WO2011085630A1 - 混合维度编解码方法和装置 - Google Patents

混合维度编解码方法和装置 Download PDF

Info

Publication number
WO2011085630A1
WO2011085630A1 PCT/CN2010/080410 CN2010080410W WO2011085630A1 WO 2011085630 A1 WO2011085630 A1 WO 2011085630A1 CN 2010080410 W CN2010080410 W CN 2010080410W WO 2011085630 A1 WO2011085630 A1 WO 2011085630A1
Authority
WO
WIPO (PCT)
Prior art keywords
processed
dimension
variable
variables
spectral
Prior art date
Application number
PCT/CN2010/080410
Other languages
English (en)
French (fr)
Inventor
蒋三新
刘佩林
应忍冬
肖玮
Original Assignee
华为技术有限公司
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
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Publication of WO2011085630A1 publication Critical patent/WO2011085630A1/zh
Priority to US13/457,238 priority Critical patent/US20120215525A1/en

Links

Classifications

    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L19/00Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
    • G10L19/02Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders
    • G10L19/032Quantisation or dequantisation of spectral components
    • G10L19/035Scalar quantisation
    • GPHYSICS
    • G10MUSICAL INSTRUMENTS; ACOUSTICS
    • G10LSPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L19/00Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
    • G10L19/002Dynamic bit allocation

