PatentDe  


Dokumentenidentifikation EP0967718 03.01.2008
EP-Veröffentlichungsnummer 0000967718
Titel Digitale Filter und entsprechende Filterverfahren
Anmelder Canon K.K., Tokio/Tokyo, JP
Erfinder Majani, Eric, 35000 Rennes, FR
Vertreter TBK-Patent, 80336 München
DE-Aktenzeichen 69937589
Vertragsstaaten DE, FR, GB
Sprache des Dokument EN
EP-Anmeldetag 10.06.1999
EP-Aktenzeichen 994014124
EP-Offenlegungsdatum 29.12.1999
EP date of grant 21.11.2007
Veröffentlichungstag im Patentblatt 03.01.2008
IPC-Hauptklasse H03H 17/02(2006.01)A, F, I, 20051017, B, H, EP

Beschreibung[en]

The present invention concerns in general terms the transformation of digital signals, and concerns more particularly symmetrical biorthogonal filters with a compact support and satisfying the requirements for perfect reconstruction.

Such filters are already known. The invention aims to provide novel filters whose coding efficiency, or performance, that is to say the output/distortion ratio, is greater than that of known filters.

To this end, the invention proposes a digital filter, and a corresponding filtering method, adapted to transform an input digital signal into two output digital signals,

  • the input digital signal including input samples,
  • the first output digital signal including output samples of odd ranking, each output sample of odd ranking being equal to the input sample of the same ranking decreased by the approximate value of a sum of input samples of even ranking respectively multiplied by first filtering coefficients,
  • the second output digital signal including output samples of even ranking, each output sample of even ranking being equal to the input sample of the same ranking increased by the approximate value of a sum of output samples of odd ranking respectively multiplied by second filtering coefficients,
characterised in that
  • the second filtering coefficients are one set chosen according to the length of the filter from all the following sets of coefficients: [5/16, 5/16], [-1/16, 5/16. 5/16. -1/16] and [1/256, -15/256, 78/256, 78/256, -15/256, 1/256].

These filtering coefficients make it possible to construct a family of analysis filters whose performance is better than that of known filters.

According to a preferred characteristic, the first filtering coefficients are one set chosen according to the length of the filter from all the following sets of coefficients: [S, S], [-1/16, 9/16, 9/16 -1/16], [3/256, -25/256, 150/256, 150/256, -25/256, 3/256], [-5/2048, 49/2048, -245/2048, 1225/2048, 1225/2048, -245/2048, 49/2048, -5/2048].

The filters thus constructed are analysis filters of odd length.

The invention also proposes a digital filter, and a corresponding filtering method, adapted to transform two input digital signals into one output digital signal,

  • a first input signal including input samples of odd ranking and a second input signal including input signals of even ranking,
  • the output digital signal including output samples of even ranking and odd ranking,
  • each output sample of even ranking being equal to the input sample of the same ranking decreased by the approximate value of a sum of input samples of odd ranking respectively multiplied by third filtering coefficients,
  • each output sample of odd ranking being equal to the input sample of the same ranking increased by the approximate value of a sum of output samples of even ranking respectively multiplied by fourth filtering coefficients,
characterised in that
  • the third filtering coefficients are one set chosen according to the length of the filter from all the following sets of coefficients: [5/16, 5/16], [-1/16, 5/16, 5/16, -1/16] and [1/256, -15/256, 78/256, 78/256, -15/256, 1/256].

These filtering coefficients make it possible to construct a family of synthesis filters, corresponding to the above analysis filters, and whose performance is better than that of known synthesis filters.

According to a preferred characteristic, the fourth filtering coefficients are one set chosen according to the length of the filter from all the following sets of coefficients: [S, S], [-1/16, 9/16, 9/16-1/16], [3/256, -25/256, 150/256, 150/256, -25/256, 3/256]. [-5/2048, 49/2048, -245/2048, 1225/2048, 1225/2048, -245/2048, 49/2048, -5/2048].

