CN102055974B

CN102055974B - Data compressing and uncompressing method, data compressing and uncompressing device and data compressing and uncompressing system

Info

Publication number: CN102055974B
Application number: CN 201010512524
Authority: CN
Inventors: 马鸿飞; 夏雨; 钟政良; 周佳; 刘怡
Original assignee: Huawei Technologies Co Ltd
Current assignee: Huawei Technologies Co Ltd
Priority date: 2010-10-14
Filing date: 2010-10-14
Publication date: 2013-04-17
Anticipated expiration: 2030-10-14
Also published as: CN102055974A; WO2012048662A1

Abstract

The embodiment of the invention discloses a data compressing and uncompressing method, a data compressing and uncompressing device and a data compressing and uncompressing system. The embodiment of the data compressing and uncompressing method comprises the following steps: obtaining the frame of initial data S and a compressing and encoding rule; conducting compressing and encoding on the frame of initial data S according to the compressing and encoding rule, so as to obtain a first-layer encoding error E1 and a first-layer encoding C1; conducting compressing and encoding on the E1 according to the compressing and encoding rule, so as to obtain a second-layer encoding C2; and encapsulating the C1 and the C2, so as to obtain a compressed encoding frame. The invention has the advantages that layered compressing is realized; the method, the device and the system are suitable for compressing not only multimedia information source signals, but also common data; when the method is applied to data transmission, the transmission efficiency of communication signals and data and the communication channel utilization ratio can be improved; and when the method is applied to data storage, the storage efficiency of the communication signals and the data and the utilization ratio of storage equipment can be improved.

Description

Data compression, decompression method, Apparatus and system

Technical field

The present invention relates to communication technical field, particularly data compression, decompression method, Apparatus and system.

Background technology

Along with the very big growth that comprises data service and multimedia communication service demand, the data volume of transmitting by communication system and communication network just constantly increases.Resource-constrained and arm and a leg communication channel can't satisfy this transmission demand that grows with each passing day.So be necessary the research compaction coding method is reduced the data of channel bearing.

Below will describe with regard to the information source compression algorithm:

As shown in Figure 1, be the schematic diagram of sampled data encoder.The frame initial data that the subset separator of encoder will be inputted resolves into first and second liang of subsets; Estimator utilizes the first subset data to estimate the second subset data, and then the valuation that deducts the second subset that obtains of the second subset obtains grouping error data.The first subset data obtains subset mantissa through a derivative coder processes and the subset index sends to formatting module, and error information obtains error mantissa through another derivative coder processes and error extension also sends to formatting module; Two above-mentioned derivative encoders are also exported subset derivative and subset Ha Fuman (Huffman) table that reflects the first subset coded message simultaneously, and the error derivative of reflection error information coded message and error Huffman table, these derivative information have determined to represent the minimum memory space of one group of floating data needs.The header encoder is shown subset derivative and subset Huffman table, error derivative and the error Huffman that receives and be combined into header from the coding parameter that reflects the encoder encodes setting to give formatting module; The packed data formatting module becomes coded frame data with error extension according to certain format combination with header, subset mantissa and subset index and the error mantissa that receives, and forms packed data output.

As shown in Figure 2, be the schematic diagram of sampled data decoder.The packed data analysis module of decoder resolves into header, subset mantissa and subset index and error mantissa and error extension with the compressed coded frames that receives; The header information decoder device is separated into subset derivative and subset Huffman table, error derivative and error Huffman table with the header of receiving; In two integrated decoders, one is utilized subset derivative and subset Huffman table that subset mantissa and subset index are reconstituted the first subset data, and one utilized error derivative and error Huffman table that error mantissa and error extension are reconstituted error information; Estimator estimates the second subset data according to the first subset data that decoding obtains, and with itself and error information addition, obtains the second subset data, and the second last subset data and the first subset data are processed by the subset colligator and obtained reconstruct data output.

Multimedia source signal (such as voice, audio frequency, image and vision signal) can utilize the mature information source encryption algorithm of industry to carry out compressed encoding before sending into channel; Yet the data between each logical block mostly are general data (nonperiodic signal, random signal are called general signal) between each chip or in the sheet between communication terminal, between the inner each several part of communication system or the parts, in the plate, these data do not possess the correlation properties of multimedia source, can not adopt general information source compression algorithm that it is compressed; Owing to only can source signal be compressed, applicability is narrow.

Summary of the invention

The technical problem that the embodiment of the invention will solve provides data compression, decompression method, Apparatus and system; Realization is to the compression of general signal and source signal.

A kind of data compression method comprises:

Obtain the frame of initial data S, obtain compression encoding rules;

According to compression encoding rules the frame of initial data S is carried out compressed encoding, obtain ground floor encoding error E1 and ground floor coding C1;

According to compression encoding rules E1 is carried out compressed encoding and obtain second layer coding C2;

With the coded frame of C2, C1 encapsulation after obtaining compressing.

A kind of uncompressing data comprises:

Obtain coded bit stream B1 and obtain coded frame, obtain decoding decompress(ion) rule;

The coded frame decomposition is obtained ground floor coding C1 and second layer coding C2;

According to decoding decompress(ion) rule the second layer C2 decoding decompress(ion) of encoding is obtained ground floor encoding error E1;

The ground floor coding C1 decompress(ion) of decoding is obtained decoded data S ';

With decoded data S ' and ground floor encoding error E1 addition, obtain reconstruct data Sr1.

A kind of data compression device comprises:

Ground floor compressed encoding module is used for compression encoding rules the frame of initial data S is carried out compressed encoding, obtains ground floor encoding error E1 and ground floor coding C1;

Second layer compressed encoding module is used for according to compression encoding rules E1 being carried out compressed encoding and obtains second layer coding C2;

The frame package module is for the coded frame after C2, C1 encapsulation is obtained compressing.

A kind of data decompression compression apparatus comprises:

Frame deblocking module is used for obtaining coded bit stream B1 and obtains coded frame; The coded frame decomposition is obtained ground floor coding C1 and second layer coding C2;

Second layer decoding decompression module is used for decoding decompress(ion) rule the second layer C2 decoding decompress(ion) of encoding is obtained ground floor encoding error E1;

Ground floor decoding decompression module obtains decoded data S ' for decompress(ion) that ground floor coding C1 is decoded; With decoded data S ' and ground floor encoding error E1 addition, obtain reconstruct data Sr1.

A kind of communication system, comprise: but the data sending terminal and the data receiver that connect with communication mode, it is characterized in that the data compression device that described data sending terminal provides for the embodiment of the invention, the data decompression compression apparatus that the described receiving terminal embodiment of the invention provides.

Technique scheme has following beneficial effect: according to compression encoding rules initial data is carried out compressed encoding, obtain ground floor encoding error E1 and ground floor coding C1; According to compression encoding rules E1 is carried out compressed encoding and obtain second layer coding C2; C2, C1 combination is obtained coded bit stream B1.Realized compressed in layers, the compression that is applicable to the multimedia source signal also is applicable to the compression of general data, and this compression method is applied to transfer of data, can improve signal of communication and data transfer efficient, improves channel utilization; Be applied to the data storage, then can improve signal of communication and data storage efficient, improve the memory device utilance.

Description of drawings

In order to be illustrated more clearly in the technical scheme of the embodiment of the invention, the accompanying drawing of required use was done to introduce simply during the below will describe embodiment, apparently, accompanying drawing in the following describes only is some embodiments of the present invention, for those of ordinary skills, under the prerequisite of not paying creative work, can also obtain according to these accompanying drawings other accompanying drawing.