Definitions

  • the present invention relates to the field of codec technologies, and in particular, to a hybrid dimension codec method and apparatus. Background technique
  • Context-based arithmetic coding is a high-performance lossless compression processing method that uses the short-term stability of the signal to establish a statistical model describing the correlation of adjacent frames or adjacent frames, and selects the best through analysis and real-time. The model thus achieves a highly efficient lossless coding effect.
  • Context refers to the coefficient of the codec that has been obtained before the current code to be coded based on the fact that there is some connection between the corresponding frequency points of adjacent frames or the adjacent frequency points of the same frame.
  • Unified Speech and Audio Coding is the ongoing standard for the Moving Picture Experts Group (MPEG). Among them, context-based arithmetic coding is adopted, and the spectral coefficient of each frame is formed from a low frequency to a high frequency every 4 groups to form a 4-dimensional vector space 4-tuples as a coding object.
  • MPEG Moving Picture Experts Group
  • Embodiments of the present invention provide a hybrid dimension codec method and apparatus, which adopt different dimensions for different spectrum coefficients, and improve codec efficiency through multi-dimensional hybrid codec.
  • the embodiment of the invention provides a hybrid dimension codec method, including:
  • Calculating at least one variable set based on the processed spectral coefficients The processing dimension of the spectral coefficient to be processed is determined according to the relationship between the at least one variable set and the corresponding set of threshold values.
  • the to-be-processed spectral coefficients are encoded or decoded in the dimension according to the selected dimension.
  • the embodiment of the invention provides a hybrid dimension codec device, including:
  • variable acquisition module configured to calculate at least one variable set according to the processed spectral coefficient
  • dimension determining module configured to determine a spectrum to be processed according to a relationship between the at least one variable set obtained by the variable acquiring module and the corresponding threshold set The processing dimension of the coefficient.
  • Codec module for encoding or decoding the to-be-processed spectral coefficients in the dimension according to the selected dimension.
  • At least one variable set is calculated according to the processed spectral coefficient, and the processing dimension of the to-be-processed spectral coefficient is determined according to the relationship between the at least one variable set and the corresponding threshold set, according to the selected dimension.
  • the technical means for encoding or decoding the spectral coefficients to be processed in the dimension by using different processing dimensions for different spectral coefficients improves the coding and decoding efficiency.
  • FIG. 1 is a schematic diagram of a context model according to an embodiment of the present invention
  • FIG. 2 is a flowchart of an embodiment of a hybrid dimension codec method according to the present invention.
  • FIG. 3 is a schematic diagram of a 16-th order context model according to an embodiment of the present invention.
  • FIG. 4 is a flow chart of an embodiment of a method for determining a dimension in a hybrid dimension codec method according to the present invention
  • FIG. 5 is a flow diagram of still another embodiment of a dimension determining method in a hybrid dimension codec method according to the present invention. Cheng Tu
  • FIG. 6 is a flow chart of an embodiment of a hybrid dimension codec method according to the present invention in combination with a location determining dimension;
  • FIG. 7 is a flowchart of an embodiment of a hybrid dimension codec method according to the present invention.
  • Figure 8 is a context model in the embodiment shown in Figure 7;
  • FIG. 9 is a flow chart of a dimension determining method in the embodiment shown in FIG. 7;
  • Figure 10 is another context model in the embodiment shown in Figure 7;
  • FIG. 11 is a flow chart of another dimension determining method in the embodiment shown in FIG. 7;
  • Figure 12 is still another context model in the embodiment shown in Figure 7;
  • FIG. 13 is a flowchart of an embodiment of a hybrid dimension codec method according to the present invention.
  • FIG. 14 is a flow chart of a dimension determining method in the embodiment shown in FIG. 13;
  • FIG. 15 is a flow chart of another dimension determining method in the embodiment shown in FIG. 13;
  • FIG. 16 is a schematic structural diagram of an embodiment of a hybrid dimension codec apparatus according to the present invention
  • FIG. 17 is a schematic structural diagram of another embodiment of a hybrid dimension codec apparatus according to the present invention.
  • the technical solution provided by the embodiment of the present invention can be applied to the coding end and the decoding end.
  • the coding and decoding expressions are unified into a process. Therefore, the to-be-processed spectrum coefficient that appears in the embodiment of the present invention includes the spectrum to be coded.
  • the coefficients and the spectral coefficients to be decoded, the processed spectral coefficients include the encoded spectral coefficients and the decoded spectral coefficients.
  • FIG. 1 is an embodiment of a context model.
  • the spectrum coefficient X to be processed in the figure, the processed spectrum coefficients 1 to 8, may include N spectral coefficients, and N may take an integer greater than or equal to 1, ie The number of pending spectral coefficients participating in the dimension selection may be N.
  • each frame includes a plurality of spectral coefficients, and before processing the spectral coefficients of each frame, the context needs to be input first.
  • the context of the input can be the context of the previous frame or the context of the first few frames.
  • you can enter only the context of the previous frame that is, the processed spectrum coefficients of the previous frame are mapped according to the length of the frame to be processed.
  • the context of storing the previous frame will be described as an example in the following embodiments.
  • the implementation of the solution provided by the embodiment of the present invention can be implemented by using the multi-frame spectral coefficient as the implementation of the context.
  • 2 is a flowchart of an embodiment of a hybrid dimension codec method according to the present invention. As shown in FIG. 2, the embodiment of the present invention includes:
  • the method provided by the embodiment of the present invention calculates at least one variable set according to the processed spectral coefficient, and determines a processing dimension of the to-be-processed spectral coefficient according to the relationship between the at least one variable set and the corresponding threshold set, according to the selected dimension pair.
  • the technical means for encoding or decoding the spectrum coefficient to be processed in the dimension by using different processing dimensions for different spectral coefficients, improves coding and decoding efficiency.
  • the hybrid dimension codec method of the present invention is further described in conjunction with FIG. 2, and the embodiment of the present invention includes:
  • the number of spectral coefficients to be processed may be N. Before selecting the processing dimension of the spectral coefficients to be processed, it is necessary to determine the number of spectral coefficients to be processed that participate in the dimension selection.
  • 4 or 2 are used as an example for description. The implementation of the other embodiments may be implemented by referring to the solution provided by the embodiment of the present invention.
  • the number of spectral coefficients to be processed is obtained to determine the encoded spectral coefficients used as context.
  • a 16-order context model is taken as an example. As shown in FIG. 3, a black square indicates a determined spectrum coefficient to be encoded.
  • the spectrum coefficient to be encoded is four, and the coded spectrum coefficient as a context may be a black circle.
  • the first frame has 12 spectral coefficients and 4 spectral coefficients of the frame to be processed.
  • 16 coded spectral coefficients can be complemented by adding zeros, and some implementations are implemented. In the manner, it can also be solved by distinguishing the positions of the spectral coefficients to be encoded. Similarly, for other context models, you can refer to the above principle, no longer - for example.
  • At least one variable set may be calculated according to the determined processed spectral coefficient according to the determined number of spectral coefficients to be processed and the context model, the processed spectral coefficients are determined.
  • variable may be the amount of the difference of the position, the energy, the mean, the variance, the mean square error, the minimum variance, the inclination, the dispersion, the dispersion, etc.
  • variable category is only an example.
  • the variable categories used to implement the technical idea of the present invention are all within the protection scope of the embodiments of the present invention.
  • a variable in a variable set can include multiple variables, and multiple variables can belong to one of the above categories, or can belong to different categories. For example, suppose the variable set includes 5 variables, which can be 5 energy variables, or 2 mean variables and 3 variance variables, or 1 energy variable, 2 mean variable, and 2 variance variables. It can be seen that the types of variables and the number of variables included in the variable set can be set according to actual conditions.
  • S102. Determine a processing dimension of the spectral coefficient to be processed according to the relationship between the at least one variable set and the corresponding threshold set.
  • the set of thresholds is set corresponding to the set of variables, and the set of thresholds includes a plurality of thresholds, and the plurality of thresholds are constants, and the values may be obtained through experiments.
  • the set of variables includes a first set of variables and a second set of variables, corresponding to the first set of thresholds and the second set of thresholds, respectively, such that at least one set of variables is
  • the relationship of the corresponding threshold set is a relationship between a variable or a combination of variables in the first variable set and a corresponding threshold in the first set of thresholds, and a combination of variables or variables in the second set of variables and the second set of values
  • the corresponding threshold value relationship may be equal to the second threshold value set in some embodiments, that is, different variable sets may correspond to the same threshold set.
  • the value of the dimension represents the number of spectral coefficients that can be processed at one time.
  • 1D processing means processing one spectral coefficient at a time
  • 4D processing means processing 4 spectral coefficients at a time
  • 16-dimensional processing means processing 16 spectral coefficients at a time.
  • the value of the dimension can be determined according to the number of spectral coefficients to be processed. If the number of spectral coefficients to be processed is two, only one-dimensional or two-dimensional processing can be used. If the spectrum coefficient to be processed is four, then 1 can be used. Dimensional, 2D or 4D processing; If the spectrum factor to be processed is 16, you can use more dimensions to process, such as 16-dimensional processing.
  • the number of available dimensions is at least two, and the specific quantity may be determined according to the variable set and the set number of the wide value set. If only one variable set and one wide value set are included, the number of dimensions can be selected as two; for example, when two variable sets and two wide value sets are included, the dimension The number of the number can be selected; when more sets are included, it can be determined according to the actual situation, and will not be described here.
  • the selection range of the processing dimension of the spectrum coefficient to be processed is also determined.
  • the specific is based on at least one variable set and the corresponding threshold value.
  • the codec in the corresponding dimension may be performed. For example, if you select a 4-dimensional processing dimension, you can perform 4-dimensional encoding, and if you select a 2-dimensional processing dimension, you can perform 2-dimensional encoding. In encoding, if the number of the aforementioned spectral coefficients is four, four spectral coefficients are regarded as one 4-dimensional vector for encoding when performing 4-dimensional encoding; when two-dimensional encoding is performed, two spectral coefficients are regarded as one The 2D vector is encoded. The content of the codec in the dimension is determined according to the selected dimension. For details, refer to the content of the prior art, and details are not described herein again.
  • the codec mode here is not limited, and arithmetic coding may be employed, or any other lossless coding or entropy coding may be employed. A variety of decoding methods can also be used to understand the code side.
  • the method provided by the embodiment of the present invention calculates at least one variable set according to the processed spectral coefficient, and determines a processing dimension of the to-be-processed spectral coefficient according to the relationship between the at least one variable set and the corresponding threshold set, according to the selected dimension pair.
  • the technical means for encoding or decoding the spectrum coefficient to be processed in the dimension by using different processing dimensions for different spectral coefficients, improves coding and decoding efficiency.
  • FIG. 4 is a flowchart of an embodiment of a method for determining a dimension in a hybrid dimension codec method according to the present invention. As shown in FIG. 4, an embodiment of the present invention includes:
  • the at least one variable set includes a first variable set and a second variable set
  • the corresponding threshold set includes the first threshold set and the second set of values as an example, indicating that at least one is calculated according to the processed spectral coefficients.
  • the number of dimensions is determined to be three, and the values of the dimensions are respectively referred to as a first dimension, a second dimension, and a third dimension.
  • the value of the dimension may have multiple options, such as allowing Any integer in the condition, the value of the above three dimensions can be determined according to the actual situation in the specific implementation.
  • the embodiments of the present invention are described in the following, but are not limited to the embodiments of the present invention.
  • FIG. 5 is a flowchart of still another embodiment of a method for determining a dimension in a hybrid dimension codec method according to the present invention.
  • an embodiment of the present invention includes:
  • the at least one variable set and the corresponding threshold set include only one variable set and one wide value set as an example, and the at least one variable set is calculated according to the processed spectral coefficient; according to at least one variable set and the corresponding width
  • the number of the dimensions is determined to be two, and the values of the dimensions are respectively referred to as the fourth dimension and the fifth dimension, and the values of the two dimensions may be determined according to actual conditions in the specific implementation.
  • S402. Determine whether the variable or combination of variables in the variable set is all less than a corresponding threshold in the set of thresholds, and then execute S404, otherwise execute S406.
