lattice vector quantization of variable rate of source signal

Country Status (1)

Country Link
CN (1) CN102801427B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070168197A1 (en) * 2006-01-18 2007-07-19 Nokia Corporation Audio coding
CN101165777A (en) * 2006-10-18 2008-04-23 宝利通公司 Fast lattice vector quantization
CN101266795A (en) * 2007-03-12 2008-09-17 华为技术有限公司;上海交通大学 An implementation method and device for grid vector quantification coding
CN102081926A (en) * 2009-11-27 2011-06-01 中兴通讯股份有限公司 Method and system for encoding and decoding lattice vector quantization audio

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070168197A1 (en) * 2006-01-18 2007-07-19 Nokia Corporation Audio coding
CN101165777A (en) * 2006-10-18 2008-04-23 宝利通公司 Fast lattice vector quantization
CN101266795A (en) * 2007-03-12 2008-09-17 华为技术有限公司;上海交通大学 An implementation method and device for grid vector quantification coding
CN102081926A (en) * 2009-11-27 2011-06-01 中兴通讯股份有限公司 Method and system for encoding and decoding lattice vector quantization audio

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
张勇等: "基于高斯格型矢量量化的导谱频率参数量化方法", 《数据采集与处理》, vol. 24, no. 5, 30 September 2009 (2009-09-30) *

Also Published As

Publication number Publication date Type
CN102801427B (en) 2015-05-06 grant

Similar Documents

Publication Publication Date Title
US20080181296A1 (en) Per multi-block partition breakpoint determining for hybrid variable length coding
US20010017941A1 (en) Method and apparatus for table-based compression with embedded coding
Laroia et al. Robust and efficient quantization of speech LSP parameters using structured vector quantizers
US6349152B1 (en) Table-based compression with embedded coding
US6345126B1 (en) Method for transmitting data using an embedded bit stream produced in a hierarchical table-lookup vector quantizer
US7433824B2 (en) Entropy coding by adapting coding between level and run-length/level modes
US6128346A (en) Method and apparatus for quantizing a signal in a digital system
US7822601B2 (en) Adaptive vector Huffman coding and decoding based on a sum of values of audio data symbols
US7680670B2 (en) Dimensional vector and variable resolution quantization
US20050285764A1 (en) Method and system for multi-rate lattice vector quantization of a signal
US20070168197A1 (en) Audio coding
US20080097749A1 (en) Dual-transform coding of audio signals
US20080097755A1 (en) Fast lattice vector quantization
US20090175550A1 (en) Successively Refinable Lattice Vector Quantization
US20090228284A1 (en) Method and apparatus for encoding/decoding multi-channel audio signal by using a plurality of variable length code tables
Tai et al. Two fast nearest neighbor searching algorithms for image vector quantization
Man et al. A family of efficient and channel error resilient wavelet/subband image coders
JP2004258603A (en) Entropy encoding adapting encoding between level mode and run length/level mode
Wang et al. Context-based adaptive arithmetic coding in time and frequency domain for the lossless compression of audio coding parameters at variable rate
US8731947B2 (en) Coding method, decoding method, codec method, codec system and relevant apparatuses
CN101223570A (en) Frequency segmentation to obtain bands for efficient coding of digital media
Kovács A fast classification based method for fractal image encoding
Rawat et al. A Hybrid Image Compression Scheme Using DCT and Fractal Image Compression.
RU2419246C1 (en) Method to compress and recover fixed halftone video images
Kossentini et al. Image coding using entropy-constrained residual vector quantization

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
C14 Grant of patent or utility model