Fig. 1 is the principle schematic of prior art sampled data encoder;

Fig. 2 is the principle schematic of prior art sampled data decoder;

Fig. 3 is embodiment of the invention data compression device schematic diagram;

Fig. 4 A is embodiment of the invention ground floor compressed encoding module diagram;

Fig. 4 B is embodiment of the invention second layer compressed encoding module diagram;

Fig. 5 A is the undistorted codec schematic diagram of the embodiment of the invention;

Fig. 5 B is embodiment of the invention second layer compressed encoding module diagram;

Fig. 5 C is the embodiment of the invention the 3rd layer compression coding module schematic diagram;

Fig. 6 is embodiment of the invention data compression device schematic diagram;

Fig. 7 is embodiment of the invention data decompression compression apparatus schematic diagram;

Fig. 8 is embodiment of the invention second layer decoding decompression module schematic diagram;

Fig. 9 is embodiment of the invention ground floor decoding decompression module schematic diagram;

Figure 10 is embodiment of the invention data decompression compression apparatus schematic diagram;

Figure 11 A is embodiment of the invention second layer uncompressed encoding module diagram;

Figure 11 B is the 3rd layer of uncompressed encoding module diagram of the embodiment of the invention;

Figure 11 C is embodiment of the invention distortion coding structure schematic diagram;

Figure 11 D is the undistorted coding structure schematic diagram of the embodiment of the invention;

Figure 12 A is embodiment of the invention communication system schematic diagram;

Figure 12 B is embodiment of the invention communication system schematic diagram;

Figure 13 A is embodiment of the invention storage system schematic diagram;

Figure 13 B is embodiment of the invention storage system schematic diagram;

Figure 14 A is embodiment of the invention data compression method schematic flow sheet;

Figure 14 B is the Time Domain Amplitude example schematic of embodiment of the invention initial data S;

Figure 14 C is embodiment of the invention DCT coefficient schematic diagram;

Figure 15 A is that the embodiment of the invention is based on the Coding Compression Algorithm schematic flow sheet of dct transform;

Figure 15 B is the schematic flow sheet of embodiment of the invention second layer coding C2;

Figure 15 C is the probability distribution example schematic of embodiment of the invention DCT coefficient;

Figure 15 D is the probability distribution example schematic of embodiment of the invention error E 1;

Figure 15 E is the example schematic of 16 code word probability distribution among the embodiment of the invention 4 bits of encoded Parameters of Time-frequency Fields coding Cr;

Figure 15 F is the example schematic of 16 code word probability distribution among the embodiment of the invention 4 bits of encoded time errors coding Ce;

Figure 16 is embodiment of the invention uncompressing data schematic flow sheet;

Figure 17 is that the embodiment of the invention is based on the decoding decompression algorithm schematic flow sheet of dct transform.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the invention, the technical scheme in the embodiment of the invention is clearly and completely described, obviously, described embodiment only is the present invention's part embodiment, rather than whole embodiment.Based on the embodiment among the present invention, those of ordinary skills belong to the scope of protection of the invention not making the every other embodiment that obtains under the creative work prerequisite.

The embodiment of the invention will disclose compression, coding and transmission method and the device of efficient communication signal and general data.Wherein the effect of encoder is that the initial data S that inputs is carried out compressed encoding, and the coded bit stream that then compression is obtained is sent into transmission or deposited storage medium in; The effect of decoder is that the compression bit stream that comes self-channel or storage medium is decoded and decompressed, and obtains reconstruct data.The scheme of the embodiment of the invention both can be applicable to signal of communication and data compression and transmission, also can be applied to go in the compressed encoding of other general data and the communications, can also be applied to go in the compressed encoding of the multimedia messagess such as voice and audio frequency and the communications.

The below introduces respectively the Code And Decode method and describes.

The embodiment of the invention provides a kind of data compression device, as shown in Figure 3, comprising:

Need to prove, the statement of all uses " first " and " second " all is in order to distinguish two non-identical entities of same names or non-identical parameter in the embodiment of the invention, as seen " first " and " second " only conveniently should not be construed as restriction to the embodiment of the invention for what explain, and subsequent embodiment is no longer one by one explanation to this.

Ground floor compressed encoding module 301, the frame for obtaining initial data S obtains compression encoding rules; According to compression encoding rules the frame of initial data S is carried out compressed encoding, obtain ground floor encoding error E1 and ground floor coding C1;

Compression encoding rules can be frame originating point information Hi; It also can be predefined compression encoding rules.Will not limit this embodiment of the invention, in subsequent embodiment, all describe as an example of frame originating point information Hi example.Frame originating point information Hi can comprise length information, error checking and correction information of frame etc.; The frame originating point information that needs to be used for compression can obtain from frame, also can store by predefined mode will not limit this embodiment of the invention.

Carry out in the process of compressed encoding at the frame to initial data S, there is error in the coded frame of the result of compressed encoding before with respect to compressed encoding, and for example: above-mentioned E1 is that the result (ground floor coding C1) of compressed encoding is with respect to the error of original encoding frame.

Second layer compressed encoding module 302 is used for according to frame originating point information Hi E1 being carried out compressed encoding and obtains second layer coding C2; Above-mentioned compressed encoding generally has compression and two parts of encoding, and these two parts can be carried out respectively, also can carry out simultaneously.In losing genuine encryption algorithm, can carry out respectively, in undistorted coding, can carry out simultaneously.

Frame package module 303 is for the coded frame after C2, C1 encapsulation is obtained compressing.

If compression encoding rules is Hi, above-mentioned frame package module 303 then, concrete coded frame after being used for C2, Hi and C1 encapsulation obtained compressing.If compression encoding rules is predefined compression encoding rules, Hi can be encapsulated in the coded frame.

Just can obtain coded bit stream B1 by the coded frame after the compression in addition.The structure that is encapsulated the coded frame after obtaining compressing by C2, Hi and C1 can be with reference to Figure 11 C in the subsequent embodiment.

Among Fig. 3, the initial data S of input is admitted to ground floor compressed encoding module 301 and carries out compressed encoding, according to frame originating point information, generate ground floor encoding error E1 and ground floor coding C1, ground floor encoding error E1 and the ground floor coding C1 that obtains sent into respectively second layer compressed encoding module 302 and frame package module 303; According to frame originating point information, ground floor encoding error E1 obtains second layer coding C2 through the compressed encoding of second layer compressed encoding module 302; At last, frame package module 303 is added frame originating point information Hi with ground floor C1, frame originating point information Hi and the second layer coding C2 combination of encoding, and obtains complete coded frame output, forms coded bit stream B1.

Particularly, shown in Fig. 4 A, above-mentioned ground floor compressed encoding module 301 comprises:

Time frequency analysis module 3011 is used for that initial data S is carried out time frequency analysis and obtains frequency domain parameter Fp;

Quantization encoding module 3012 is used for according to frame originating point information Hi frequency domain parameter Fp being quantized to obtain frequency domain parameter coding Cr; Because the value that initial data S time frequency analysis is obtained frequency domain parameter Fp has a lot and is distributed in the confirmable scope, by quantizing and with the frequency domain parameter Fp value of phase with a Parametric Representation, can reducing data volume like this; What quantize is embodied in the subsequent embodiment and will be described in more detail.

Decoding inverse quantization module 3013 is for the frequency domain parameter Fp ' that obtains decoding according to frequency domain parameter coding Cr execution decoding inverse quantization;

Time-frequency synthesis module 3014 is used for decoding frequency domain parameter Fp ' is carried out the synthetic decoded data S ' that obtains of time-frequency;

Subtraction block 3015 is used for initial data S and decoded data S ' are subtracted each other, and obtains ground floor encoding error E1;

Entropy coding module 3016 is used for that frequency domain parameter coding Cr is carried out the entropy coding and obtains ground floor coding C1.

In Fig. 4 A, initial data S obtains frequency domain parameter Fp through time frequency analysis module 3011, under the guidance of frame originating point information, 3012 pairs of frequency domain parameters of quantization encoding module quantize to produce frequency domain parameter coding Cr, and the frequency domain parameter coding is sent into decoding inverse quantization module 3013 and entropy coding module 3016 simultaneously; Decoding inverse quantization module 3013 is carried out the local decode frequency domain parameter Fp ' that obtains decoding according to frequency domain parameter coding, the decoding frequency domain parameter is again through the synthetic decoded data S ' that obtains of time-frequency, initial data S and decoded data S ' are subtracted each other the error that obtains, be referred to as in the present embodiment ground floor encoding error E1; Simultaneously, 3016 pairs of frequency domain parameter codings of entropy coding module Cr carries out the coding that the entropy coding obtains, and is referred to as in the present embodiment ground floor coding C1.