  • the above setting of the comparison relationship is not limited to the embodiment of the present invention, and may be less than or equal to less than or equal to.
  • the effect of the position of the spectral coefficient to be processed on the dimension is considered, i.e., the processing dimension of the spectral coefficient to be processed is selected based on the processed spectral coefficient and the location of the spectral coefficient to be processed.
  • the hybrid dimension codec method of this embodiment includes: performing time-frequency transform on the input signal to obtain a spectral coefficient to be encoded; selecting a to-be-processed according to the position of the spectral coefficient to be processed and the processed spectral coefficient in order from low frequency to high frequency; The processing dimension of the spectral coefficients; the encoding or decoding of the spectral coefficients to be processed according to the selected dimensions.
  • the position of the spectral coefficient to be processed represents the position in the frame to be processed.
  • a position threshold can be set according to the spectrum coefficient to be processed and the position threshold. The relationship determines the different dimension selection processes.
  • the first range may include multiple Dimensions; the spectrum to be processed that is less than the position threshold in the frame to be processed
  • the coefficient can be set to a second range, and the number of dimensions set in the second range can be less, and can be set to at least one.
  • the dimensions in the second range can also be differentiated, for example, Some points set the range of one dimension selection, while others set the range of another dimension selection.
  • the determination of the range of dimension selection herein can be determined in conjunction with the method of the previous embodiment.
  • the processing dimension selection range of the spectral coefficients to be processed according to the position By selecting the processing dimension selection range of the spectral coefficients to be processed according to the position, the spectral coefficients of some special regions can be obtained with sufficient context for analysis.
  • the above-mentioned setting of the comparison relationship is not limited to the embodiment of the present invention, and may be greater than or equal to less than or equal to less than or equal to.
  • the location value is 4 (representing the 5th pending coefficient of the frame to be processed). Taking a specific set of values as an example, the above embodiment is further described with reference to FIG. 6:
  • S502 The method for selecting a processing dimension of a spectral coefficient to be processed according to the processed spectral coefficient according to the foregoing embodiment, and determining, in the first scope, that the processing dimension is 4D, 2D, or 1D.
  • S504 The method for selecting a processing dimension of a spectral coefficient to be processed according to the processed spectral coefficient provided by the foregoing embodiment, and determining, in the second range, that the processing dimension is two-dimensional or one-dimensional.
  • the selection range of the dimension is 2D or 1D.
  • FIG. 7 is a flowchart of an embodiment of a hybrid dimension codec method according to the present invention. As shown in FIG. 7, the embodiment of the present invention takes an encoding end as an example, and uses energy as a variable to combine specific values by location and energy. Further explain the technical solutions described above:
  • the context is input, and the context is the length of the frame to be encoded according to the current frame to be encoded. Degree mapping comes. Then, according to the position of the current spectral coefficient to be encoded, according to the encoded spectral coefficient, the dimension of the spectral coefficient to be encoded is selected, and the encoded spectral coefficient is encoded according to the selected dimension. After the encoding of the currently to be encoded spectral coefficients is completed, the position counter is incremented to update the current spectral coefficients to be encoded, and then the process of selecting the spectral dimensions to be encoded according to the position of the currently to be encoded spectral coefficients is repeated until the current encoding of the frame to be encoded is completed. After the current encoding of the to-be-coded frame is completed, the context is refreshed to prepare for the encoding of the next frame.
  • the current spectrum coefficient to be encoded when the spectrum coefficient to be encoded is in the 0, 2 position, the current spectrum coefficient to be encoded may be selected in one-dimensional or two-dimensional encoding; when the spectrum coefficient to be encoded is in the 1, 3 position, the current spectrum to be encoded The coefficient is encoded in 1D; when the spectrum coefficient to be encoded is at a position after 4, 4D, 2D, or 1D encoding can be selected.
  • ena, enb, enc, end are the spectral coefficients to be encoded, that is, the selected spectral coefficients to be encoded are four; the other 16 data: ra, rb, rc, rd, va, vb, Vc, vd, la, lb, lc, Id, era, crb, crc, crd are 16 context spectral coefficients of the spectral coefficients to be encoded, ie selected encoded spectral coefficients; here predicted by the 16 encoded spectral coefficients The dimensions of the spectral coefficients ena, enb, enc, end to be encoded.
  • the energy is a variable
  • the variable set includes a first variable set ev, esl er, el, a second variable set wsO, wsl, ws2
  • the threshold set includes a first threshold set a, b, the second set of values c, d, e, where a, b, c, d, e are constants, the values of which are obtained experimentally.
  • the variables ev, esl, er, el, wsO, wsl, ws2 are the energy of multiple coded spectral coefficients in the adjacent region of the spectral coefficient to be processed.
  • the calculation method is as follows:
  • Ev
  • a 3; esl
  • a 3 ; el
  • a 3 ; er
  • a 3 ; wsO
  • a 3 ; Wsl
  • Ws2 er+
  • ena is treated as a 1D vector, encoded by 1D vector.
  • the energy set ev, esl, er, el is obtained according to the processed spectral coefficient; if (ev+vsl) ⁇ a, and (er+el) ⁇ b, the coding dimension is determined as 4D; if not
  • the energy set wsO, wsl, ws3 is obtained; if ws0 ⁇ c and wsKd JL ws3 ⁇ e, then determine that the coding dimension is 2 dimensions; otherwise it is determined that the coding dimension is 1 dimension.
  • the spectral coefficients to be encoded are encoded in 1 or 2 dimensions, which are discussed in two cases:
  • a set of variables is ws0, wsl , and a set of threshold values is c, d.
  • wsO
  • a 3; wsl
  • Wsl is called la
  • 13 is a flowchart of another embodiment of a hybrid dimension codec method according to the present invention. As shown in FIG. 13, the embodiment of the present invention takes a decoding end as an example, and uses a mean value and a variance as variables to pass the combination of position, mean, and variance. The more specific values further explain the technical solutions described above:
  • the context is entered, which maps the decoded frame of the previous frame according to the length of the current frame to be decoded. Then, according to the position of the currently to be decoded spectral coefficient, according to the decoded spectral coefficient, the dimension of the spectral coefficient to be decoded is selected, and the decoded spectral coefficient is decoded according to the selected dimension.
  • the position counter is incremented to update the currently to be decoded spectral coefficients, and then the above process of selecting the spectrum to be decoded according to the position of the currently to be decoded spectral coefficients is repeated until the decoding of the currently to be decoded frame is completed. After the decoding of the current frame to be decoded is completed, the context is refreshed to prepare for decoding of the next frame.
  • the current spectrum to be decoded when the spectrum coefficient to be decoded is in the 0, 2 position, the current spectrum to be decoded may be selected to be 1D or 2D decoding; when the spectrum coefficient to be decoded is in the 1, 3 position, the current spectrum to be decoded
  • the coefficient uses 1D decoding; when the spectrum factor to be decoded is at a position after 4, 4D, 2D, or 1D decoding can be selected.
  • ena, enb, enc, end are the spectral coefficients to be decoded, that is, the selected spectral coefficients to be decoded are four; the other 16 spectral coefficients: ra, rb, re, rd, va, vb , vc, vd, la, lb, lc, Id, era, crb, crc, crd are the 16 context spectral coefficients of the spectral coefficients to be decoded, ie the selected decoded spectral coefficients; here through the 16 decoded spectral coefficients Predict the spectral coefficient ena to be decoded, The dimensions of enb, enc, end.
  • the mean and the variance are variables
  • the variable set includes the first variable set vv, mv, vr, mr, the second variable set vsO, vs 1 , msO, ms 1
  • the threshold set includes The first set of values a, b, the second set of values c, d, where a, b, c, d are constants, the values of which are obtained experimentally.
  • vv, vr, vsO, vsl are the variances of multiple coded spectral coefficients in the adjacent region of the spectral coefficient to be processed
  • mv, mr, ms0, ms 1 are the multiple encoded spectra in the adjacent region of the spectral coefficient to be processed.
  • the mean of the coefficients is calculated as follows:
  • Vv ((
  • Vr ((
  • Ms0 (
  • Vs0 ((
  • Msl (
  • Vsl ((
  • 2D decoding ie: ena, enb is treated as a 2D vector, decoded by 2D vector; otherwise, 1 is selected.
  • Dimension decoding about: ena is treated as a 1-dimensional vector, decoded as a 1-dimensional vector.
  • the mean and variance sets vv, mv, vr, mr are obtained according to the processed spectral coefficients; if (vv+mv) ⁇ a and (w+mr) ⁇ b, the coding dimension is determined. 4D; if not If the condition of (vv+mv) ⁇ a JL(vr+mr) ⁇ b is met, the mean and variance set vsO, vsl, msO, msl are obtained; if (vsO+vsl) ⁇ c and (1 ⁇ 0+11 81) ⁇ Then determine that the coding dimension is 2 dimensions; otherwise it is determined that the coding dimension is 1 dimension.
  • the spectral coefficients to be decoded are decoded using 1D or 2D vectors, which are discussed in two cases:
  • variable set is vsO, vsl, msO, msl
  • set of thresholds is c, d.
  • vsO ((
  • Msl (
  • vsl ((
  • Ms0 (
  • vsO ((
  • Msl (
  • vsl ((
  • select 2D decoding at this time which will be: ena, enb Make a 2D vector and decode it according to the 2D vector; otherwise, select 1D decoding, ie: ena is treated as a 1D vector and decoded as a 1D vector.
  • FIG. 16 is a schematic structural diagram of an embodiment of a hybrid dimension codec apparatus according to the present invention. As shown in FIG. 16, an embodiment of the present invention includes:
  • variable acquisition module 701 configured to calculate at least one variable set according to the processed spectral coefficient
  • the dimension determining module 702 is configured to determine a processing dimension of the spectral coefficient to be processed according to the relationship between the at least one variable set obtained by the variable obtaining module 701 and the corresponding threshold set.
  • Codec module 703 For encoding or decoding the to-be-processed spectral coefficients in the dimension according to the selected dimension.
  • the device provided by the embodiment of the present invention calculates at least one variable set according to the processed spectral coefficient, and determines a processing dimension of the to-be-processed spectral coefficient according to the relationship between the at least one variable set and the corresponding threshold set, according to the selected dimension pair.
  • the technical means for encoding or decoding the spectrum coefficient to be processed in the dimension by using different processing dimensions for different spectral coefficients, improves coding and decoding efficiency.
  • variable obtained by the variable obtaining module 701 includes at least one of position, energy, mean, variance, mean square error, minimum variance, inclination, dispersion, and dispersion.
  • the dimension determining module 702 is further configured to determine a processing dimension of the spectral coefficient to be processed in conjunction with the location of the spectral coefficient to be processed. If the position of the to-be-processed spectral coefficient in the to-be-processed frame is greater than or equal to the positional threshold, the processing dimension is selected within the first range. If the position of the to-be-processed spectral coefficient in the to-be-processed frame is less than the positional threshold, the processing dimension is selected in the second range.
  • the variable acquisition module 701 includes a first variable acquisition unit 801 and a second variable acquisition unit 802.
  • the first variable obtaining unit 801 is configured to acquire a first variable set corresponding to the first threshold set;
  • the second variable acquiring unit 802 configured to acquire the second threshold The set corresponds to the second set of variables.
  • the following variable acquisition module 701 includes a first variable acquisition unit 801 and a second variable acquisition unit.
  • the first variable obtaining unit 801 obtains the first variable set according to the processed spectral coefficient; the dimension determining module 702 determines whether the variable or the variable combination in the first variable set is all smaller than the corresponding threshold in the first threshold set, and if the result If yes, determining that the processing dimension of the spectral coefficient to be processed is the first dimension;
  • the second variable acquisition unit 802 obtains the second variable set according to the processed spectral coefficient, and then the dimension determining module 702 determines whether the variable or the variable combination in the second variable set is all smaller than the second threshold set. In the corresponding threshold, if the result is yes, the processing dimension of the spectral coefficient to be processed is determined to be the second dimension; if the result is no, the processing dimension of the spectral coefficient to be processed is determined to be the third dimension.
  • variable obtaining module 701 is only used to obtain a variable set, and then does not need to be subdivided into units. In this case, the variable obtaining module 701 obtains a variable set according to the processed spectral coefficients, and the dimension determining module 702 determines the obtained variable. Whether the variable or combination of variables in a variable set is all smaller than the corresponding threshold in the corresponding threshold set. If the result is yes, the processing dimension of the spectral coefficient to be processed is determined to be the fourth dimension; if the result is no, the pending is determined. The processing dimension of the spectral coefficients is the fifth dimension.
  • the hybrid dimension codec device in the foregoing embodiment is used to implement the foregoing hybrid dimension codec method, so that the method for performing the method on the mixed-dimension codec device is used in the foregoing method embodiment.
  • the detailed description is only briefly described, and details are not described herein. For details, refer to the foregoing method embodiments.
  • One of ordinary skill in the art can understand all or part of the process in implementing the above embodiments. This may be accomplished by a computer program instructing the associated hardware, which may be stored in a computer readable storage medium, which, when executed, may include the flow of an embodiment of the methods described above.
  • the storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Computational Linguistics (AREA)
  • Signal Processing (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)