The filters thus constructed are synthesis filters of odd length.

The invention also proposes a digital filter, and a corresponding filtering method, adapted to transform an input digital signal into two output digital signals,

  • the input digital signal including input samples,
  • an intermediate digital signal including intermediate samples of odd ranking, each intermediate sample of odd ranking being equal to the input sample of the same ranking decreased by the approximate value of a sum of input samples of even ranking respectively multiplied by fifth filtering coefficients,
  • the first output digital signal including output samples of even ranking, each output sample of even ranking being equal to the input sample of the same ranking increased by the approximate value of a sum of intermediate samples of odd ranking respectively multiplied by sixth filtering coefficients,
  • the second output digital signal including output samples of odd ranking, each output sample of odd ranking being equal to the intermediate sample of the same ranking decreased by the approximate value of a sum of intermediate samples of even ranking respectively multiplied by seventh filtering coefficients,
characterised in that
  • the seventh filtering coefficients are one set chosen according to the length of the filter from all the following sets of coefficients: [5/16, 0, -5/16], [-1/16, 6/16, 0, -6/16, 1/16], and [1/256, -16/256, 93/256, 0, -93/256, 16/256, -1/256].

These filtering coefficients make it possible to construct a family of analysis filters whose performance is better than that of known filters.

According to a preferred characteristic, the sixth filtering coefficients are one set chosen according to the length of the filter from all the following sets of coefficients: [1/16, S, -1/16], [-3/256, 22/256, S, -22/256, 3/256], [5/2048, -44/2048, 201/2048, S, -201/2048, 44/2048, -5/2048].

The filters thus constructed are analysis filters of even length.

According to another preferred characteristic, there is a single fifth filtering coefficient and it is equal to one.

Implementation is thus simplified.

The invention also concerns a digital filter, and a corresponding filtering method, adapted to transform two input digital signals into one output digital signal,

  • a first input signal including input samples of odd ranking and a second input signal including input signals of even ranking,
  • the output digital signal including output samples of even ranking and odd ranking,
  • an intermediate digital signal including intermediate samples of odd ranking, each intermediate sample of odd ranking being equal to the input sample of the same ranking increased by the approximate value of a sum of input samples of even ranking respectively multiplied by eighth filtering coefficients,
  • each output sample of even ranking being equal to the input sample of the same ranking decreased by the approximate value of a sum of intermediate samples of odd ranking respectively multiplied by ninth filtering coefficients,
  • each output sample of odd ranking being equal to the intermediate sample of the same ranking increased by the approximate value of a sum of output samples of even ranking respectively multiplied by tenth filtering coefficients,
characterised in that
  • the eighth filtering coefficients are one set chosen according to the length of the filter from all the following sets of coefficients: [5/16, 0, -5/16], [-1/16, 6/16, 0, -6/16, 1/16], and [1/256, -16/256, 93/256, 0, -93/256, 16/256, -1/256].

These filtering coefficients make it possible to construct a family of synthesis filters, corresponding to the above analysis filters, and whose performance is better than that of known filters.

According to a preferred characteristic, the ninth filtering coefficients are one set chosen according to the length of the filter from all the following sets of coefficients: [1/16, S, -1/16], [-3/256, 22/256, S, -22/256, 3/256], [5/2048, -44/2048, 201/2048, S -201/2048, 44/2048, -5/2048].

The filters thus constructed are synthesis filters of even length.

According to another preferred characteristic there is a single tenth filtering coefficient and it is equal to one. Implementation is thus simplified.

According to preferred and alternative characteristics, the approximate value is the identity function, or the approximate value is a function of a real variable which supplies the integer which is closest to the variable, or the approximate value is a function of a real variable which supplies the integer below the variable, or the approximate value is a function of a real variable which supplies the integer higher than the variable, or again the approximate value is a function of a real variable broken down into sub-variables, the sum of which is equal to the variable, which supplies a sum of approximate values of the sub-variables, each of the approximate values of the sub-variables being as previously defined.

