CN1206470A - Compression method and apparatus for seismic data - Google Patents

Abstract

Methods and apparatus for seismic data compression are described, making use of a local discrete sine/cosine transform of the type IV applied to a data set which is limited by a stationary window function which reduces the overlap to the adjoining windows.

Description

The compression method of geological data and device

The present invention relates to the compression method and the device of geological data.

Data compression (or reduction) is not lose the Digital Signal Processing of main information in order to reduce processed data volume in processing.This mainly is to realize by the redundancy of removing in the data, and may relate to the inessential part of casting out data.The possibility of result of this compression causes some loss of data accuracy.Allow the data compression of accurate reconstruct raw data often to be called harmless data compression in the literature.The data compression that relates to certain reduction of accuracy is called the data compression that diminishes." rounding-off method " reaches " down-sampling method " (down sampling) is the common example of data compression; These two methods normally diminish.

The detection of geological data need be carried out a large amount of seismic experiments to obtain image under the reliable face of land.The response that each experiment will use suitable sound source to produce sound wave and measured the earth by a large amount of receivers.Large-scale seismological observation has so just produced the lot of data that is generally digital format, and these data need be transmitted, storage and processing.For the ease of this lot of data is handled, can adopt data compression.

Usually the data compression technique that diminishes that adopts during geological data is surveyed is the grouping forming method.This relates to maintenance and the transmission and the processing of adjacent reception devices a large amount of in the group of fixed size, rather than individual other measurement.

Compressing data is not mainly used the grouping forming method.The grouping forming method suppresses neighbourhood noise at random, and suppresses the ripple such as the low superficial velocities such as ground roll in the earthquake of land.This grouping forming method has weakened the high spatial frequency composition of data.Yet this weakens and is to carry out in coarse mode, has promptly only partly suppressed to be the ripple of slow propagation on the surface and to have changed remaining data.Thereby, there is adequate cause just to save the grouping forming method from the detection phase, and the output of each receiver of individual record.At this moment allow to adopt more exquisite method to reach coherent noise at random to reduce.Yet abandoning grouping formation rule in the detection phase has increased each stage data volume to be processed subsequently greatly.

At IEEE Int.SYM.Circuits ﹠amp; Systym, New Orleans, LA, 1-3 May1990, Vol.2, among the 1573-6, A.Spanias, people such as S.Johnson have set forth several methods based on conversion that are used for the geological data compression.These methods comprise discrete Fourier transform (DFT) (DFT), discrete cosine transform (DCT), Walsh-Hardamard conversion (WHT), and Karhunen-Loeve conversion (KLT).Yet, can be used between several different conversion at the DCT of slip frame described in this publication and that be applied to N data point and to carry out relatively.When as data compression method, the slip frame produces a large amount of redundant datas at transform domain.

Thereby the purpose of this invention is to provide a kind of method that is used to compress geological data.Another object of the present invention is that a kind of method of not using the compression geological data of grouping forming method will be provided.

The invention provides the first order compression that wherein the discrete triangle of the local space of IV type or time (that is, sine or cosine) conversion is used for seismic signals.The discrete sin/cos conversion of IV type itself is known.For example exist by H.S.Malvar: " Lapped transforms for efficienttransform/subband coding ", IEEE ASSP, vol.38, no.6 has provided general elaboration among the June 1990.The result that local space or time discrete cosine/sine transform obtain is the conversion coefficient that obtains and less correlativity compacter more than raw data.In following data processing step, preferably adopt this two character.

Adopted the compactedness of conversion coefficient in a treatment step, this step can be described to the re-quantization (requantization) or the step that rounds off.The purpose of this step is to keep selected coefficient with high precision, and keeps other coefficient with lower precision, thereby describes the required data volume of coefficient and realize further data compression so that reduce.

The correlativity of the conversion coefficient that reduces provides the chance that adopts encoding scheme, so that further reduction will be stored or data quantity transmitted.Adoptable encoding scheme itself is known for example Huffmann coding or amplitude coding.