Description

混合维度编解码方法和装置 本申请要求了 2010年 1月 13 日提交的, 申请号为 201010042764.9,发明 名称为 "混合维度编解码方法和装置" 的中国专利申请的优先权, 其全部内容 通过引用结合在本申请中。 技术领域
本发明涉及编解码技术领域, 尤其涉及一种混合维度编解码方法和装置。 背景技术
基于上下文的算术编码是一种高性能的无损压缩处理方法,该方法利用信 号的短时稳定性,建立描述相邻帧或相邻若干帧相关性的统计模型,通过分析、 实时地选择最优的模型, 从而获得高效率的无损编码效果。 上下文(context ) 指基于相邻帧对应频点间或者同一帧临近频点间存在某种联系这个事实,当前 待编解码的系数之前已经得到的编解码的系数。
统一语音和音频编码( Unified Speech and Audio Coding, USAC )是移动 图像专家组( Moving Picture Experts Group, MPEG )正在进行的标准。 其中 采用了基于上下文的算术编码, 该方法每一帧的频谱系数从低频到高频每 4个 一组, 形成一个 4维向量空间 4-tuples, 作为编码对象。
在实现本发明创造的过程中, 发明人发现: 现有技术中采取 4-tuples方法 的固定维度编解码方法, 限制了编解码效率的提升。 发明内容
本发明实施例提供一种混合维度编解码方法和装置,对不同的频谱系数采 用不同的维度, 通过多维混合编解码, 提高编解码效率。
本发明实施例提供一种混合维度编解码方法, 包括:
根据已处理频谱系数计算得出至少一个变量集合; 根据至少一个变量集合与对应的阔值集合的关系,确定待处理频谱系数的 处理维度。
根据选择的维度对所述待处理频谱系数进行该维度下的编码或解码。 本发明实施例提供一种混合维度编解码装置, 包括:
变量获取模块: 用于根据已处理频谱系数计算得出至少一个变量集合; 维度确定模块:用于根据所述变量获取模块获得的至少一个变量集合与对 应的阔值集合的关系, 确定待处理频谱系数的处理维度。
编解码模块:用于根据选择的维度对所述待处理频谱系数进行该维度下的 编码或解码。
本发明实施例提供的技术方案,采用根据已处理频谱系数计算得出至少一 个变量集合,根据至少一个变量集合与对应的阔值集合的关系, 确定待处理频 谱系数的处理维度,根据选择的维度对所述待处理频谱系数进行该维度下的编 码或解码的技术手段,通过针对不同的频谱系数采用不同的处理维度,提高了 编解码效率。 附图说明 为了更清楚地说明本发明实施例中的技术方案,下面将对实施例描述中所 需要使用的附图作简单地介绍, 显而易见地, 下面描述中的附图仅仅是本发明 的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动性的前提 下, 还可以根据这些附图获得其他的附图。
图 1为本发明实施例提供的一种上下文模型示意图;
图 2为本发明混合维度编解码方法一个实施例的流程图;
图 3为本发明实施例提供的 16阶上下文模型示意图;
图 4 为本发明混合维度编解码方法中维度确定方法的一个实施例的流程 图;
图 5 为本发明混合维度编解码方法中维度确定方法的又一个实施例的流 程图;
图 6 为本发明混合维度编解码方法一个结合位置确定维度的实施例的流 程图;
图 7为本发明混合维度编解码方法的一个实施例的流程图;
图 8为图 7所示实施例中的一种上下文模型;
图 9为图 7所示实施例中的一个维度确定方法流程图;
图 10为图 7所示实施例中的另一种上下文模型;
图 11为图 7所示实施例中的另一个维度确定方法流程图;
图 12为图 7所示实施例中的又一种上下文模型;
图 13为本发明混合维度编解码方法的一个实施例的流程图;
图 14为图 13所示实施例中的一个维度确定方法流程图;
图 15为图 13所示实施例中的另一个维度确定方法流程图;
图 16为本发明混合维度编解码装置一个实施例的结构示意图; 图 17为本发明混合维度编解码装置另一个实施例的结构示意图。 具体实施方式 下面将结合本发明实施例中的附图,对本发明实施例中的技术方案进行清 楚、 完整地描述, 显然, 所描述的实施例仅仅是本发明一部分实施例, 而不是 全部的实施例。基于本发明中的实施例, 本领域普通技术人员在没有做出创造 性劳动前提下所获得的所有其他实施例, 都属于本发明保护的范围。
本发明实施例提供的技术方案可以应用于编码端和解码端,为使实施例简 要清楚,将编码和解码的表述统一为处理, 所以本发明实施例中出现的待处理 频谱系数包括待编码频谱系数和待解码频谱系数,已处理频谱系数包括已编码 频谱系数和已解码频谱系数。
下面结合图 1 给出一种上下文模型用于简要介绍本发明实施例中应用到 的上下文。假设当前待处理频谱系数为 X, 则可以取临近的已处理频谱系数作 为上下文,图中的数字 1至 8表示已处理频谱系数与待处理频谱系数的相关程 度, 1表示相关程度最强, 8表示相关程度最弱。 从图中可以看出, 待处理频 谱系数与待处理帧的临近的频谱系数相关,也与待处理帧之前的若干帧的已处 理频谱系数有关。 这里需要说明的是, 图 1是一个上下文模型的实施例, 图中 待处理频谱系数 X, 已处理频谱系数 1至 8中, 可以包括 N个频谱系数, N 可以取大于等于 1的整数, 即参与维度选择的待处理频谱系数可以为 N个。
本发明实施例中, 每一帧包括多个频谱系数, 处理每一帧的频谱系数前, 需要首先输入上下文。输入的上下文可以为前一帧的上下文,也可以是前几帧 的上下文。 为节省存储空间, 可以只输入前一帧的上下文, 即将前一帧已处理 频谱系数根据待处理帧的长度映射而来。 为叙述简明,后续实施例中将以储存 前一帧的上下文为例进行说明。以多帧频谱系数作为上下文的实现方式可以参 考本发明实施例提供的方案实现, 不再一一举例。 图 2为本发明混合维度编解码方法一个实施例的流程图, 如图 2所示, 本 发明实施例包括:
S 100、 根据已处理频谱系数计算得出至少一个变量集合。
S102、根据所述至少一个变量集合与对应的阔值集合的关系,确定待处理 频谱系数的处理维度。
S 104、 根据选择的维度对所述待处理频谱系数进行该维度下的编码或解 码。
本发明实施例提供的方法,采用根据已处理频谱系数计算得出至少一个变 量集合,根据至少一个变量集合与对应的阔值集合的关系,确定待处理频谱系 数的处理维度,根据选择的维度对所述待处理频谱系数进行该维度下的编码或 解码的技术手段, 通过针对不同的频谱系数采用不同的处理维度,提高了编解 码效率。 结合图 2对本发明混合维度编解码方法进行进一步说明,本发明实施例包 括:
S 100、 根据已处理频谱系数计算得出至少一个变量集合。
如前所述, 待处理的频谱系数可以是 N个, 在选择待处理频谱系数的处 理维度之前, 需要确定参与维度选择的待处理频谱系数的个数, 为叙述方便, 后续实施例中将以 4个或 2个为例进行说明,选择其他个数的实现方式可以参 考本发明实施例提供的方案实现, 不再一一举例。
获得待处理频谱系数的个数后即可以确定用于作为上下文的已编码频谱 系数。 现以 16阶上下文模型为例进行说明, 如图 3所示, 黑色方块表示确定 的待编码频谱系数, 这里待编码频谱系数为 4个,作为上下文的已编码频谱系 数可以为采用黑色圓形表示的前一帧 12个频谱系数及待处理帧 4个频谱系数。
这里需要说明的是, 在某些实施方式中, 采用上述 16阶的上下文模型时, 如果待编码频谱系数为最初的几个点, 可以通过添零补齐 16个已编码频谱系 数, 某些实施方式中, 也可以通过区分待编码频谱系数的位置的方式解决。 同 理, 对于其他上下文模型, 可以参照上述原理处理, 不再——举例。
本发明实施例中, 可以根据确定的待处理频谱系数的个数与上下文模型、 确定已处理频谱系数, 根据确定的已处理频谱系数计算得出至少一个变量集 合。
本发明实施例中, 变量可以为位置、 能量、 均值、 方差、 均方差、 最小方 差、 倾斜度、 离散度、 离差等表示数据差异的量, 需要说明的是, 这里的变量 类别只是举例说明, 凡能够根据已处理频谱系数计算得出, 用于实现本发明技 术思想的变量类别均在本发明实施例保护范围之内。变量集合中的变量可以包 括多个变量, 多个变量可以同属于上述类别中的一类,也可以分属于不同的类 别。 举例来说, 假设变量集合包括 5个变量, 可以是 5个能量变量, 也可以是 2个均值变量和 3个方差变量, 或者是 1个能量变量、 2个均值变量和 2个方 差变量。 可见, 变量集合包括的变量种类和变量个数可以根据实际情况设定。 S102、根据至少一个变量集合与对应的阔值集合的关系,确定待处理频谱 系数的处理维度。
本发明实施例中, 相应于变量集合设定阔值集合, 阔值集合中包括若干阔 值, 该若干阔值为常量, 其取值可以通过实验得到。
根据至少一个变量集合与对应的阔值集合的关系包括多种情况,举例说明
^口下:
1、 存在一个变量集合和一个阔值集合的情况: 即将一个变量集合中的变 量或变量组合与一个阔值集合中的相应阔值进行比较。