Shown in Fig. 4 B, second layer compressed encoding module 302 can comprise quantization encoding module 3021 and entropy coding module 3022.Ground floor encoding error E1 is admitted to second layer compressed encoding module 302, through obtaining error coding Ce behind quantization encoding module 3021 codings, again through obtaining second layer compressed encoding C2 behind entropy coding module 3022 codings.

Further, above-mentioned second layer compressed encoding module 302 also is used for before E1 is carried out compressed encoding E1 being classified by bit-depth, and sorted residual error is carried out non-homogeneous scalar quantization by the scope of its bit-depth.In subsequent embodiment, will be described in more detail at this point.Wherein, what bit-depth was described is the precision of data, can refer to represent the number of significant digit of data, or the figure place of binary number.Wherein, the purpose of classification is the quality that is to improve quantification; By the mode of bit-depth classification for example: bit-depth is the class that is divided into of 10 bits, 9 bits be divided into a class, etc.

Further, shown in Fig. 5 A, owing to having used quantification and the coding that can bring encoding error to process among the above embodiment, causing encoding device is the distortion encoder.The present embodiment has further designed undistorted codec.

Above-mentioned second layer compressed encoding module 302 is used for ground floor encoding error E1 being carried out data compression coding, the second layer encoding error E2 that obtains; E1 carries out data compression coding to the ground floor encoding error, obtains second layer coding C2; Also comprise:

The 3rd layer compression coding module 501 is used for according to frame originating point information Hi E2 being carried out compressed encoding and obtains the 3rd layer of coding C3;

Above-mentioned frame package module 303, concrete for the coded frame after C3, C2, Hi and C1 encapsulation are obtained compressing.

In Fig. 5 A, the initial data S of input is input to ground floor compressed encoding module 301 and carries out data compression coding, will obtain ground floor encoding error E1 and ground floor coding C1 and give respectively second layer compressed encoding module 302 and frame package module 303; Ground floor encoding error E1 is admitted to second layer compressed encoding module 302 and carries out data compression coding, and the second layer encoding error E2 that obtains and second layer coding C2 are given respectively the 3rd layer compression coding module 501 and frame package module 303; Second layer encoding error E2 is input to the 3rd layer compression coding module 303 and carries out data compression coding, generates second layer coding C3; At last, frame package module 303 is combined encode C1, second layer coding C2 and the 3rd layer of coding C3 of ground floor, adds frame originating point information Hi, obtains complete coded frame output, forms coded bit stream B2.

Particularly, shown in Fig. 5 B, above-mentioned second layer compressed encoding module 302 comprises:

Time frequency analysis module 3021 is used for that E1 is carried out time frequency analysis and obtains frequency domain parameter F1p;

Quantization encoding module 3022 is used for according to frame originating point information Hi frequency domain parameter F1p being quantized to obtain frequency domain parameter coding C1r;

Decoding inverse quantization module 3023 is used for carrying out the local decode frequency domain parameter F1p ' that obtains decoding according to frequency domain parameter coding C1r;

Time-frequency synthesis module 3024 is used for decoding frequency domain parameter F1p ' is carried out the synthetic error E 1 ' of decoding that obtains of time-frequency;

Subtraction block 3025 is used for E1 is subtracted each other with decoding error E 1 ', obtains second layer encoding error E2;

Entropy coding module 3026 is used for that frequency domain parameter coding C1r is carried out the entropy coding and obtains second layer coding C2.

In Fig. 5 B, input data E1 obtains frequency domain parameter F1p through time frequency analysis module 3021, under the guidance of frame originating point information, 3022 pairs of frequency domain parameters of quantization encoding module quantize to produce frequency domain parameter coding C1r, and field parameter coding C1r is sent into decoding inverse quantization module 3023 and entropy coding module 3026 simultaneously; Decoding inverse quantization module 3023 is carried out the local decode frequency domain parameter F1p ' that obtains decoding according to field parameter coding C1r, decoding frequency domain parameter F1p ' is again through the synthetic error E 1 that obtains decoding of time-frequency "; will input data E1 and subtract each other with decoding error E 1 ', generate second layer encoding error E2.

Shown in Fig. 5 C, provided the schematic diagram of the 3rd layer compression coding module 503 in the undistorted encoder.Second layer encoding error E2 is admitted to the 3rd layer compression coding module 501, obtains error coding C2e through quantization encoding module 5011, passes through entropy coding module 5012 again, obtains the 3rd layer compression coding C3.

Further, as shown in Figure 6, said apparatus also comprises:

Sending module 601 is used for that above-mentioned coded bit stream B1 is sent to transmission medium and transmits, and perhaps above-mentioned coded bit stream B1 is sent to storage medium stores.

Among the above embodiment, because coded bit stream B1 is obtained by the frame after the compression, thereby the relative initial data of above-mentioned coded bit stream B1 is less, can save transfer resource in transmission course, and can save memory space during storage.

Accordingly, the embodiment of the invention also provides a kind of data decompression compression apparatus, as shown in Figure 7, comprising:

Frame deblocking module 701 is used for obtaining coded bit stream B1 and obtains coded frame; The coded frame decomposition is obtained ground floor coding C1, second layer coding C2;

Second layer decoding decompression module 702 is used for decoding decompress(ion) rule the second layer C2 decoding decompress(ion) of encoding is obtained ground floor encoding error E1; Above-mentioned decoding decompress(ion) rule can be that predefined decoding decompress(ion) rule also can be frame originating point information Hi; If decoding decompress(ion) rule is frame originating point information Hi, then above-mentioned frame deblocking module 701 specifically is used for obtaining coded bit stream B1 and obtains coded frame; The coded frame decomposition is obtained ground floor coding C1, frame originating point information Hi and second layer coding C2; Compression encoding rules in above-mentioned decoding decompress(ion) rule and the previous embodiment is corresponding relation.

Ground floor decoding decompression module 703 obtains decoded data S ' for decompress(ion) that ground floor coding C1 is decoded; With decoded data S ' and ground floor encoding error E1 addition, the data that obtain are reconstruct data Sr1.

Receive coded frame from coded bit stream B1 after, frame deblocking module 701 becomes second layer coding C2, frame originating point information Hi and ground floor coding C1 with the information decomposition in the coded frame; Second layer encoded data stream C2 is admitted to second layer decoding decompression module 702, and according to frame originating point information, the decoding decompress(ion) obtains ground floor encoding error E1; Ground floor encoding error E1 and ground floor coding C1 are admitted to ground floor decoding decompression module 703 together, generating solution code data S ' behind the ground floor coding C1 process decoding decompress(ion); Last decoded data S ' and ground floor encoding error E1 addition obtain reconstruct data Sr1 output.

Particularly, as shown in Figure 8, second layer decoding decompression module 702 comprises:

Entropy decoder module 7021 is used for that second layer compressed encoding C2 is carried out the entropy decoding and obtains error coding Ce;

Decoding inverse quantization module 7022 is used for that error coding Ce is carried out inverse quantization and obtains ground floor encoding error E1.

The second layer compressed encoding C2 of input obtains error coding Ce through entropy decoder module 7021, obtains ground floor encoding error E1 through inverse quantization module 7022 again.

Particularly, as shown in Figure 9, ground floor decoding decompression module 703 comprises:

Entropy decoder module 7031 is used for that ground floor coding C1 is carried out the entropy decoding and obtains frequency domain parameter coding Cr;

Decoding contravariant quantization modules 7032 is used for according to frame originating point information Hi frequency domain parameter coding Cr being carried out the inverse quantization frequency domain parameter Fp ' that obtains decoding;

Time-frequency synthesis module 7033 is used for decoding frequency domain parameter Fp ' is carried out the synthetic decoded data S ' that obtains of time-frequency;

Addition module 7034 is used for decoded data S ' and ground floor encoding error E1 addition are obtained reconstruct data Sr1.