The seismic signals that adopts this method generally is the tracking figure that obtains from a plurality of receivers, for example obtains from underground syren or hydrophone.The arrangement capable of being combined of these receivers, all these is to know in the prior art.For example, one of them is the 3 this traditional dimension land seismology layouts that are arranged in the linear array of the underground syren in several parallel lines.Use partial transformation to allow the data that are included in the some receivers on each bar line are compressed in the method.Know as institute in this specialty, partial transformation is a kind of like this conversion, and wherein conversion is applied to the window of defined tracking figure.Like this, the number that has applied the tracking figure of partial transformation in each conversion stages in succession is called spatial window, and this window may change along with the alternative types that is applied.

The partial transformation window is defined by the function of window, and window function be chosen such that make conversion be quadrature and be reversible.The selection of window function makes conversion be applied to window with adjacency overlapping center window, preferably half of overlapping these windows.

Conversion can divide two steps to carry out, and first step is a folding step, center window wherein and half window combination of adjacency and produce a folded signal, and second step is compression to cosine transform that folded signal carried out.

Except the local space conversion, preferably also apply local time's conversion to data.The combination of two kinds of partial transformation can reach ratio of compression preferably.Preferably IV type local time discrete sine/cosine transform of local time's conversion.Yet other signal transformation and decomposition also can be used, such as common local discrete cosine transform, and local Fourier transform.Can apply local space triangular transformation and local time's triangular transformation by any order.

The conversion coefficient of raw data has formed a data set after the expression conversion, and different compression methods can be applied to this data set.These compression methods can be called (again) jointly and be quantized and coding.Quantizing process generally includes scaling step and rounds off when being used for packed data.Quantizing process is in order to reduce radio-frequency component or coefficient, to keep low-frequency component with degree of precision simultaneously.

Calibration is preferably by removing with first scalar factor of expression low frequency and coming except that realizing with second scalar factor of expression high frequency.First scalar is selected less than second scalar, this be since scalar the compression that realizes of senior general is big more more.Like this, in seismic analysis the coefficient of significant especially expression low frequency will can not resemble the expression high frequency coefficient be compressed, thereby keep the former precision.

By using the uniform quantization near integer function, or statistics rounds off, or non-homogeneous re-quantization can be realized calibration.

Calibration or quantization parameter can be in time, space or the frequency of space or time and change.

In further preferred embodiment of the present invention, ratio of compression is to be determined automatically by the noise level in the seismic signal.Preferably use a part of signal or a plurality of part signal that do not comprise the signal that produces by the seismic origin to measure noise level.Like this, at " first arrive " part or the so-called noise files of the tracking figure of record before, promptly the tracking figure of the record when not having the seismic origin can be used for determining noise level.Even be more preferably, before determining the step of noise level to signal filtering, to avoid undue estimation to noise level.

Preferably select ratio of compression, promptly quantization error makes it be equal to or less than noise level.

The reduction of the precision that is caused by data compression has increased the redundance of data.Thereby according to the present invention further characteristics, the redundance of data can be used to further reduce compressed data, this is preferably by applying variable length data coding, such as amplitude coding or Huffman coding.For example, can carry out amplitude coding, make each coefficient be directly proportional with the absolute value of greatest coefficient to calibration coefficient.

Geological data compression method according to the present invention can be used for all types of geological datas, comprises land, transitional zone, 2 peacekeepings, the 3 dimension observation datas of ocean or habitata.Possible data also comprise pre-recorded data or preprocessed data, collect (shot gathers) such as impacting, and shared mid point (CMP) is collected, heap (stacks), transition segment, or single-sensor record.Also be applicable to two-dimension earthquake detection geometry, such as the land layout that comprises a line receiver, or for two dimension or the three-dimensional oceanic earthquake observation of using floating drum with hydrophone, and for the application of two vertical direction of actual area receiver layout (in the line/cross spider application).