2、 存在两个变量集合和两个阔值集合的情况: 变量集合包括第一变量集 合和第二变量集合,分别对应于第一阔值集合和第二阔值集合,从而至少一个 变量集合与对应的阔值集合的关系为第一变量集合中的变量或变量组合与第 一阔值集合中相应的阔值的关系,以及第二变量集合中的变量或变量组合与第 二阔值集合中相应的阔值的关系, 某些实施方式中, 第一阔值集合也可以与第 二阔值集合相等, 即不同的变量集合可以对应同一个阔值集合。
3、 其他情况: 如存在至少三个变量集合和至少三个阔值集合的情况; 或 变量集合和阔值集合的个数不对等的情况。前述情况可以根据本发明实施例提 供的原则处理, 在此不再——赘述。
维度的取值表示一次可以处理的频谱系数的个数,如 1维处理表示一次处 理 1个频谱系数, 4维处理表示一次处理 4个频谱系数, 16维处理表示一次处 理 16个频谱系数。 维度的取值可以根据待处理频谱系数的个数确定, 如待处 理的频谱系数为 2个, 则只能采用 1维或 2维处理; 如待处理的频谱系数为 4 个, 则可以采用 1维、 2维或 4维处理; 如待处理的频谱系数为 16个, 则可 以采用更多的维数进行处理, 比如 16维处理。
本发明实施例中, 可供选择的维度的数量至少为 2个, 具体数量可以根据 变量集合和阔值集合的集合个数确定。如只包括一个变量集合和一个阔值集合 时, 维度的数量可以选择 2个; 如包括两个变量集合和两个阔值集合时, 维度 的数量可以选择 3个; 包括更多的集合数时, 可以根据实际情况确定, 在此不 再赘述。
本发明实施例中, 维度的取值和维度的数量确定后,待处理频谱系数的处 理维度的选择范围也就确定了。具体的根据至少一个变量集合与对应的阔值集 绍。
S 104、 根据选择的维度对待处理频谱系数进行该维度下的编码或解码。 本发明实施例中,确定待处理频谱系数的维度后即可以进行相应维度下的 编解码。 比如选择了 4维的处理维度即可以进行 4维编码,选择了 2维的处理 维度即可以进行 2维编码等。 在编码时, 如果前述频谱系数的个数为 4个, 在 进行 4维编码时即将 4个频谱系数看作一个 4维向量进行编码;在进行 2维编 码时, 即将 2个频谱系数看作一个 2维向量进行编码。具体的根据选择的维度 进行该维度下的编解码方法, 可以参考现有技术的内容, 在此不再赘述。
这里的编解码方式不做限定, 可以采用算术编码,也可以采用其他任意无 损编码或熵编码。 同理解码端也可以采用多种解码方式。
本发明实施例提供的方法,采用根据已处理频谱系数计算得出至少一个变 量集合,根据至少一个变量集合与对应的阔值集合的关系,确定待处理频谱系 数的处理维度,根据选择的维度对所述待处理频谱系数进行该维度下的编码或 解码的技术手段, 通过针对不同的频谱系数采用不同的处理维度,提高了编解 码效率。
现对混合维度的编解码方法对于编解码效率的提高进行一个简单的说明: 以编码为例, 基于上下文的编码借助相邻系数的相关性获得比较高的编码效 率, 若相邻系数相关性大则编码效率高, 反之亦然。 例如, 相邻四个系数分别 为 12,2,1,2, 由于 12与其他系数相关性不高, 采用统一的 4维联合编码很难找 到合适的概率模型, 导致效率不高。 针对此种情况, 可以将上述系数区分开: {12}, {2,1,2}。 对 12单独选择合适的处理维度进行编码; 对 {2,1,2}单独选择合 适的处理维度进行编码, 效率更高。 图 4 为本发明混合维度编解码方法中维度确定方法的一个实施例的流程 图, 如图 4所示, 本发明实施例包括:
本实施例以至少一个变量集合包括第一变量集合和第二变量集合,相应的 阔值集合包括第一阔值集合和第二阔值集合为例,说明根据已处理频谱系数计 算得出至少一个变量集合; 根据至少一个变量集合与对应的阔值集合的关系, 确定待处理频谱系数的处理维度的方法。 本发明实施例中维度的数量确定为 3 个, 维度的取值分别称为第一维度、 第二维度和第三维度, 如前所述, 维度的 取值可以有多种选择,如在允许条件下的任意整数, 上述三个维度的取值在具 体实现中可以根据实际情况确定。 本发明实施例为叙述方便, 后续以 1维、 2 维和 4维举例进行说明, 但不作为本发明实施例的限定。
S300、 根据已处理频谱系数获得第一变量集合。
S302、判断第一变量集合中的变量或变量组合是否全部小于第一阔值集合 中相应的阔值, 是则执行 S304, 否则执行 S306。
S304、 确定待处理频谱系数的处理维度为第一维度。
S306、 根据已处理频谱系数获得第二变量集合。
S308、判断第二变量集合中的变量或变量组合是否全部小于第二阔值集合 中相应的阔值, 是则执行 S310, 否则执行 S312。
S310、 确定待处理频谱系数的处理维度为第二维度。
S312、 确定待处理频谱系数的处理维度为第三维度。
需要说明的是, 上述关于比较关系的设定并不作为本发明实施例的限定, 小于也可以为小于或等于。 图 5 为本发明混合维度编解码方法中维度确定方法的又一个实施例的流 程图, 如图 5所示, 本发明实施例包括: 本实施例以至少一个变量集合和对应的阔值集合只包括一个变量集合和 一个阔值集合为例,说明根据已处理频谱系数计算得出至少一个变量集合; 根 据至少一个变量集合与对应的阔值集合的关系,确定待处理频谱系数的处理维 度的方法。 本发明实施例中维度的数量确定为 2个, 维度的取值分别称为第四 维度和第五维度, 上述两个维度的取值在具体实现中可以根据实际情况确定。
S400、 根据已处理频谱系数获得一个变量集合。
S402、判断变量集合中的变量或变量组合是否全部小于阔值集合中相应的 阔值, 是则执行 S404, 否则执行 S406。
S404、 确定待处理频谱系数的处理维度为第四维度。
S406、 确定待处理频谱系数的处理维度为第五维度。
需要说明的是, 上述关于比较关系的设定并不作为本发明实施例的限定, 小于也可以为小于或等于。 某些实施方式中, 考虑待处理频谱系数的位置在确定维度时的影响, 即根 据已处理频谱系数和待处理频谱系数的位置, 选择待处理频谱系数的处理维 度。 该实施例的混合维度编解码方法包括: 对输入信号进行时频变换, 获得待 编码频谱系数; 按照从低频到高频的顺序,根据待处理频谱系数的位置和已处 理频谱系数,选择待处理频谱系数的处理维度; 根据选择的维度对待处理频谱 系数进行该维度下的编码或解码。
待处理频谱系数的位置即待处理频谱系数的下标,代表在待处理帧中所处 的位置。 某些情况下,待处理帧中初始的若干个的待处理频谱系数无法获得足 够多的上下文进行分析, 所以某些实施方式中可以设定一个位置阔值, 根据待 处理频谱系数与位置阔值的关系确定不同的维度选择流程。
对于位置阔值前后的待处理系数设定不同维度选择范围,比如对于在待处 理帧中的位置大于等于位置阔值的待处理频谱系数设定一个第一范围,第一范 围内可以包括多个维度;对于在待处理帧中的位置小于位置阔值的待处理频谱 系数可以设定一个第二范围, 第二范围内设定的维度的数量可以少一些, 最少 可以设为一个, 某些实施方式中, 第二范围中的维度也可以进行区分设定, 比 如对于某些点设定一个维度选择的范围, 另一些点设定另一个维度选择的范 围。这里的维度选择范围的确定可以结合前面实施例的方法确定。通过根据位 置区分待处理频谱系数的处理维度选择范围,可以使一些特殊区域的频谱系数 获得足够多的上下文进行分析。 需要说明的是, 上述关于比较关系的设定并不 作为本发明实施例的限定,大于等于也可以为大于,小于也可以为小于或等于。 下面假设位置阔值取值为 4 (代表待处理帧的第 5个待处理系数), 以一 组具体的数值为例, 结合图 6对上述实施例进行进一步说明:
S500、 判断待处理频谱系数的位置是否大于等于 4, 是则执行 S502, 否 则执行 S504。
S502、采用前述实施例提供的根据已处理频谱系数,选择待处理频谱系数 的处理维度的方法, 在第一范围中确定处理维度为 4维、 2维或 1维。
S504、采用前述实施例提供的根据已处理频谱系数,选择待处理频谱系数 的处理维度的方法, 在第二范围中确定处理维度为 2维或 1维。
某些实施方式中, 可以根据第二范围的设定做进一步的区分:
若待处理频谱系数在待处理帧中的位置为 0或 2 (待处理帧的第 1、 3个 待处理系数), 维度的选择范围为 2维或 1维
若待处理频谱系数在待处理帧中的位置为 1或 3 (待处理帧的第 2、 4个 待处理系数), 确定维度为 1维。 图 7为本发明混合维度编解码方法的一个实施例的流程图, 如图 7所示, 本发明实施例以编码端为例, 以能量作为变量, 通过位置和能量结合, 较为具 体的取值对前述介绍的技术方案进行进一步说明:
首先,输入上下文, 该上下文为将前一帧已编码帧根据当前待编码帧的长 度映射而来。 然后, 根据当前待编码频谱系数的位置, 根据已编码频谱系数, 选择待编码频谱系数的维度,根据选择的维度对待编码频谱系数进行编码。 当 前待编码频谱系数编码完成后, 位置计数器增加以更新当前待编码频谱系数, 然后, 重复上述根据当前待编码频谱系数的位置选择待编码频谱维度的过程, 直到当前待编码帧编码完成。 当前待编码帧编码完成后, 刷新上下文, 为下一 帧的编码做准备。