Ground floor encoding error E1 and ground floor coding C1 are admitted to ground floor decoding decompression module 703 together, decoding obtains frequency domain parameter coding Cr to ground floor coding C1 through entropy, the frequency domain parameter coding is admitted to decoding inverse quantization module 7032, obtain frequency domain parameter Fp according to the header inverse quantization, by frequency domain parameter Fp, time-frequency synthesis module 7033 produces decoded data S '; Last decoded data S ' and ground floor encoding error E1 addition obtain reconstruct data Sr1 output.

Further, as shown in figure 10, above-mentioned frame deblocking module 701 also obtains the 3rd layer of coding C3 with the coded frame decomposition; Also comprise:

The 3rd decoding decompression module 1001 is used for according to frame originating point information Hi, and the 3rd layer of coding C3 decompress(ion) of decoding is obtained second layer encoding error E2;

Above-mentioned second layer uncompressed encoding module 702 is used for according to frame originating point information Hi the second layer coding C2 decompress(ion) of decoding is obtained the error E 1 ' of decoding, and the error E of will decode 1 ' and second layer encoding error E2 addition obtain ground floor encoding error E1.

Decoding device receives coded frame from coded bit stream B2 after, frame deblocking module 701 becomes frame originating point information Hi, ground floor coding C1, second layer coding C2 and the 3rd layer of coding C3 with the information decomposition in the coded frame; The 3rd layer of encoded data stream C3 is admitted to the 3rd layer decoder decompression module 1001, and according to frame originating point information, the decoding decompress(ion) obtains second layer encoding error E2; Second layer encoding error E2 and second layer coding C2 are admitted to second layer decoding decompression module 702 together, generating solution code error E1 ' behind the second layer coding C2 process decoding decompress(ion), decoding error E 1 ' and second layer encoding error E2 addition obtain ground floor encoding error E1; The ground floor encoding error E1 that obtains and ground floor coding C1 are sent into ground floor decoding decompression module 703 decompress(ion) of decoding, generating solution code data S ' behind the ground floor coding C1 process decoding decompress(ion), decoded data S ' and ground floor encoding error E1 addition obtain reconstruct data Sr2 output.

Particularly, shown in Figure 11 A, above-mentioned second layer uncompressed encoding module 702 comprises:

Entropy decoder module 7021 is used for that second layer coding C2 is carried out the entropy decoding and obtains frequency domain parameter coding C1r;

Decoding inverse quantization module 7022 is used for according to frame originating point information Hi frequency domain parameter coding C1r being carried out the inverse quantization frequency domain parameter F1p ' that obtains decoding;

Time-frequency synthesis module 7023 is used for decoding frequency domain parameter F1p ' is carried out the synthetic error E 1 that obtains decoding of time-frequency;

Addition module 7024 is used for will decode error E 1 ' and second layer encoding error E2 addition, obtains ground floor encoding error E1.

In Figure 11 A, second layer encoding error E2 and second layer coding C2 are admitted to second layer decoding decompression module 702 together, decoding obtains frequency domain parameter coding C1r to second layer coding C2 through entropy, the frequency domain parameter coding is admitted to decoding inverse quantization module 7022, obtain frequency domain parameter F1p according to the header inverse quantization, by frequency domain parameter, time-frequency synthesis module 7023 produces decoding error E 1 '; The error E of decoding at last 1 ' and second layer encoding error E2 addition obtain ground floor encoding error E1 output.Correspondingly, the structure of undistorted coded frame and the coded bit stream B2 that is comprised of continuous undistorted coded frame are provided by following table.As seen, compare with the distortion coded frame is arranged, in undistorted coded frame, increased second layer coded portion.

Shown in Figure 11 B, provided the structural representation of the 3rd layer of uncompressed encoding module 1001 in the undistorted decoder.The 3rd layer compression coding C3 of input obtains error coding C2e through entropy decoder module 10011, obtains second layer encoding error E2 through decoding inverse quantization module 10012 again.

Frame package module 303 is added frame originating point information with the ground floor coding C1 and the second layer coding C2 combination that obtain in the encoder, forms a complete coded frame, and Figure 11 C has provided the structural representation of distortion coded frame.The different coding frame is arranged according to the time sequencing of output, just formed the coded bit stream B1 of encoder output, shown in Figure 11 C.

Frame originating point information among Figure 11 C is the optional information in the coded frame, and when codec was operated in fixing mode of operation, frame originating point information can remove need not; But during the pattern that becomes when codec is operated in, frame originating point information is carrying the coding mode of current encoded frame, and this must pass to decoder end, so can not save.The information that frame originating point information may comprise has the information such as frame length, coding mode, code rate, quantization encoding form, and this is need to be according to concrete application demand and definite.

The structure of undistorted coded frame and the coded bit stream B2 that is comprised of continuous undistorted coded frame are provided by Figure 11 D.As seen, compare with the distortion coded frame is arranged, in undistorted coded frame, increased second layer coded portion.

The embodiment of the invention also provides a kind of communication system in addition, can be with reference to figure 12A, comprise: but the data sending terminal 1201 and the data receiver 1202 that connect with communication mode, the data decompression compression apparatus of any one that the data compression device of any one that provides for the embodiment of the invention of above-mentioned data sending terminal 1201 wherein, above-mentioned data receiver 1202 provide for the embodiment of the invention.Below two systems be that example describes with regard to data receiver 1202 for communication equipment or memory device respectively for example.

The embodiment of the invention also provides a kind of communication system, shown in Figure 12 A, comprise: but the data sending terminal 1201 and the data receiver 1202 that connect with communication mode, the data decompression compression apparatus of any one that the data compression device of any one that provides for the embodiment of the invention of above-mentioned data sending terminal 1201 wherein, above-mentioned receiving terminal 1202 provide for the embodiment of the invention.

Shown in Figure 12 B, one of communication system for example, the embodiment of the invention is disclosed to have distortion and undistorted codec can be applied among the system shown in Figure 12 B, and encoder 1201A carries out remote data communication by all kinds of transmission channels or network 1303A with decoder 1202A and (communicates by letter such as data, file transfer, voice communication, voice communication etc.), the data communication of process and internal system (such as the IQ transfer of data in the common public radio interface and zoom out etc.), in the equipment or the transfer of data (such as the exchanges data of AD/DA converter and other parts) between the plate inner module etc.Data flow is: initial data S, and encoded device 1201A coding obtains coded bit stream B, and coded bit stream B arrives decoder 1202A through transmission channel 1303A, obtains reconstruct data Sr through decoding

The embodiment of the invention also provides a kind of storage system, as shown in FIG. 13A, comprise: encoding device 1301 and decoding device 1302, the data decompression compression apparatus of any one that the data compression device of any one that provides for the embodiment of the invention of above-mentioned encoding device 1301 wherein, above-mentioned decoding device 1302 provide for the embodiment of the invention.

Shown in Figure 13 B, one of storage system for example, the embodiment of the invention is disclosed have distortion and undistorted codec can be applied in shown in Figure 13 B among the system that storage is used, such as the efficient storage of data, digital media storage, digital library etc.The flow direction of data is: the encoded device 1301A of initial data S coding obtains coded bit stream B, and coded bit stream is sent to storage device 1302A and stores, and coded bit stream B is gone out out to be entered decoder 1303A decoding and obtain reconstruct data Sr.

The embodiment of the invention provides a kind of data compression method, shown in Figure 14 A, comprising:

1401: obtain the frame of initial data S, obtain compression encoding rules;

Compression encoding rules can be frame originating point information Hi; Also can be that predefined compression encoding rules will not limit this embodiment of the invention, in subsequent embodiment, all describe as an example of frame originating point information Hi example.