Use the special chip group, can accomplish before field work chamber (field boxes) or probe vehicles or ship transmission in single receiver packed data.Can also be in data transmission use, for example use to the processing enter transmission or for the interstage for data in farther " downstream " of data processing.After transmission and/or stores compressed data, can use identical reverse step and with the reverse sequence reconstruct or the raw data that decompresses out.

The present invention also is to be used for carrying out the device of method described here.

The present invention is that also data are subjected to the seismological observation mode according to above-mentioned one or more method compressions.

From following detailed description and accompanying drawing, these or its its feature of the present invention, preferred embodiment and mutation thereof, possible purposes and advantage can be received and understood by this area ordinary skill.

Fig. 1 represents the typical window according to preferred window function design.

Fig. 2 represents to describe the calibration estimated or the curve of quantization error in spatial frequency domain when using the method according to this invention.

Fig. 3 represents to describe when using the method according to this invention the calibration actual in spatial frequency domain or the curve of quantization error.

In general, in 3 traditional dimension land seismology layouts, receiver is arranged in the straight line receiver array of several parallel lines.In this embodiment, data compression method applies along each receiver is capable, and the capable reason of can coverlet staying alone of each receiver.This application is called the application by row.

Traditional grouping forming method process is to each group of receivers summation, is expressed as on mathematics $\left[1\right]\stackrel{\→}{g}\left(i\right)=\frac{1}{N_g}\underset{n=0}{\overset{N_{g-1}}{\mathrm{\Σ}}}\stackrel{\→}{S}(n+i{N}_g);$ $i=O,...,I_g-1$

N wherein _gBe the number of every group of receivers, I _gBe the capable group number of every receiver, vectorial g is included in the data that form in groups in the i group, and vectorial s (n) comprises the data of measuring among the receiver n.

As seen, the sample number that grouping forms in the data is a factor 1/N of original data volume in equation [1] _g

Geological data compression method according to the present invention does not rely on this grouping forming method, and this method is the roughening method that only keeps the space low-frequency component of received signal.On the other hand, this method has avoided keeping all data from each receiver.New method has kept prior space low-frequency component with high accuracy, and has kept the spatial high-frequency composition with (but still being important) accuracy that reduces.The reduction of accuracy means the less position of every sample data needs, thereby has realized the compression of data.

The method according to this invention relates to following several stages that will illustrate:

Stage 1: spatial alternation

IV type Local Cosine Transform, i.e. cosine transform in data apply the window of finite population receiver, ${\stackrel{\→}{C}}_{\mathrm{km}}^{\[2\]}=\underset{n=m/2}{\overset{3M/2-1}{\mathrm{\Σ}}}\stackrel{\→}{s}(n+\mathrm{nM})h\left(n\right)\sqrt{\frac{2}{M}}\cos \left(\frac{\mathrm{\π}}{M}(k+\frac{1}{2})(n+\frac{1}{2})\right)$

For k=0 ..., M-1, and for m=0 ..., P-1.

In equation [2], vectorial C _KmBe local space DCT-IV coefficient, vectorial S (n) is data measured among the receiver n, and M is the receiver number in every window in the Local Cosine Transform, and h (n) is a window function, and P is the capable window number of every receiver.

In this concrete equation, suppose that M is an even number, suitable modification M is made in conversion may be selected to be odd number certainly.If window function h (n) meets the following conditions, be quadrature and reversible 0≤h (n)≤1h (n)=0 then with up conversion $\mathrm{for\; n}\&le;-\frac{M}{2}\mathrm{and\; n}\&GreaterEqual;\frac{3M}{2},$ [3]h(n)=h(M-n-1),h(n)2+h(n+M)2=1 $\mathrm{for}-\frac{M}{2}\&le;n<\frac{M}{2}\&CenterDot;$

Conversion in the formula [2] is called DCT-TV (the 4th type discrete cosine transform), and has and the analogous effective realization of fast Fourier transform (FFT).Conversion in the formula [2] will be called local DCT-IV.For local DCT-IV, complexity of calculation is directly proportional with window number purpose product, and carries out the required workload of DCT-IV and be directly proportional with (N/M) * M log (M) ∝ N, and wherein N is a signal length, the receiver number that promptly every receiver is capable, and M is a length of window.This and FFT are more favourable, and the latter carries out the whole length of signal, will need the multi-pass operations that is directly proportional with N log (N).Main difference between this conversion and the common local DCT is that it can be used in overlapping window.Common local DCT is limited in disjoint rectangular window.