本发明实施例中, 设定当待编码频谱系数处于 0, 2位置时, 当前待编码 频谱系数可以选择 1维或 2维编码; 当待编码频谱系数处于 1, 3位置时, 当 前待编码频谱系数采用 1维编码; 当待编码频谱系数处于 4以后的位置, 可选 择 4维、 2维、 或 1维编码。
设当前待编码频谱系数的位置变量为 pst, 按照上述流程图, 下面提供三 个分支的实例:
情况 1 : 当 8 4时
上下文模型如图 8所示, 其中: ena, enb, enc, end是待编码频谱系数, 即 选择待编码频谱系数为 4个; 其它 16个数据: ra, rb, rc, rd, va, vb, vc, vd, la, lb, lc, Id, era, crb, crc, crd是待编码频谱系数的 16个上下文频谱系数, 即选择的已 编码频谱系数; 这里通过这 16个已编码频谱系数来预测待编码频谱系数 ena, enb, enc, end的维度。
如图 9所示,本实施例中以能量为变量,变量集合包括第一变量集合 ev, esl er, el, 第二变量集合 wsO, wsl, ws2; 阔值集合包括第一阔值集合 a, b, 第二阔 值集合 c, d, e, 其中 a, b, c, d, e为常量, 其取值通过实验得到。
变量 ev, esl, er, el, wsO, wsl, ws2为待处理的频谱系数相邻区域内多个已编 码频谱系数的能量, 计算方法如下所示:
ev=|va|A3+|vb|A3+|vc|A3+|vd|A3; esl=|cra|A3+|crb|A3+|crc|A3+|crd|A3 ; el =|la|A3+ |lb|A3+ |lc|A3+ |ld|A3 ; er=|ra|A3+|rb|A3+|rc|A3+|rd|A3 ; wsO=|va|A3+|vb|A3+|crc|A3+|crd|A3 ; wsl=|lc|A3+ |ld|A3+|vc|A3+|vd|A3 ;
ws2= er+|la|A3+ |lb|A3。
在本实施例中, 可以设定阔值分别为: a = b = 8, c = 64, d = 133, e = 216: 1、 若: ra= 1 , rb=0, rc=0, rd= 1; va= 1 , vb=0, vc= 1 , vd=0;
la=0, lb=0, lc=l, ld=0; cra=0, crb=l, crc=l, crd=0; 则经计算可得: ev=2, er=2, el=l, esl=2;
则:(ev+esl)=4<a 且 (er+el)=3<b, 此时选择 4维编码,即将: ena, enb, enc, end看做一个 4维向量, 按 4维向量进行编码。
2、 若: ra=l, rb=0, rc=0, rd=l; va=l, vb=0, vc=l, vd=0;
la=2, lb=0, lc=l, ld=0; cra=0, crb=l, crc=l, crd=0; 则经计算可得: ev=2, er=2, el=8, esl=2;
贝' J : (ev+esl)=4<8 JL (er+el)=9>8 , 此时只能选择 2或 1维编码。
进一步计算可得:
ws0=2, wsl=2, ws2=10
此时: ws0<c, wsl<d, ws2<e, 符合 2维编码的条件, 选择 2维编码, 即 将: ena, enb看做一个 2维向量, 按 2维向量进行编码。
3、 若: ra= 1 , rb=0, rc=0, rd= 1; va= 1 , vb=0, vc= 1 , vd=3;
la=l, lb=0, lc=5, ld=0; cra=0, crb=l, crc=l, crd=0; 则经计算可得: ev=9, er=28, el=126, esl=2;
贝' J : (ev+esl)=31>a JL (er+el)=128>b, 此时只能选择 2或 1维编码。
进一步计算可得:
ws0=2, wsl=134, ws2=4
此时: ws0<c, wsl〉d, ws2<e, 不符合 2维编码的条件, 选择 1维编码, 即将: ena看做一个 1维向量, 按 1维向量进行编码。
由上述流程可见, 本发明实施例中, 根据已处理频谱系数获得能量集合 ev, esl, er, el; 若 (ev+vsl)<a, 且 (er+el)<b, 则确定编码维度为 4维; 若不符合
(ev+vsl)<a, 且 (er+el)<b的条件, 则获得能量集合 wsO, wsl, ws3; 若 ws0<c且 wsKd JL ws3<e, 则确定编码维度为 2维; 否则确定编码维度为 1维。
情况 2: 当 pst=0或 2时
在该情况下,待编码频谱系数采用 1维或 2维向量编码, 这里分两种情况 进行讨论:
1、当 pst=0时,上下文模型如图 10所示,这里对待编码频谱系数: ena, enb 采用 1维或 2维编码, 其具体判断流程如图 11所示:
在本实施例中, 一个变量集合为 ws0, wsl , 一个阔值集合为 c, d。 设定 阔值分别为: c=35, d=152, 能量变量 wsO和 wsl的计算方法如下:
wsO=|va|A3+|vb|A3; wsl=|la|A3+ |lb|A3+ |lc|A3+ |ld|A3+|vc|A3+|vd|A3; 若: va=2, vb=l, vc=0, vd=l; la=2, lb=l, lc=2, ld=l时:
ws0=9<35; wsl=19<152, 则此时选择 2维编码, 即将: ena, enb看做一个 2维向量, 按 2维向量进行编码。
若: va=2, vb=l, vc=2, vd=3; la=6, lb=0, lc =0, ld=l时:
ws0=9<35; wsl=252>152, 则此时选择 1维编码, 即将: ena看做一个 1 维向量, 按 1维向量进行编码。
2、当 pst=2时,上下文模型如图 12所示,这里对待编码频谱系数: ena, enb 采用 1维或 2维编码, 其具体判断流程图可以采用如图 11所示的方法。
在本实施例中,设定常量分别为: c=27, d=343, wsO和 wsl的计算方法如 下:
wsO=|va|A3+|vb|A3++|crc|A3+|crd|A3;
wsl叫 la|A3+ |lb|A3+ |lc|A3+ |ld|A3+|vc|A3+|vd|A3+|rc|A3+|rd|A3 ;
若: va=2, vb=l, vc=0, vd=l; la=2, lb=l, lc=2, ld=l, rc=l, rd=0, slc=l,sld=0 时:
ws0=10<27; wsl=20<343 , 则此时选择 2维编码, 即将: ena, enb看#丈一个 2维向量, 按 2维向量进行编码。
若: va=2, vb=l, vc=2, vd=0; la=6, lb=0, lc =0, ld=5, rc=l, rd=0, slc=l,sld=0 时:
ws0=10<27; wsl=350>343 , 则此时选择 1维编码, 将将: ena看做一个 1 维向量, 按 1维向量进行编码。
情况 3: pst=l, 3时, 强制使用 1维编码。 图 13 为本发明混合维度编解码方法的另一个实施例的流程图, 如图 13 所示, 本发明实施例以解码端为例, 以均值和方差作为变量, 通过位置、 均值 和方差的结合, 较为具体的取值对前述介绍的技术方案进行进一步说明:
首先,输入上下文, 该上下文为将前一帧已解码帧根据当前待解码帧的长 度映射而来。 然后, 根据当前待解码频谱系数的位置, 根据已解码频谱系数, 选择待解码频谱系数的维度,根据选择的维度对待解码频谱系数进行解码。 当 前待解码频谱系数解码完成后, 位置计数器增加以更新当前待解码频谱系数, 然后, 重复上述根据当前待解码频谱系数的位置选择待解码频谱维度的过程, 直到当前待解码帧解码完成。 当前待解码帧解码完成后, 刷新上下文, 为下一 帧的解码做准备。
本发明实施例中, 设定当待解码频谱系数处于 0, 2位置时, 当前待解码 频谱系数可以选择 1维或 2维解码; 当待解码频谱系数处于 1, 3位置时, 当 前待解码频谱系数采用 1维解码; 当待解码频谱系数处于 4以后的位置, 可选 择 4维、 2维、 或 1维解码。
设当前待解码频谱系数的位置变量为 pst, 按照上述流程图, 下面提供三 个分支的实例:
情况 1 : 当 8 4时
上下文模型如图 8所示, 其中: ena, enb, enc, end是待解码频谱系数, 即 选择待解码频谱系数为 4个;其它 16个频谱系数: ra, rb, re, rd, va, vb, vc, vd, la, lb, lc, Id, era, crb, crc, crd是待解码频谱系数的 16个上下文频谱系数,即选择的 已解码频谱系数;这里通过这 16个已解码频谱系数来预测待解码频谱系数 ena, enb, enc, end的维度。
如图 14所示, 本实施例中以均值和方差为变量, 变量集合包括第一变量 集合 vv, mv, vr, mr, 第二变量集合 vsO, vs 1 , msO, ms 1; 阔值集合包括第一阔值 集合 a, b, 第二阔值集合 c, d, 其中 a, b, c, d为常量, 其取值通过实验得到。