1402: according to compression encoding rules above-mentioned coded frame is carried out compressed encoding, obtain ground floor encoding error E1 and ground floor coding C1;

Therefore above-mentioned 1402 can be: according to Hi above-mentioned coded frame is carried out compressed encoding, obtain ground floor encoding error E1 and ground floor coding C1; More specifically, in above-mentioned 1402, obtain ground floor encoding error E1 and ground floor coding C1 according to Hi, comprising:

Initial data S is carried out time frequency analysis obtain frequency domain parameter Fp;

According to frame originating point information Hi frequency domain parameter Fp is quantized to obtain frequency domain parameter coding Cr,

Carry out the decoding inverse quantization frequency domain parameter Fp ' that obtains decoding according to frequency domain parameter coding Cr, decoding frequency domain parameter Fp ' is carried out time-frequency is synthetic to obtain decoded data S ', initial data S and decoded data S ' are subtracted each other, obtain ground floor encoding error E1; Frequency domain parameter coding Cr is carried out the entropy coding obtain ground floor coding C1.

More specifically, above-mentionedly initial data S carried out time frequency analysis obtain frequency domain parameter Fp, comprising:

Every frame to initial data S carries out discrete cosine transform (Discrete Cosine Transform, DCT) direct transform, obtains frequency domain parameter Fp; The method that need to prove time frequency analysis can also be: revise cosine transform (Modified Discrete Cosine Transform, MDCT), fast Fourier transform (FastFourier Fransform, FFT), wavelet transform etc. (Discrete Wavelet Transform, DWT) etc.DCT is that one of them should not be construed as restriction to the embodiment of the invention for example.

Above-mentionedly according to frame originating point information Hi frequency domain parameter Fp is quantized to obtain frequency domain parameter coding Cr, comprising:

According to frequency domain parameter Fp amplitude distribution, to the frequency domain parameter Fp processing of classifying, then quantize to obtain frequency domain parameter coding Cr;

Above-mentioned compressed encoding is non-homogeneous scalar quantization coding.

More specifically, above-mentioned entropy is encoded to: at least one item in Ha Fuman Huffman coding, arithmetic coding, other unfixed-length coding.

More specifically, above-mentioned classification is processed and comprised: the frequency domain parameter Fp to full dynamic range advances non-homogeneous scalar quantization, perhaps, the frequency domain parameter Fp in the setting range is carried out non-homogeneous scalar quantization.

1403: according to compression encoding rules E1 is carried out compressed encoding and obtain second layer coding C2;

Particularly, above-mentioned 1403 can be: according to frame originating point information Hi E1 is carried out compressed encoding and obtain second layer coding C2.

Further, in 1403, before E1 is carried out compressed encoding, also comprise:

E1 is classified by bit-depth, sorted residual error is carried out non-homogeneous scalar quantization by the scope of its bit-depth.

Particularly, above-mentioned 1403 carry out compressed encoding according to frame originating point information Hi to E1 obtains second layer coding C2, comprising: ground floor encoding error E1 is carried out quantization encoding obtain error coding Ce, error coding Ce is carried out the entropy coding obtain second layer compressed encoding C2.

1404: the coded frame after C2 and C1 encapsulation obtained compressing.Just can obtain coded bit stream B1 by the coded frame after the compression in addition.

If compression encoding rules is Hi, then 1404 can be coded frame after C2, Hi and C1 encapsulation obtained compressing.If compression encoding rules is predefined compression encoding rules, Hi can be encapsulated in the coded frame.

Further, the coded frame of above-mentioned 1404 encapsulation after obtaining compressing comprise: select variable bit rate or fixed rate to encapsulate coded frame after obtaining compressing according to the coded frame characteristic complexity of initial data S.

Further, before the above-mentioned coded frame after C2, Hi and C1 encapsulation is obtained compressing, also comprise:

E1 carries out data compression coding to the ground floor encoding error, the second layer encoding error E2 that obtains;

E1 carries out data compression coding to the ground floor encoding error, obtains second layer coding C2;

Above-mentioned 1403 carry out compressed encoding according to frame originating point information Hi to E1 obtains second layer coding C2, comprising:

According to frame originating point information Hi E2 is carried out compressed encoding and obtain the 3rd layer of coding C3;

Above-mentioned 1404 coded frame after C2, Hi and C1 encapsulation obtained compressing comprise:

Above-mentioned coded frame after C3, C2, Hi and C1 encapsulation are obtained compressing.

Particularly, above-mentioned ground floor encoding error E1 is carried out data compression coding, the second layer encoding error E2 that obtains comprises:

E1 is carried out time frequency analysis obtain frequency domain parameter F1p, according to frame originating point information Hi frequency domain parameter F1p is quantized to obtain frequency domain parameter coding C1r;

Carry out the local decode frequency domain parameter F1p ' that obtains decoding according to frequency domain parameter coding C1r, decoding frequency domain parameter F1p ' is carried out the synthetic error E 1 ' of decoding that obtains of time-frequency, E1 is subtracted each other with decoding error E 1 ', obtain second layer encoding error E2.

Further, obtain also comprising after the coded bit stream B1: send above-mentioned coded bit stream B1 by transmission channel, perhaps above-mentioned coded bit stream B1 is stored in the storage medium.

Initial data S in the embodiment of the invention can be the data of any source and any type, such as general data sequence, sampling signal of communication, sampling time sequence, voice signal, audio signal etc.The effect of data compression coding module is that input initial data S is carried out compressed encoding, it can adopt distinct methods to the requirement of compression performance and cost according to application system, the normal method that adopts has frequency domain and transform process method, as: dct transform, wavelet transformation (Wavelet Transfirm, WT) etc., the waveform coding method that also can have time domain to process, as: pulse code modulation (Pulse Code Modulation, PCM) and adaptive difference pulse code modulation (Adaptive Differential Pulse Code Modulation, ADPCM) etc., can also adopt the prediction of algorithm more complicated, such as linear prediction (Linear Prediction, the method such as LP), can also adopt simultaneously the higher but vector of algorithm more complicated of compression ratio, for example quantization vector quantizes (Vector Quantization, VQ) method etc.In order to reduce hard-wired cost, the data compression coding module in this encoder has adopted the fairly simple but Coding Compression Algorithm based on dct transform of better performances of algorithm.

Dct transform is a kind of time-frequency conversion method commonly used, and the signal of being everlasting is processed and image uses in the middle of processing, for the compression that data and signal are diminished.Why dct transform can be used for carrying out data compression, be because it it has good concentration of energy characteristic for general data or signal, be that data are done after the dct transform, the DCT coefficient that energy can focus on fraction gets on, and is conducive to carry out efficient data compression and quantization encoding.As shown in Figure 14B, the Time Domain Amplitude example of an initial data S of expression, being provided by Figure 14 C of its corresponding DCT coefficient, Figure 14 C is the amplitude example of DCT coefficient.Can find out, compare with the time domain initial data, the concentration of energy of signal is in the part coefficient behind the dct transform.

Coding Compression Algorithm flow process based on dct transform is provided by Figure 15 A.Idiographic flow is:

1501: the initial data S to every frame input carries out the DCT direct transform, obtains frequency domain parameter Fp, i.e. the DCT coefficient;

1502: in order to make algorithm data are had adaptivity, frequency domain parameter Fp is carried out preliminary treatment.

1503: for the quantification after making it more accurate, before quantizing according to preliminary treatment after the amplitude distribution of DCT coefficient, with the processing of classifying of DCT coefficient.The distribution of DCT coefficient generally is non-homogeneous distribution, provided the afterwards probability distribution example (probability distribution of other class coefficients is similar with it) of a class DCT coefficient of classification such as Figure 15 C, can find out, the distribution of DCT coefficient is not wait very much general distribution, should adopt non-homogeneous scalar quantization encryption algorithm to its Coding Compression Algorithm, what DCT coefficient compressed encoding module adopted here is the non-homogeneous scalar quantization algorithm of Lloyd-Max.