As seen, each organizes coefficient { C in equation [2] _KmK=0 ..., M-1 } calculating need the contribution data of 2M receiver, M wherein is from the window own, M/2 is from the window of each adjacency.But the number of whole conversion coefficients equals primary data sample number, i.e. p*M=N.The beginning of data and end can or be used beginning separately and finish the window function processing by assumption period.

Can the while calculation of transform coefficients in a plurality of windows.In addition, there is a kind of effective implementation method of carrying out the conversion in the equation [2] in two steps suddenly:

1. determine the folded signal vector f among each window m _mFolding step:

For 0≤n＜M/2; And $\&lsqb;4^{\&prime;}\&rsqb;\stackrel{\&RightArrow;}{s}(n+\mathrm{mM})h\left(n\right)-\stackrel{\&RightArrow;}{s}(2M-n-1+\mathrm{mM})h(2M-n-1)$

For M/2≤n＜M.

2. folded signal f _mCosine transform be such $\&lsqb;5\&rsqb;{\stackrel{\&RightArrow;}{c}}_{\mathrm{km}}=\underset{n=0}{\overset{M-1}{\mathrm{\&Sigma;}}}{\stackrel{\&RightArrow;}{f}}_m\left(n\right)\sqrt{\frac{2}{M}}\cos \left(\frac{\mathrm{\&pi;}}{M}(k+\frac{1}{2})(n+\frac{1}{2})\right)$

For k=0 ..., M-1, and m=0 ..., P-1.

At IEEE ASSP, vol.38, No.6., June 1990, among the Lapped transform forefficient tranform/subband coding, H.S.Maler has discussed the character and the implementation method of local DST-IV (the local sine transform of the 4th type), and this conversion equates with local DCT-IV except cosine is replaced by sine.

Stage 2: time change

Local DCT-IV coefficient remains the function of writing time.In the method that is proposed, also apply the DCT-IV of local time to them.The length of window and window function are to select independently of one another in the length of employed window from local space DCT-IV and the window function.

Though other ordinate refers to the time no longer simply, vectorial C _KmRepresentation constant.

Stage 3: re-quantization

The convenience of quadrature reversible transformation is, they are to keep energy, and promptly they satisfy the Parseval theorem.This means, in the transform domain square the quantization error square error that equals to try to achieve in the original domain.For the square error in the original domain is so equally.Square error for all data energy also is like this.Yet this retention properties does not exist for the maximum absolute amplitude of data.

Quantize to relate to the scope of the data amplitudes of division such as coefficient etc., and according to employed concrete quantization method by the amplitude that rounds off so that specify another amplitude to reduce the data volume of appearance.

Prevailing quantization method is the uniform quantization method: amplitude range is divided into equal differential, and the alignment amplitude that rounds off.The result who does to obtain like this is point of fixity (integer) expression of data sample.If (such as △) is enough little for differential size, then quantization error is with variance or energy △ ²/ 12 equally distributed white noises.If use the uniform quantization method in transform domain, then the quantization error in the original domain also will show as the white noise of constant energy.Quantization error not necessarily evenly distributes in the original domain.If use the non-uniform quantizing device, then these sayings are false.It is contemplated that the quantizer that has the decline precision for the amplitude that increases, such as the quantizer that uses in the floating point representation.