变量 vv, vr, vsO, vsl为待处理的频谱系数相邻区域内多个已编码频谱系数 的方差, mv, mr,ms0, ms 1为待处理的频谱系数相邻区域内多个已编码频谱系 数的均值, 计算方法如下所示:
mv=(|va|+|vb|+|vc|+|vd|+|cra|+|crb|+|crc|+|crd|)/8;
vv=((|va|-mv)A2+(|vb卜 mv)A2+(|vc|-mv)A2+(|vd卜 mv)A2+(|cra卜 mv)A2+(|crb卜 m v)A2+(|crc卜 mv)A2+(|crd卜 mv)A2)/8;
mr=(|la|+ |lb|+ |lc|+ |ld|+|ra|+|rb|+|rc|+|rd|)/8;
vr=((|la卜 mr)A2+(|lb卜 mr)A2+(|lc卜 mr)A2+ (岡 -mr)A2+(|ra卜 mr)A2+(|rb卜 mr)A2+(|r c|-mr)A2+(|rd卜 mr)A2)/8;
ms0=(|va|+|vb|+|crc|+|crd|)/4;
vs0=((|va卜 msO)A2+(|vb卜 msO)A2+(|crc卜 msO)A2+(|crd卜 msO)A2)/4;
msl=(|la|+|lb|+|lc|+|ld|+|ra|+|rb|+|rc|+|rd|+|vc|+|vd|+|cra|+|crb|)/12;
vsl=((|la|-msl)A2+(|lb卜 msl)A2+(|lc卜 msl)A2+(|ld|-msl)A2+(|ra卜 msl)A2+(|rb卜 msl)A2+(|rc卜 msl)A2+(|rd卜 msl)A2+(|vc卜 msl)A2+(|vd卜 msl)A2+(|cra卜 msl)A2+(|crb| -msl)A2)/12。
若 (vv+mv)<a,且 (w+mr)<b, 则选择 4维解码: 即将: ena, enb, enc, end看 做一个 4维向量,按 4维向量进行解码。如果不满足 (vv+mv)<a,且 (vr+mr)<b, 则选择 2或 1维解码。
进一步的,如果 ( 80+¥81)<(;且(11 80+11 81)< 则选择 2维解码, 即将: ena, enb看做一个 2维向量, 按 2维向量进行解码; 否则选择 1维解码, 即将: ena 看做一个 1维向量, 按 1维向量进行解码。
由上述流程可见, 本发明实施例中,根据已处理频谱系数获得均值与方差 集合 vv, mv, vr, mr; 若 (vv+mv)<a且 (w+mr)<b, 则确定编码维度为 4维; 若不 符合 (vv+mv)<a JL(vr+mr)<b 的条件, 则获得均值与方差集合 vsO, vsl, msO, msl ; 若 (vsO+vsl)<c且(1^0+11 81)< 则确定编码维度为 2维; 否则确定编码 维度为 1维。
情况 2: 当 pst=0或 2时
在该情况下,待解码频谱系数采用 1维或 2维向量解码, 这里分两种情况 进行讨论:
1、当 pst=0时,上下文模型如图 10所示,这里对待解码频谱系数: ena, enb 采用 1维或 2维解码, 其具体判断流程如图 15所示:
在本实施例中, 变量集合为 vsO, vsl , msO, msl , 阔值集合为 c, d。 vsO, vsl , msO, msl的计算方法如下:
ms0=(|va|+|vb|)/2;
vsO=((|va卜 msO)A2+(|vb卜 msO)A2)/2;
msl=(|la|+|lb|+|lc|+|ld|+|vc|+|vd|)/6;
vsl=((|la|-msl)A2+|lb卜 msl)A2+|lc卜 msl)A2+|ld卜 msl)A2+(|vc|-msl)A2+(|vd|-ms 1)Λ2)/6。
若 00+1^0)<(;且( 81+11 81)< 则此时选择 2维解码, 即将: ena, enb看 做一个 2维向量, 按 2维向量进行解码; 否则选择 1维解码, 即将: ena看做 一个 1维向量, 按 1维向量进行解码。
2、当 pst=2时,上下文模型如图 12所示,这里对待解码频谱系数: ena, enb 采用 1维或 2维解码, 其具体判断流程可以采用如图 15所示的方法。
在本实施例中, vsO, vsl , msO, msl的计算方法如下:
ms0=(|va|+|vb|+|crc|+|crd|)/4;
vsO=((|va卜 msO)A2+(|vb卜 msO)A2+(|crc卜 msO)A2+(|crd卜 msO)A2)/4;
msl=(|la|+|lb|+|lc|+|ld|+|vc|+|vd|+|rc|+|rd|)/8;
vsl=((|la|-msl)A2+|lb卜 msl)A2+|lc卜 msl)A2+|ld卜 msl)A2+(|vc|-msl)A2+(|vd|-ms l)A2+(|rc卜 msl)A2+(|rd卜 msl)A2)/8。
若 00+1^0)<(;且( 81+11 81)< 则此时选择 2维解码, 即将: ena, enb看 做一个 2维向量, 按 2维向量进行解码; 否则选择 1维解码, 即将: ena看做 一个 1维向量, 按 1维向量进行解码。
情况 3: pst= l, 3时, 强制使用 1维解码。 图 16 为本发明混合维度编解码装置一个实施例的结构示意图, 如图 16 所示, 本发明实施例包括:
变量获取模块 701 : 用于根据已处理频谱系数计算得出至少一个变量集 合;
维度确定模块 702: 用于根据所述变量获取模块 701获得的至少一个变量 集合与对应的阔值集合的关系, 确定待处理频谱系数的处理维度。
编解码模块 703: 用于根据选择的维度对所述待处理频谱系数进行该维度 下的编码或解码。
本发明实施例提供的装置,采用根据已处理频谱系数计算得出至少一个变 量集合,根据至少一个变量集合与对应的阔值集合的关系,确定待处理频谱系 数的处理维度,根据选择的维度对所述待处理频谱系数进行该维度下的编码或 解码的技术手段, 通过针对不同的频谱系数采用不同的处理维度,提高了编解 码效率。
本发明实施例中, 变量获取模块 701获得的变量包括位置、 能量、 均值、 方差、 均方差、 最小方差、 倾斜度、 离散度, 和离差中的至少一个。
某些实施方式中,所述维度确定模块 702还用于结合待处理频谱系数的位 置,确定待处理频谱系数的处理维度。 若待处理频谱系数在待处理帧中的位置 大于等于位置阔值,在第一范围内选择处理维度。 若待处理频谱系数在待处理 帧中的位置小于位置阔值, 在第二范围内选择处理维度。
某些实施方式中, 如图 17所示, 所述变量获取模块 701包括第一变量获 取单元 801和第二变量获取单元 802。 第一变量获取单元 801用于获取与第一 阔值集合对应的第一变量集合; 第二变量获取单元 802: 用于获取与第二阔值 集合对应的第二变量集合。
下面以变量获取模块 701包括第一变量获取单元 801和第二变量获取单元
802为例, 简单说明一下本发明实施例提供的混合维度编解码装置执行上述实 施例提供的方法的流程:
第一变量获取单元 801根据已处理频谱系数获得第一变量集合; 维度确定模块 702 判断所述第一变量集合中的变量或变量组合是否全部 小于第一阔值集合中相应的阔值,若结果为是, 则确定待处理频谱系数的处理 维度为第一维度;
若结果为否,则由第二变量获取单元 802根据已处理频谱系数获得第二变 量集合,然后由维度确定模块 702判断判断第二变量集合中的变量或变量组合 是否全部小于第二阔值集合中相应的阔值, 若结果为是, 则确定待处理频谱系 数的处理维度为第二维度; 若结果为否, 则确定待处理频谱系数的处理维度为 第三维度。
某些实施方式下, 变量获取模块 701只用于获取一个变量集合, 则不必再 细分为单元,此时,变量获取模块 701根据已处理频谱系数获得一个变量集合, 维度确定模块 702 判断获得的一个变量集合中的变量或变量组合是否全部小 于对应阔值集合中相应的阔值, 若结果为是, 则确定待处理频谱系数的处理维 度为第四维度; 若结果为否, 则确定待处理频谱系数的处理维度为第五维度。
需要说明的是, 上述关于比较关系的设定并不作为本发明实施例的限定, 如小于也可以为小于或等于, 大于也可以为大于或等于。
由于前述方法实施例已对混合维度编解码方法进行了比较详细的说明,上 述实施例中的混合维度编解码装置用于实现前述混合维度编解码方法,所以对 混合维度编解码装置执行方法时的具体细节只进行简单说明, 在此不再赘述, 可以参考前述方法实施例的内容。 本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程, 可以通过计算机程序来指令相关的硬件来完成,所述的程序可存储于一计算机 可读取存储介质中, 该程序在执行时, 可包括如上述各方法的实施例的流程。 其中,所述的存储介质可为磁碟、光盘、只读存储记忆体(Read-Only Memory, ROM )或随机存储记忆体 ( Random Access Memory, RAM )等。
以上所述仅是本发明的具体实施方式,应当指出,对于本技术领域的普通 技术人员来说, 在不脱离本发明原理的前提下, 还可以做出若干改进和润饰, 这些改进和润饰也应视为本发明的保护范围。