The invention also concerns a signal processing device including the filter as previously defined, or the means of implementing the corresponding method.

The invention also concerns a signal processing device having at least two filters as previously defined, the output signal of one of the filters being the input signal of the other filter.

The invention also concerns a digital apparatus including the signal processing device.

The advantages of the signal processing device and of the digital apparatus are identical to those disclosed previously.

An information storage means which can be read by a computer or microprocessor, whether or not integrated into the device, possibly removable, stores a program implementing the filtering method.

The characteristics and advantages of the present invention will emerge more clearly from a reading of a preferred embodiment illustrated by the accompanying drawings, in which:

  • Figure 1 depicts schematically a data processing device comprising the invention,
  • Figure 2 depicts schematically another data processing device comprising the invention,
  • Figure 3 depicts an embodiment of the data processing device comprising the invention,
  • Figure 4 depicts an embodiment of a transformation circuit included in the data processing device of Figure 1,
  • Figure 5 depicts an embodiment of an elementary transformation unit comprising the present invention,
  • Figure 6 depicts an embodiment of an elementary transformation unit comprising the present invention,
  • Figure 7 depicts sets of filters used in the elementary transformation unit of Figures 5 and 6,
  • Figure 8 depicts sets of filters used in the elementary transformation units of Figures 5 and 6,
  • Figure 9 depicts preferred combinations of sets of filters used in the elementary transformation units of Figures 5 and 6,
  • Figure 10 depicts an embodiment of an elementary transformation unit comprising the present invention,
  • Figure 11 depicts an embodiment of an elementary transformation unit comprising the present invention,
  • Figure 12 depicts sets of filters used in the elementary transformation unit of Figures 10 and 11,
  • Figure 13 depicts sets of filters used in the elementary transformation unit of Figures 10 and 11,
  • Figure 14 depicts preferred combinations of sets of filters used in the elementary transformation units of Figures 10 and 11.

According to a chosen embodiment depicted in Figure 1, a data processing device according to the invention is a data coding device 2 which has an input 24 to which a source 1 of non-coded data is connected. The data processing device can be integrated into a digital apparatus, such as a computer, a digital photographic apparatus or a facsimile machine, for example.

The source 1 has for example a memory means, such as a random access memory, a hard disk, a diskette or a compact disc, for storing non-coded data, this memory means being associated with a suitable reading means for reading the data therein. A means for recording the data in the memory means can also be provided. The source 1 can also be integrated into the digital apparatus.

It will more particularly be considered hereinafter that the data to be coded are a series of digital samples representing an image IM.

The source 1 supplies a digital image signal SI at the input of the coding circuit 2. The image signal SI is a series of digital words, for example bytes. Each byte value represents a pixel of the image IM, here with 256 grey levels, or black and white image. The image can be a multispectral image, for example a colour image having components in three frequency bands, of the red-green-blue or luminance and chrominance type. Each band is then processed in a similar fashion to the monospectral image.

Means 3 using coded data are connected at the output 25 of the coding device 2.

The user means 3 include for example means of storing coded data, and/or means of transmitting coded data.

The coding device 2 conventionally has, starting from the input 24, a transformation circuit 21, relating particularly to the present invention, and several example embodiments of which will be detailed hereinafter. The transformations envisaged here are breakdowns into frequency sub-bands of the data signal so as to effect an analysis of the signal.

The transformation circuit 21 is connected to a quantization circuit 22. The quantization circuit uses a quantization known per se, for example a scalar quantization, or a vector quantization, of the coefficients, or groups of coefficients, of the frequency sub-bands supplied by the circuit 21.

The circuit 22 is connected to an entropic coding circuit 23, which effects an entropic coding, for example a Huffman coding, or an arithmetic coding, of the data quantized by the circuit 22.

Figure 2 depicts another data processing device according to the invention, in the form of a device 5 for decoding data coded by the device 2.