If reach big ratio of compression, then quantizer becomes coarse for some part of (in transform domain) data.The stochastic analysis of this time error begins to lose efficacy, and the filtering to the data in the transform domain occurs.If yet employed conversion provides the good compression of geological data content, all as if use DCT-IV such, can avoid the important contents of data to be subjected to the effect of this filtering.

The coefficient of local space DCT-IV is represented the local space frequency content of geological data, i.e. coefficient vector C in the m window _Km, K=0 ..., M-1 representation space frequency content.Index k has determined the spatial frequency of being considered.Low k is corresponding to low local space frequency, and high k is corresponding to high local space frequency.

This method relates to the re-quantization that separates of low and high local space frequency content.This can pass through C _Km' rule round off (or uniform quantization) carry out so that provide $\&lsqb;6\&rsqb;{\stackrel{\&RightArrow;}{c}}_{\mathrm{km}}=\mathrm{NINT}({\stackrel{\&RightArrow;}{c}}_{\mathrm{km}}/{\mathrm{\&delta;}}_L){\mathrm{\&delta;}}_L$

For 0≤k≤k _m-1, and $\&lsqb;7\&rsqb;{\stackrel{\&RightArrow;}{c}}_{\mathrm{km}}=\mathrm{NINT}({\stackrel{\&RightArrow;}{c}}_{\mathrm{km}}/{\mathrm{\&delta;}}_H){\mathrm{\&delta;}}_H$

For k _m≤ k≤M-1,

Wherein equation [6] is used for low local space frequency, and equation [7] is used for high local space frequency.In these two equations, NINT is nearest integer function, k _mBe the number of high-precision local DCT-IV coefficient, and δ _L, δ _HBe to be used for scalar that spatial frequency is rounded off.In above expression formula, at δ _L, δ _HQuantization error becomes bigger when becoming big.In order to keep low local space frequency on precision, to be higher than high local space frequency content, select these scalars to make δ _L＜δ _HScalar or quantization error are big more, and the figure place of the every sample that then needs is few more, thereby can reach bigger compression.

Thereby the value of regulating these scalars provides a kind of method of automatic selection ratio of compression.At first, a method of adjusting scalar is an estimated value will determining noise in the geological data that is recorded.This can not have the tracer signal of seismic signal partly to realize by more known, i.e. recorded data part before arriving for the first time more preferably, or the data that write down during so-called noise impact.From these " no signal " data, can derive an estimated value of noise by common statistical method.If provide this estimated value, just can be by the scalar of compression being regulated with its predetermined relation.And the coefficient that precision reduces is many more (to be k _mMore little), then the compression that can reach is big more.For example, if the data signal levels of given-10dB and-noise level (using DATA REASONING) of 50dB from the data division of no signal, then these scalars can be set, so that reach 18: 1 ratio of compression at least.

In the part and method windowing, needed separation can not be perfect between low (traditional) and high (interpolation) wave number band.The leakage in a small amount of the quantizing noise of wave number band from high to low is inevitable.In order to make the noise leakage minimum, carefully select window design for quantizing noise in the local cosine coefficient.Also to reach satisfied low-level leakage by sacrificing some compression performance.In the following stage 0 The Window Design is discussed.

Stage 4: amplitude coding

For further packed data, can adopt the reduction precision, thereby reduce the figure place of every sample, this has increased the redundance in the data.This is by to conversion coefficient (C _Km') amplitude coding utilize.For this purpose, form the computing of a small amount of (being generally 8) coefficient.In each computing, maximum absolute value has determined the figure place that coefficient will use.In the code, must be placed on each computing of coefficient the back of employed figure place, perhaps more particularly, be placed on the back of the code that the Huffman coding of the required figure place of each computing obtains.

Stage 0: parameter setting and window design

This stage need be selected window function and re-quantization parameter.

The low spatial frequency content of this method retention data reduces the precision of high spatial frequency content simultaneously.As noted earlier, this is to accomplish by the precision of the local DCT-IV coefficient of the local space frequency content that changes the expression geological data.Use by row above, can predict that in theory re-quantization local space frequency (being local space DCT-IV coefficient) is to the action effect along the spatial frequency content of the capable data of whole receiver.