Claims

权 利 要 求 书
1、 一种混合维度编解码方法, 其特征在于, 所述方法包括:
根据已处理频谱系数计算得出至少一个变量集合;
根据至少一个变量集合与对应的阔值集合的关系,确定待处理频谱系数的 处理维度。
根据选择的维度对所述待处理频谱系数进行该维度下的编码或解码。
2、 根据权利要求 1所述的方法, 其特征在于, 所述变量包括:
位置、 能量、 均值、 方差、 均方差、 最小方差、 倾斜度、 离散度, 和离差 中的至少一个。
3、 根据权利要求 1所述的方法, 其特征在于:
结合待处理频谱系数的位置, 确定待处理频谱系数的处理维度。
4、 根据权利要求 3所述的方法, 其特征在于, 所述方法包括:
若待处理频谱系数在待处理帧中的位置大于等于位置阔值,在第一范围内 选择处理维度。
5、 根据权利要求 1或 4所述的方法, 其特征在于:
所述至少一个变量集合包括第一变量集合和第二变量集合;
所述第一变量集合和第二变量集合分别与第一阔值集合和第二阔值集合 对应。
6、 根据权利要求 5所述的方法, 其特征在于, 所述根据已处理频谱系数 计算得出至少一个变量集合; 根据至少一个变量集合与对应的阔值集合的关 系, 确定待处理频谱系数的处理维度包括:
根据已处理频谱系数获得第一变量集合;
若所述第一变量集合中的变量或变量组合全部小于第一阔值集合中相应 的阔值, 则确定待处理频谱系数的处理维度为第一维度。
7、 根据权利要求 5所述的方法, 其特征在于, 所述根据已处理频谱系数 计算得出至少一个变量集合; 根据至少一个变量集合与对应的阔值集合的关 系, 确定待处理频谱系数的处理维度包括:
根据已处理频谱系数获得第一变量集合;
若所述第一变量集合中的变量或变量组合不全部小于第一阔值集合中相 应的阔值, 则获得第二变量集合;
若所述第二变量集合中的变量或变量组合全部小于第二阔值集合中相应 的阔值, 则确定待处理频谱系数的处理维度为第二维度。
8、 根据权利要求 5所述的方法, 其特征在于, 所述根据已处理频谱系数 计算得出至少一个变量集合; 根据至少一个变量集合与对应的阔值集合的关 系, 确定待处理频谱系数的处理维度包括:
根据已处理频谱系数获得第一变量集合;
若所述第一变量集合中的变量或变量组合不全部小于第一阔值集合中相 应的阔值, 则获得第二变量集合;
若所述第二变量集合中的变量或变量组合不全部小于第二阔值集合中相 应的阔值, 则确定待处理频谱系数的处理维度为第三维度。
9、 根据权利要求 3所述的方法, 其特征在于, 所述方法包括: 若待处理频谱系数在待处理帧中的位置小于位置阔值,在第二范围内选择 处理维度。
10、根据权利要求 1或 9所述的方法, 其特征在于, 所述根据已处理频谱 系数计算得出至少一个变量集合;根据至少一个变量集合与对应的阔值集合的 关系, 确定待处理频谱系数的处理维度包括:
根据已处理频谱系数获得一个变量集合;
若获得的一个变量集合中的变量或变量组合全部小于对应阔值集合中相 应的阔值, 则确定待处理频谱系数的处理维度为第四维度。
11、 根据权利要求 1或 9所述的方法, 其特征在于, 所述根据已处理频谱 系数计算得出至少一个变量集合;根据至少一个变量集合与对应的阔值集合的 关系, 确定待处理频谱系数的处理维度包括:
根据已处理频谱系数获得一个变量集合;
若获得的一个变量集合中的变量或变量组合不全部小于对应阔值集合中 相应的阔值, 则确定待处理频谱系数的处理维度为第五维度。
12、 一种混合维度编解码装置, 其特征在于, 所述装置包括:
变量获取模块: 用于根据已处理频谱系数计算得出至少一个变量集合; 维度确定模块:用于根据所述变量获取模块获得的至少一个变量集合与对 应的阔值集合的关系, 确定待处理频谱系数的处理维度。
编解码模块:用于根据选择的维度对所述待处理频谱系数进行该维度下的 编码或解码。
13、 根据权利要求 12所述的装置, 其特征在于:
所述维度确定模块: 还用于结合待处理频谱系数的位置, 确定待处理频谱 系数的处理维度。
14、根据权利要求 12所述的装置, 其特征在于, 所述变量获取模块包括: 第一变量获取单元: 用于获取与第一阔值集合对应的第一变量集合; 第二变量获取单元: 用于获取与第二阔值集合对应的第二变量集合。
PCT/CN2010/080410 2010-01-13 2010-12-28 混合维度编解码方法和装置 WO2011085630A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US13/457,238 US20120215525A1 (en) 2010-01-13 2012-04-26 Method and apparatus for mixed dimensionality encoding and decoding

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201010042764.9 2010-01-13
CN2010100427649A CN102131081A (zh) 2010-01-13 2010-01-13 混合维度编解码方法和装置

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US13/457,238 Continuation US20120215525A1 (en) 2010-01-13 2012-04-26 Method and apparatus for mixed dimensionality encoding and decoding

Publications (1)

Publication Number Publication Date
WO2011085630A1 true WO2011085630A1 (zh) 2011-07-21

Family

ID=44268952

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2010/080410 WO2011085630A1 (zh) 2010-01-13 2010-12-28 混合维度编解码方法和装置

Country Status (3)

Country Link
US (1) US20120215525A1 (zh)
CN (1) CN102131081A (zh)
WO (1) WO2011085630A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102667923B (zh) 2009-10-20 2014-11-05 弗兰霍菲尔运输应用研究公司 音频编码器、音频解码器、用于将音频信息编码的方法、用于将音频信息解码的方法
SG182466A1 (en) 2010-01-12 2012-08-30 Fraunhofer Ges Forschung Audio encoder, audio decoder, method for encoding and audio information, method for decoding an audio information and computer program using a modification of a number representation of a numeric previous context value
CN103379320B (zh) * 2012-04-16 2016-11-23 华为技术有限公司 视频图像码流处理方法和设备
CN105898334B (zh) * 2016-06-22 2017-12-05 合肥工业大学 一种应用于视频编解码的dc预测电路及其方法

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1871501A (zh) * 2003-10-23 2006-11-29 松下电器产业株式会社 频谱编码装置、频谱解码装置、音响信号发送装置、音响信号接收装置及其使用方法
CN101176147A (zh) * 2005-05-13 2008-05-07 松下电器产业株式会社 语音编码装置以及频谱变形方法
CN101582259A (zh) * 2008-05-13 2009-11-18 华为技术有限公司 立体声信号编解码方法、装置及编解码系统

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5748786A (en) * 1994-09-21 1998-05-05 Ricoh Company, Ltd. Apparatus for compression using reversible embedded wavelets
EP1444688B1 (en) * 2001-11-14 2006-08-16 Matsushita Electric Industrial Co., Ltd. Encoding device and decoding device
US6934677B2 (en) * 2001-12-14 2005-08-23 Microsoft Corporation Quantization matrices based on critical band pattern information for digital audio wherein quantization bands differ from critical bands
EP1489599B1 (en) * 2002-04-26 2016-05-11 Panasonic Intellectual Property Corporation of America Coding device and decoding device
JP3881943B2 (ja) * 2002-09-06 2007-02-14 松下電器産業株式会社 音響符号化装置及び音響符号化方法
US7454076B2 (en) * 2004-06-15 2008-11-18 Cisco Technology, Inc. Hybrid variable length coding method for low bit rate video coding
US7953604B2 (en) * 2006-01-20 2011-05-31 Microsoft Corporation Shape and scale parameters for extended-band frequency coding
US7756281B2 (en) * 2006-05-20 2010-07-13 Personics Holdings Inc. Method of modifying audio content
US7774205B2 (en) * 2007-06-15 2010-08-10 Microsoft Corporation Coding of sparse digital media spectral data
CN101409830B (zh) * 2007-10-10 2010-10-13 华为技术有限公司 Dct系数块相似性判断、图像加解密方法及装置

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1871501A (zh) * 2003-10-23 2006-11-29 松下电器产业株式会社 频谱编码装置、频谱解码装置、音响信号发送装置、音响信号接收装置及其使用方法
CN101176147A (zh) * 2005-05-13 2008-05-07 松下电器产业株式会社 语音编码装置以及频谱变形方法
CN101582259A (zh) * 2008-05-13 2009-11-18 华为技术有限公司 立体声信号编解码方法、装置及编解码系统

Also Published As

Publication number Publication date
CN102131081A (zh) 2011-07-20
US20120215525A1 (en) 2012-08-23

Similar Documents

Publication Publication Date Title
CN107809642B (zh) 用于视频图像编码和解码的方法、编码设备和解码设备
US10003792B2 (en) Video encoder for images
CN103329522B (zh) 用于使用字典编码视频的方法
US20210329267A1 (en) Parallelized rate-distortion optimized quantization using deep learning
US10681388B2 (en) Compression of occupancy or indicator grids
WO2014166434A1 (zh) 深度图像的编解码方法和编解码装置
US11936909B2 (en) Prediction method, encoder, decoder, and computer storage medium
CN104641643A (zh) 在分层层级中信号编码、解码和重构期间残差数据的分解
WO2017201678A1 (zh) 图像预测方法和相关设备
JP5908914B2 (ja) 運動行列を用いた映像の符号化および復号のための方法および装置
CN111742552B (zh) 环路滤波的方法与装置
WO2011085630A1 (zh) 混合维度编解码方法和装置
EP3069449B1 (en) Split gain shape vector coding
Al-Mahmood et al. Lossless image compression based on predictive coding and bit plane slicing
CN105556850B (zh) 编码器、解码器及其运行方法
CN107113426B (zh) 使用广义图形参数执行基于图形的变换的方法和设备
US10382711B2 (en) Method and device for processing graph-based signal using geometric primitives
EP3151559A1 (en) Method for coding and decoding a plurality of picture blocks and corresponding devices
CN106559668B (zh) 一种基于智能量化技术的低码率图像压缩方法
Emara et al. Image compression using advanced optimization algorithms
US20150085923A1 (en) Method for Improving Compression Efficiency of Distributed Source Coding Using Intra-Band Information
CN109845259A (zh) 编码和解码图像参数的方法、编码和解码图像参数的设备及其相应的计算机程序
CN117915107B (zh) 图像压缩系统、图像压缩方法、存储介质与芯片
Ochotorena et al. Regression-based intra-prediction for image and video coding
WO2023231872A1 (zh) 编码方法、解码方法、装置及设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 10842912

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 10842912

Country of ref document: EP

Kind code of ref document: A1