Means 4 using coded data are connected at the input 54 of the decoding device 5. The means 4 include for example coded data memory means, and/or means of receiving the coded data which are adapted to receive the coded data transmitted by the transmission means 3.

Means 6 using decoded data are connected at the output 55 of the decoding device 5. The user means 6 are for example image display means, or sound restoration means, according to the nature of the data processed.

The decoding device 5 overall performs operations which are the reverse of those of the coding device 2. The device 5 has an entropic decoding circuit 51, which effects an entropic decoding corresponding to the coding of the circuit 23. The circuit 51 is connected to a dequantization circuit 52, corresponding to the quantization circuit 22. The circuit 52 is connected to a reverse transformation circuit 53, corresponding to the transformation circuit 21. The reverse transformation circuit 53 relates more particularly to the present invention. Several example embodiments will be detailed below. The transformations envisaged here effect a synthesis of the digital signal, from frequency sub-bands.

The coding device and the decoding device can be integrated into the same digital apparatus, for example a digital camera. In this case, the data processing device effects the coding and decoding of the data.

With reference to Figure 3, an example of a device 10 implementing the invention is described. This device is adapted to transform a digital signal and, according to the examples developed below, to analyse it, or to synthesise it, or to analyse and then synthesise it.

The device 10 is here a microcomputer having a communication bus 101 to which there are connected:

  • a central unit 100,
  • a read only memory 102,
  • a random access memory 103,
  • a screen 104,
  • a keyboard 114,
  • a hard disc 108,
  • a diskette drive 109 adapted to receive a diskette 110,
  • an interface 112 for communicating with a communication network 113,
  • an input/output card 106 connected to a microphone 111.

The hard disc 108 stores the programs implementing the invention, as well as the data to be coded and the data coded according to the invention. These programs can also be read on the diskette 110, or received via the communication network 113, or stored in read only memory 102.

More generally, the programs according to the present invention are stored in a storage means. This storage means can be read by a computer or microprocessor. This storage means is integrated or not into the device, and may be removable. For example, it may include a magnetic tape, a diskette or a CD-ROM (fixed-memory compact disc).

When the device is powered up, the programs according to the present invention are transferred into the random access memory 103, which then contains the executable code of the invention and the variables necessary for implementing the invention.

The device 10 can receive data to be coded from a peripheral device 107, such as a digital photographic apparatus, or a scanner, or any other means of acquiring or storing data.

The device 10 can also receive data to be coded from a distant device, via the communication network 113, and transmit coded data to a distant device, still via the communication network 113.

The device 10 can also receive data to be coded from the microphone 111. These data are then a sound signal.

The screen 104 enables a user notably to display the data to be coded, and serves, with the keyboard 114, as a user interface.

With reference to Figure 4, the transformation circuit 21, or analysis circuit, is a two-level dyadic breakdown circuit. The circuit 21 is, in this embodiment, a conventional set of filters, respectively associated with decimators by two, which filter the image signal in two directions, respectively vertical and horizontal, into sub-bands signals of high and low spatial frequencies. The relationship between a high-pass filter and a low-pass filter is determined by the conditions for the perfect reconstruction of the signal. Different examples of filters will be envisaged hereinafter. It should be noted that the vertical and horizontal breakdown filters are not necessarily identical, although in practice this is generally the case. The circuit 21 here has two successive analysis units B1 and B2 for breaking down the image IM into sub-band signals according to two resolution levels.

In general terms, the resolution of a signal is the number of samples per unit length used in order to represent this signal. In the case of an image signal, the resolution of a sub-band signal is related to the number of samples per unit length used for representing this sub-band signal horizontally and vertically. The resolution depends on the number of decimations performed, on the decimation factor and on the resolution of the initial image.