1504: quantize.In technical descriptioon before, the DCT coefficient quantization is that classification is processed, and concrete can be divided into two classes: a class is directly carried out the non-homogeneous scalar quantization of Lloyd-Max to the data of full dynamic range, and this processing method is relatively simple.Another kind of is carried out the non-homogeneous scalar quantization of Lloyd-Max to the data in a part of scope, namely sets first two threshold values, is respectively little coefficient thresholding and large coefficient thresholding; It then is to keep reconstruction SNR in the situation that make little coefficient in the tolerance interval as much as possible that little coefficient thresholding is established rules really, it then is to keep bit rate in the situation that make large coefficient in the tolerance interval as much as possible that large coefficient thresholding is established rules really, with the DCT coefficient zero setting of amplitude less than little coefficient thresholding, with the DCT coefficient more accurately method representation of amplitude greater than large coefficient thresholding, only the DCT coefficient between little coefficient and large coefficient thresholding just participates in the non-homogeneous scalar quantization of Lloyd-Max, thereby keeps to greatest extent reconstruction SNR on the basis of the minimum bit of transmission.Wherein, can represent with the form that index adds mantissa greater than the DCT coefficient of large coefficient thresholding, or utilize the meticulousr non-homogeneous scalar quantization table of Lloyd-Max to quantize.

Specific embodiment: preliminary treatment is namely unified displacement to the DCT coefficient, determine the DCT coefficient is normalized to 15 needed displacement figure places according to the absolute value maximum of this frame DCT coefficient, move at most 7, namely do not participate in displacement less than 128 coefficient, according to the displacement figure place full frames of data is carried out to shifting left again; Four distribution situations according to DCT coefficient absolute value maximum binary representation after the displacement, the DCT coefficient is divided into four classes, be that DCT coefficient maximum four divides a class into for 1000-1001,1010-1011 divides a class into, 1100-1101 divides a class into, 1110-1111 divides a class into, and wherein each class data is close on amplitude, can be relatively accurate during quantification.Setting little coefficient thresholding during quantification is 1024, namely all is set to 0 less than 1024 coefficient, indicates with 1 quantized interval; Setting large coefficient thresholding is 1% maximum place, and every frame keeps 2 large coefficients, and these large coefficients come quantization means with the non-homogeneous scalar quantization of the Lloyd-Max of 5 bits, indicate with 1 quantized interval; DCT coefficient between little coefficient thresholding and large coefficient thresholding quantizes with the positive and negative respectively non-homogeneous scalar quantization of Lloyd-Max in 7 intervals.For whole DCT coefficient, utilizing altogether 16 intervals is that 4 bits quantize.

1505: frequency domain parameter Fp obtains frequency domain parameter coding Cr behind quantization encoding described above, the frequency domain parameter Fp ' that obtains decoding behind the frequency domain parameter coding process local decode inverse quantization,

1506: the decoding frequency domain parameter obtains decoded data S ' by time-frequency is synthetic again through reprocessing, and initial data S and decoded data S ' are subtracted each other, and generates ground floor encoding error E1.In actual applications, the distribution of error E 1 generally is non-homogeneous distribution, is the probability distribution example of error E 1 shown in Figure 15 D.Therefore should adopt the non-uniform quantizing encryption algorithm to the Coding Compression Algorithm of error E 1, what here adopt also is the non-homogeneous scalar quantization algorithm of Lloyd-Max.

DCT coefficient after 1507:1504 quantizes carries out the entropy coding, obtains ground floor coding (ground floor compressed encoding).

Shown in Figure 15 B, for carrying out the flow process of second layer coding C2 (second layer compressed encoding).

1501A: find in the actual observation, the error dynamics scope differs larger between every frame, and only has the very large residual error of several amplitudes in every frame, and remainder residual error amplitude all fluctuates in relative its less scope.Therefore, before error E 1 is carried out quantization encoding, error is classified by bit-depth.

1502A: sorted residual error is carried out the non-homogeneous scalar quantization of Lloyd-Max by the scope of its bit-depth, make like this residual quantization in each class more accurate.

Specific embodiment: in order to remove the impact that accidental several large values are judged all the other most of residual error bit-depths of this frame in the frame, the third-largest value (after taking absolute value) of generally getting every frame residual error is judged the bit-depth of this frame residual error, namely judge the third-largest value whether less than 32, less than 64, less than 128 or greater than 128, then according to bit-depth residual error is divided into four classes.Sorted residual error is carried out the non-homogeneous scalar quantization of Lloyd-Max by the scope of its bit-depth, be that first kind quantized interval scope is 0-32, Equations of The Second Kind quantized interval scope is 0-64, and the 3rd class quantization interval range is 0-128, and the 4th class quantization interval range is greater than 128.Wherein the non-homogeneous scalar quantization of Lloyd-Max indicates with 15 quantized intervals, also have 1 quantized interval to be used for indicating 2 large values of residual error that every frame keeps, be the first value of residual error and Second Largest Value, these large values are come quantization means with the non-homogeneous scalar quantization of the Lloyd-Max of 5 bits.

1503A: the error coding Ce that the frequency domain parameter coding Cr that is produced by ground floor compressed encoding module and second layer compressed encoding module produce generally also has the stronger general characteristic that do not wait, Figure 15 E provided the example of 16 code word probability distribution among 4 bits of encoded Parameters of Time-frequency Fields coding Cr, and Figure 15 F has provided the example of 16 code word probability distribution among the 4 bits of encoded time errors coding Ce.Can find out, the probability that frequency domain parameter coding Cr and error coding Ce code word occur is unequal, this just provides foundation for adopting the entropy coding to carry out further data compression, therefore data entropy coding module and Error Entropy coding module adopt entropy coding method respectively frequency domain parameter coding Cr code word and error coding Ce code word to be carried out undistorted entropy coding separately, obtain respectively ground floor coding C1 and second layer coding C2.The entropy coding can adopt Huffman coding, arithmetic coding or other unfixed-length coding algorithm or their combinational algorithm.

Accordingly, the embodiment of the invention also provides a kind of uncompressing data, as shown in figure 16, comprising:

1601: obtain coded bit stream B1 and obtain coded frame, obtain decoding decompress(ion) rule; The coded frame decomposition is obtained ground floor coding C1 and second layer coding C2;

1602: according to decoding decompress(ion) rule the second layer C2 decoding decompress(ion) of encoding is obtained ground floor encoding error E1;

Above-mentioned decoding decompress(ion) rule can be that predefined decoding decompress(ion) rule also can be frame originating point information Hi; If decoding decompress(ion) rule is frame originating point information Hi, then above-mentioned 1601 can be: obtain coded bit stream B1 and obtain coded frame; The coded frame decomposition is obtained ground floor coding C1, frame originating point information Hi and second layer coding C2;

Particularly, according to frame originating point information Hi the second layer C2 decoding decompress(ion) of encoding is obtained ground floor encoding error E1 in 1602, comprising:

Second layer compressed encoding C2 is carried out the entropy decoding obtain error coding Ce, error coding Ce is carried out inverse quantization obtain ground floor encoding error E1.

Further, in 1601, the coded frame decomposition is also obtained the 3rd layer of coding C3; Above-mentioned 1602 obtain ground floor encoding error E1 according to frame originating point information Hi with the second layer C2 decoding decompress(ion) of encoding, and comprising:

According to frame originating point information Hi, the 3rd layer of coding C3 decompress(ion) of decoding obtained second layer encoding error E2;

To the second layer coding C2 decompress(ion) of decoding, obtain the error E 1 ' of decoding according to frame originating point information Hi, the error E of will decode 1 ' and second layer encoding error E2 addition obtain ground floor encoding error E1.

1603: the ground floor coding C1 decompress(ion) of decoding is obtained decoded data S ';

Particularly, the ground floor coding C1 decompress(ion) of decoding is obtained decoded data S ' in 1603, comprising:

Ground floor coding C1 is carried out the entropy decoding obtain frequency domain parameter coding Cr, according to frame originating point information Hi frequency domain parameter coding Cr is carried out the inverse quantization frequency domain parameter Fp ' that obtains decoding, decoding frequency domain parameter Fp ' is carried out the synthetic decoded data S ' that obtains of time-frequency.