In spatial frequency band, be considered to low and be considered to make difference between high.By the capable observation of whole receiver the time, the precision of the low spatial frequency content of data reduces the precision of high local space frequency, and this is inevitable.The loss of the precision that causes at low spatial frequency is by window function h (also relating to its length 2M), high precision k _mLocal DCT-IV coefficient number and by δ _LAnd δ _HThe selected precision of decision is determined.

This method relates to the window design process of being made up of following steps:

-regulation is along the capable low spatial frequency band of receiver;

-preferably with data in the predetermined estimated value of noise the threshold value that maximum in the low spatial frequency band can be accepted loss of significance is set relatively;

-selection M, k _m, δ _LAnd δ _H

-the window function h (from equation [3]) of all permissions is made the loss of significance minimum of low spatial frequency band;

If the loss of significance of-gained is lower than threshold value, then stop this process, otherwise by high precision (k _m) increase the number of local DCT-IV coefficient and repeat the step of front.

The amount that influences the leakage rate of quantization noise level (△) is: local by index (k _m), be used for DCT-IV length M, and be at last length of window (≤2M) and shape.To leak minimum in order making, to require △ as much as possible little, and k _mBig as much as possible with M.Yet, require the big as much as possible and k of △ in order to make compression maximum _mAs much as possible little.M and △ fix in window design.Use is to k _mInitial selected, h is reduced to minimum to noise leakage at window.If noise surpasses leak threshold, then increase by index k _mAnd calculate new window.

Propose to use the following example of real data on a small scale now:

-128 receivers, thereby the spatial frequency number is 64;

-spatial window length M=16, thereby spatial window function length 2M=32;

-spatial window is counted P=8;

-at every receiver of 4ms time 1024 samples are arranged;

-time window length is 64 samples;

-24 fixed point sample values.

Select the low spatial frequency band that it is made up of 4 lowest spatial frequency (8 real-valued fourier coefficients).This means, because window number is 8, so every window high precision (is k _m) local DCT-IV coefficient (that is K, _m) number be at least 1.

The threshold value that low spatial frequency band quantization error is set is-115dB.Low quantization error is set is-68dB for-119dB and high quantization error.During using this example preliminary experiment, if high precision (k _m) number of local DCT-IV coefficient is increased to 4 from 1, then error is only fallen below the threshold value.This means 25% (4 of per 16 windows) that kept coefficient with high precision.Showed designed window among Fig. 1.The quantization error of prediction together is shown among Fig. 2 with representing the rectangular curve of cutting apart between low and the high spatial frequency.

Shown in Figure 3 from the quantization error that real data obtained.Between prediction and actual quantization error, can be observed good coincideing.

Raw data is 24 point of fixity.The figure place that is used for obtaining the every local DCT-IV coefficient of required precision is 26 for low DCT-IV coefficient (25%), is 17 for high DCT-IV coefficient (75%).This is the initial compression of 19.25 in an every sample of average out to.Yet the redundance of increase can make the amplitude coding in the stage 4 that this is reduced to 4.4 in every sample.

In another embodiment, fundamental purpose is not to make data compression method inconsistent with the data that obtain with classic method.This means that a large amount of branch group of receivers will keep with high precision, and the accuracy representing of remaining geological data to reduce.The embodiment of this embodiment is simpler than first embodiment.This method that relates to FFT and linear array is as follows:

Stage 1: spatial alternation $\&lsqb;8\&rsqb;{\stackrel{\&RightArrow;}{f}}_{\mathrm{km}}=\underset{n=0}{\overset{M-1}{\mathrm{\&Sigma;}}}{\stackrel{\&RightArrow;}{s}}_m\left(n\right)\mathrm{e\; exp}\left(\frac{-2\mathrm{\&pi;ink}}{M}\right)$

For k=0 ..., M-1, and to all m.