Particularly, above-mentioned 1603 obtain ground floor encoding error E1 according to frame originating point information Hi with the second layer C2 decoding decompress(ion) of encoding, and comprising:

Second layer coding C2 is carried out the entropy decoding obtain frequency domain parameter coding C1r, according to frame originating point information Hi frequency domain parameter coding C1r is carried out the inverse quantization frequency domain parameter F1p ' that obtains decoding, decoding frequency domain parameter F1p ' is carried out the synthetic error E 1 ' of decoding that obtains of time-frequency; The error E of will decoding 1 ' and second layer encoding error E2 addition obtain ground floor encoding error E1.

1604: with decoded data S ' and ground floor encoding error E1 addition, obtain reconstruct data Sr1.

Correspondence is based on the flow process of the compressed encoding of dct transform, and the embodiment of the invention also provides the decoding decompression algorithm flow process based on dct transform to be provided by Figure 17.

1701: second layer coding C2 obtains error coding Ce through the entropy decoder module;

1702: obtain ground floor encoding error E1 through inverse quantization module again;

1703: decoding obtains frequency domain parameter coding Cr to ground floor coding C1 through entropy;

1704: frequency domain parameter coding is admitted to inverse quantization module, according to the header inverse quantization frequency domain parameter Fp ' that obtains decoding;

1705: decoding frequency domain parameter Fp ' reverts to original frequency domain parameter characteristic through reprocessing

1706: the time-frequency synthesis module carries out the DCT inverse transformation according to this decoding frequency domain parameter Fp ' and obtains decoded data S '.

1707: last decoded data S ' and ground floor encoding error E1 addition, obtain reconstruct data Sr1, finish the decoding decompression procedure.

The embodiment of the invention gives, the explanation that coded-bit distributes:

As above, have one to be coding mode in the frame originating point information, the possible coding mode of codec has two kinds, i.e. variable bit rate and fixed rate.Under variable rate mode, the code rate of encoder output bit flow is not fixed, and is that if input coding frame characteristic is complicated, just the many bits of needs carry out the coding of certain mass to it according to the characteristic changing of encoder input data; If opposite input coding frame characteristic is simple, with regard to needing many bits it is carried out the coding of certain mass.Under fixed rate mode, require the code rate of encoder output bit flow changeless, at this moment just need to carry out reasonable distribution to the limited coded-bit of every frame, be in the situation that bit number is limited, accomplish that coding quality is best.

Bit Allocation in Discrete under the fixed rate mode adopts the method for iteration, and the specific implementation step is as follows.If the sample frequency of data is Fs, the frame length of coded frame is Lf, and the code rate that codec is set is Rb, and frame rate Rf and every frame bit number Bf just can be expressed as so:

(formula 1)

(formula 2)

Under fixed rate mode, code rate is that Rb is given constant, and when the sample frequency of coded frame is after Fs and frame length Lf determine, the available bit number Bf of so every frame data coding has just determined.The purpose of Bit Allocation in Discrete is exactly the part reasonable distribution bit number that need to encode for every frame, makes coding efficiency reach best, and the bit number sum of simultaneously each several part distribution should equal every frame number available bit number Bf.

If the bit number of frame originating point information is Bh, the required bit number of ground floor coding C1 is B _C1, the required bit number of second layer coding C2 is B _C2, the purpose of Bit Allocation in Discrete is to make so:

Bf=Bh+B _C1+ B _C2(bit) (formula 3)

For the situation of undistorted codec, be B if establish the 3rd layer of required bit number of coding C3 _C3, the purpose of Bit Allocation in Discrete is to make so:

Bf=Bh+B _C1+ B _C2+ B _C3(bit) (formula 4)

Processing for the iteration of carrying out Bit Allocation in Discrete, to design respectively several quantization encoding tables to the ground floor compressed encoding module in the codec and second layer compressed encoding module in advance, the coding bit rate that different forms is corresponding different, the coded-bit of every like this frame distribute just can be finished with following iterative process:

A1: initialization: with being set on the coding schedule that needs many bits in the encoder of ground floor compressed encoding module and second layer compressed encoding module, make Bf＜Bt=Bh+B _C1+ B _C2

A2: adjust respectively or simultaneously the quantization table in the encoder of number ground floor compressed encoding module and second layer compressed encoding module, required number of coded bits is reduced, and calculate and be set on the coding schedule that needs many bits, make Bt=Bh+B _C1+ B _C2

A3: if Bf＜Bt forwards A2 to); If Bf 〉=Bt stops.

For undistorted codec, coded-bit distributes can adopt similar method:

B1: initialization: with being set on the coding schedule that needs many bits in the encoder of each layer compression coding module, make Bf＜Bt=Bh+B _C1+ B _C2+ B _C3

B2: the quantization table in the encoder of each layer compression coding module respectively or simultaneously, required number of coded bits is reduced, and calculate and be set on the coding schedule that needs fewer bit, make Bt=Bh+B _C1+ B _C2+ B _C3

B3: if Bf＜Bt forwards B2 to); If Bf 〉=Bt stops.

Among above square law device, the method and system embodiment, will carry out compressed encoding to initial data according to Hi, obtain ground floor encoding error E1 and ground floor coding C1; According to frame originating point information Hi E1 is carried out compressed encoding and obtain second layer coding C2; Then coded frame after C2, Hi and C1 encapsulation obtained compressing obtains coded bit stream B1.Realized compressed in layers, the compression that is applicable to the multimedia source signal also is applicable to the compression of general data, and this compression method is applied to transfer of data, can improve signal of communication and data transfer efficient, improves channel utilization; Be applied to the data storage, then can improve signal of communication and data storage efficient, improve the memory device utilance.One of ordinary skill in the art will appreciate that all or part of step that realizes in above-described embodiment method is to come the relevant hardware of instruction to finish by program, above-mentioned program can be stored in a kind of computer-readable recording medium, the above-mentioned storage medium of mentioning can be read-only memory, disk or CD etc.

Above data compression, decompression method, the Apparatus and system that the embodiment of the invention is provided is described in detail, used specific case herein principle of the present invention and execution mode are set forth, the explanation of above embodiment just is used for helping to understand method of the present invention and core concept thereof; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, all will change in specific embodiments and applications, to sum up, this description should not be construed as limitation of the present invention.

Claims

1. a data compression method is characterized in that, comprising:

Obtain the frame of initial data S, obtain compression encoding rules;

With the coded frame of C2, C1 encapsulation after obtaining compressing;

Described compression encoding rules comprises: frame originating point information Hi;

Described the coded frame of C2, C1 encapsulation after obtaining compressing comprised: the coded frame after C2, Hi and C1 encapsulation are obtained compressing.

2. described method according to claim 1 is characterized in that, describedly obtains ground floor encoding error E1 and ground floor coding C1 according to Hi, comprising:

Carry out the decoding inverse quantization frequency domain parameter Fp ' that obtains decoding according to frequency domain parameter coding Cr, decoding frequency domain parameter Fp ' is carried out time-frequency is synthetic to obtain decoded data S ', initial data S and decoded data S ' are subtracted each other, obtain ground floor encoding error E1;

Frequency domain parameter coding Cr is carried out the entropy coding obtain ground floor coding C1.

3. described method according to claim 2 is characterized in that, describedly initial data S is carried out time frequency analysis obtains frequency domain parameter Fp, comprising:

Every frame to initial data S carries out the discrete cosine transform direct transform, obtains frequency domain parameter Fp;

Describedly according to frame originating point information Hi frequency domain parameter Fp is quantized to obtain frequency domain parameter coding Cr, comprising:

Describedly according to frame originating point information Hi described coded frame is carried out compressed encoding and comprises:

According to frame originating point information Hi described coded frame is carried out non-homogeneous scalar quantization coding.

4. according to claim 1 to the described method of 3 any one, it is characterized in that, before E1 is carried out compressed encoding, also comprise:

5. according to claim 1 to the described method of 3 any one, it is characterized in that, describedly according to frame originating point information Hi E1 carried out compressed encoding and obtain second layer coding C2, comprising:

Ground floor encoding error E1 is carried out quantization encoding obtain error coding Ce, error coding Ce is carried out the entropy coding obtain second layer compressed encoding C2.

6. described method according to claim 1 is characterized in that, before the described coded frame after C2, Hi and C1 encapsulation are obtained compressing, also comprises:

Describedly according to frame originating point information Hi E1 is carried out compressed encoding and obtains second layer coding C2, comprising:

Described coded frame after C2, Hi and C1 encapsulation is obtained compressing comprises:

Described coded frame after C3, C2, Hi and C1 encapsulation are obtained compressing.

7. described method according to claim 6 is characterized in that, described ground floor encoding error E1 is carried out data compression coding, and the second layer encoding error E2 that obtains comprises:

Carry out the local decode frequency domain parameter F1p ' that obtains decoding according to frequency domain parameter coding C1r, decoding frequency domain parameter F1p ' is carried out the synthetic error E 1 ' of decoding that obtains of time-frequency, E1 is subtracted each other with decoding error E 1 ', obtain second layer encoding error E2; Frequency domain parameter coding C1r is carried out the entropy coding obtain second layer coding C2.

8. a uncompressing data is characterized in that, comprising:

Obtain coded bit stream B1 and obtain coded frame, obtain decoding decompress(ion) rule; Decoding decompress(ion) rule comprises: frame originating point information Hi;

9. described method according to claim 8 is characterized in that, according to frame originating point information Hi the second layer C2 decoding decompress(ion) of encoding is obtained ground floor encoding error E1, comprising:

10. according to claim 8 or 9 described methods, it is characterized in that, the ground floor coding C1 decompress(ion) of decoding obtained decoded data S ', comprising:

11. according to claim 8 or 9 described methods, it is characterized in that, describedly according to frame originating point information Hi the second layer C2 decoding decompress(ion) of encoding obtained ground floor encoding error E1, comprising:

12. described method is characterized in that according to claim 8, coded frame is decomposed also obtain the 3rd layer of coding C3; Describedly according to frame originating point information Hi the second layer C2 decoding decompress(ion) of encoding is obtained ground floor encoding error E1, comprising:

According to frame originating point information Hi, to the second layer coding C2 decompress(ion) of decoding, obtain the error E 1 ' of decoding, the error E of will decode 1 ' and second layer encoding error E2 addition obtain ground floor encoding error E1.

13. a data compression device is characterized in that, comprising:

Ground floor compressed encoding module be used for to be used for according to frame originating point information Hi the frame of initial data S being carried out compressed encoding, obtains ground floor encoding error E1 and ground floor coding C1;

Second layer compressed encoding module is used for according to frame originating point information Hi E1 being carried out compressed encoding and obtains second layer coding C2;

The frame package module, concrete coded frame after being used for C2, Hi and C1 encapsulation obtained compressing.

14. described device is characterized in that according to claim 13, described ground floor compressed encoding module comprises:

The time frequency analysis module is used for that initial data S is carried out time frequency analysis and obtains frequency domain parameter Fp;

The quantization encoding module is used for according to frame originating point information Hi frequency domain parameter Fp being quantized to obtain frequency domain parameter coding Cr,

The decoding inverse quantization module is for the frequency domain parameter Fp ' that obtains decoding according to frequency domain parameter coding Cr execution decoding inverse quantization;

The time-frequency synthesis module is used for decoding frequency domain parameter Fp ' is carried out the synthetic decoded data S ' that obtains of time-frequency;

Subtraction block is used for initial data S and decoded data S ' are subtracted each other, and obtains ground floor encoding error E1;

The entropy coding module is used for that frequency domain parameter coding Cr is carried out the entropy coding and obtains ground floor coding C1.

15. according to claim 13 or 14 described devices, it is characterized in that,

Described second layer compressed encoding module also is used for before E1 is carried out compressed encoding E1 being classified by bit-depth, and sorted residual error is carried out non-homogeneous scalar quantization by the scope of its bit-depth.

16. described device is characterized in that according to claim 13,

Described second layer compressed encoding module, concrete being used for carried out data compression coding to ground floor encoding error E1, the second layer encoding error E2 that obtains; E1 carries out data compression coding to the ground floor encoding error, obtains second layer coding C2;

Described device also comprises:

The 3rd layer compression coding module is used for according to frame originating point information Hi E2 being carried out compressed encoding and obtains the 3rd layer of coding C3;

Described frame package module, concrete being used for C3, C2, Hi and C1 by the coded frame after compressing.

17. described device is characterized in that according to claim 16, described second layer compressed encoding module comprises:

The time frequency analysis module is used for that E1 is carried out time frequency analysis and obtains frequency domain parameter F1p;

The quantization encoding module is used for according to frame originating point information Hi frequency domain parameter F1p being quantized to obtain frequency domain parameter coding C1r;

The decoding inverse quantization module is used for carrying out the local decode frequency domain parameter F1p ' that obtains decoding according to frequency domain parameter coding C1r;

The time-frequency synthesis module is used for decoding frequency domain parameter F1p ' is carried out the synthetic error E 1 ' of decoding that obtains of time-frequency;

Subtraction block is used for E1 is subtracted each other with decoding error E 1 ', obtains second layer encoding error E2;

The entropy coding module is used for that frequency domain parameter coding C1r is carried out the entropy coding and obtains second layer coding C2.

18. a data decompression compression apparatus is characterized in that, comprising:

Frame deblocking module is used for obtaining coded bit stream B1 and obtains coded frame; The coded frame decomposition is obtained ground floor coding C1, frame originating point information Hi and second layer coding C2;

Second layer decoding decompression module is used for according to frame originating point information Hi the second layer C2 decoding decompress(ion) of encoding being obtained ground floor encoding error E1;

19. described device is characterized in that according to claim 18, second layer decoding decompression module comprises:

The entropy decoder module is used for that second layer compressed encoding C2 is carried out the entropy decoding and obtains error coding Ce;

The decoding inverse quantization module is used for that error coding Ce is carried out inverse quantization and obtains ground floor encoding error E1.

20. according to claim 18 or 19 described devices, it is characterized in that, ground floor decoding decompression module comprises:

The entropy decoder module is used for that ground floor coding C1 is carried out the entropy decoding and obtains frequency domain parameter coding Cr;

Decoding contravariant quantization modules is used for according to frame originating point information Hi frequency domain parameter coding Cr being carried out the inverse quantization frequency domain parameter Fp ' that obtains decoding;

Addition module is used for decoded data S ' and ground floor encoding error E1 addition are obtained reconstruct data Sr1.

21. according to claim 18 or 19 described devices, it is characterized in that, also obtain the 3rd layer of coding C3 described coded frame is decomposed; Described device also comprises:

The 3rd decoding decompression module is used for according to frame originating point information Hi, and the 3rd layer of coding C3 decompress(ion) of decoding is obtained second layer encoding error E2;

Described second layer uncompressed encoding module, concrete be used for according to frame originating point information Hi obtains the error E 1 ' of decoding to the second layer coding C2 decompress(ion) of decoding, and the error E of will decode 1 ' and second layer encoding error E2 addition obtain ground floor encoding error E1.

22. according to claim 18 or 19 described devices, it is characterized in that, described second layer uncompressed encoding module comprises:

The entropy decoder module is used for that second layer coding C2 is carried out the entropy decoding and obtains frequency domain parameter coding C1r;

The decoding inverse quantization module is used for according to frame originating point information Hi frequency domain parameter coding C1r being carried out the inverse quantization frequency domain parameter F1p ' that obtains decoding;

Addition module is used for will decode error E 1 ' and second layer encoding error E2 addition, obtains ground floor encoding error E1.

23. communication system, comprise: but the data sending terminal and the data receiver that connect with communication mode, it is characterized in that, described data sending terminal is the device of claim 13 to 15 any one, and described receiving terminal is the device of claim 18 to 20,22 any one; Perhaps, described data sending terminal is the device of claim 16 or 17, and described receiving terminal is the device of claim 21.