Window size (M) equals grouping size (N simply _g).Here grouping number is as data vector S _m(n) subscript m in provides representing that all groupings are separated to handle, and they can be at angle capable with receiver.

Stage 2: time change

For time change, as local DCT-IV service time among first embodiment (3.1 joint).

Stage 3: re-quantization

The equation [6] and [7] that are similar among first embodiment are carried out re-quantization like that, for example $\&lsqb;9\&rsqb;{\stackrel{\&RightArrow;}{f}}_{\mathrm{km}}=\mathrm{NINT}({\stackrel{\&RightArrow;}{f}}_{\mathrm{Om}}/{\mathrm{\&delta;}}_L){\mathrm{\&delta;}}_L$ (with) and $\&lsqb;10\&rsqb;{\stackrel{\&RightArrow;}{f}}_{\mathrm{km}}=\mathrm{NINT}({\stackrel{\&RightArrow;}{f}}_{\mathrm{km}}/{\mathrm{\&delta;}}_H){\mathrm{\&delta;}}_H$

For 1≤k≤M-1 (high local space frequency).

Stage 4: amplitude coding

As carrying out the amplitude coding described in first embodiment.

Owing among this embodiment, only keep a large amount of receiver groupings, do not limit the capable seismic survey geometry of receiver that uses linear array with high precision.For example can be applied to the detection geometry of area array.

Other embodiment still has following characteristics as embodiment discussed above

A) stage 1 and 2 exchanges;

B) any other method of the coefficient in re-quantization equation [6] and [7] in the stage 3.For example, statistics rounds off or non-homogeneous re-quantization;

C) in equation [6] and [7] the re-quantization parameter in time, position or frequency and change;

D) strange length of window M;

E) length of window is constant in local space or time D CT-IV;

F) DST-IV replaces DCT-IV;

G) DCT-IV of local time in the stage 1 is replaced by any other signal transformation or decomposition, replaces such as decomposing with (part) DCT, local DST, (part/short time) FFT, wavelet transform or sub-band;

H) local space DCT-IV is replaced by any other signal transformation or decomposition, such as using (part) DCT, local DST, (part/short time) FFT, wavelet transform or sub-band decomposition;

I) coding of the amplitude in the stage 4 is replaced by any other method of utilizing data redudancy.

Claims (9)

1. based on the geological data compression method of discrete trigonometric transforms, described method is characterised in that following steps:

-select window function, make conversion be applied to the center window and have overlapping with the adjacency window;

-to described data apply the IV type (DCT-IV, DST-IV) local space and/or time discrete triangular transformation are so that produce data in transform domain; And

Data in the described transform domain of-compression.

According to the process of claim 1 wherein by the precision that reduces by keeping its selected part the data in the compressed transform territory.

3. according to the method for claim 2, wherein keep the HFS of data in the transform domain by the precision that reduces.

4. according to the process of claim 1 wherein the data in the compressed transform territory by data in the transform domain are carried out more coarse re-quantization.

5. according to the data compression method of claim 1, wherein determine the noise estimation value in the unpacked data, and according to the automatic ratio of compression of selecting for this compression of described estimated value.

6. according to the data compression method of above any claim, wherein the redundance of the reduction of the data in the transform domain is used for further reducing compressed data.

7. according to the data compression method of claim 6, wherein by the data in the variable length code compressed transform territory.

8. the method for claim 1,4 or 6, further comprising the steps of:

The data of-transmission and/or store compressed; And

The contract data of the described compression of reciprocal transformation of-decompress(ion).

9. be used to compress the device of geological data, described device comprises

-generation window function makes conversion be applied to the center window and with the adjacency window overlapping device is arranged;

-be used for to described data apply the IV type (DCT-IV, DST-IV) local space and/or time discrete triangular transformation are so that produce the device of data in transform domain; And

-be used for compressing the device of the data of described transform domain.

Images