The first analysis unit B1 receives the digital image signal SI and applies it to two digital filters, respectively low-pass and high-pass 210 and 220, which filter the image signal in a first direction, for example horizontal in the case of an image signal. After passing through decimators by two D210 and D220, the resulting filtered signals are respectively applied to two low-pass filters 230 and 250, and high-pass filters 240 and 260, which filter them in a second direction, for example vertical in the case of an image signal. Each resulting filtered signal passes through a respective decimator by two D230, D240, D250 and D260. The first unit delivers as an output four sub-band signals LL1, LH1, HL1 and HH1 with the highest resolution RES1 in the breakdown.

The sub-band signal LL1 includes the image signal components, or coefficients, of low frequency, in both directions. The sub-band signal LH1 includes the image signal components of low frequency in a first direction and high frequency in a second direction. The sub-band signal HL1 includes the components of high frequency in the first direction and the components of low frequency in the second direction. Finally, the sub-band signal HH1 includes the components of high frequency in both directions.

Each sub-band signal is a set of real coefficients constructed from the original image, which contains information corresponding to a respectively vertical, horizontal and diagonal orientation of the image contours, in a given frequency band. Each sub-band signal can be assimilated to an image.

The sub-band signal LL1 is analysed by the analysis unit B2 similar to the previous one in order to supply four sub-band signals LL2, LH2, HL2 and HH2 of resolution level RES2.

Each of the sub-band signals of resolution RES2 corresponds also to an orientation in the image.

To a given analysis circuit 21 there corresponds, in a conventional fashion, a synthesis circuit whose structure is derived from that of the analysis circuit.

All the filters considered here are biorthogonal and symmetrical, with a compact support, and satisfy the requirements for perfect reconstruction.

Hereinafter, an elementary analysis unit BE having a high-pass filter, a low-pass filter and decimators, will be dealt with in particular, as well as the corresponding elementary synthesis unit, having a high-pass filter, a low-pass filter and interpolators.

Naturally, elementary analysis units BE, or respectively elementary synthesis units, can be used in combination, so that the output of one of the units is connected to the input of another unit. Thus any analysis or respectively synthesis circuit can be formed.

Figure 5 depicts a first embodiment of an elementary transformation unit effecting here an analysis filtering of a digital signal. This embodiment is equivalent to that of the unit BE, as disclosed in the article " The Lifting Scheme: A construction of second generation wavelets", by Wim Sweldens, Siam J. Math. Anal., Vol. 29, No.2, pp 511-546, 1997 .

This unit has an input E1 to which the signal to be transformed is applied. The signal to be transformed includes a series of samples {xi}, where i is a sample ranking index.

The input E1 is connected to a first decimator by two D1 which delivers the samples {x2i} of even ranking.

The input E1 is also connected to an advance AV2 followed by a second decimator D2, which delivers the samples {X2i+1} of odd ranking.

The first decimator D1 is connected to a first filter P1 which filters the samples of even ranking and supplies filtered samples to a first approximation unit A1.

The filter P1 and the approximation unit A1 will be detailed below.

The output of the approximation unit A1 and that of the second decimator D2 are connected to a first operator OP1, which performs a subtraction. The operator OP1 is connected to a first output S1 of the elementary unit, which delivers a digital signal including high-frequency samples y2i+1 calculated according to the formula y 2 i + 1 = z 2 i + 1 - APP &Sgr; m k = 1 - m p k x 2 i + k

where APP is a function produced by the approximation unit A1 giving an approximate value of a real number. The approximate value can be the real number itself, the APP function then being identity. The approximate value can be the integer below the real number, the integer above the real number or the closest integer to the real number.

Where the units P1 and A1 are combined, the sum of samples multiplied by filtering coefficients can be broken down into partial sums, the approximate value then being a sum of approximate values, as defined above, of partial sums.

In formula (1), the parameter pk is a first filtering coefficient, that is to say the current coefficient of the filter P1, the parameter k is a filtering coefficient index and the parameter m determines the length of the filter P1 (which is equal to 2.m).

The output of the operator OP1 is also connected to a second filter U1 connected to a second approximation unit A2.

The second approximation unit A2 is connected to a second operator OP2, to which the first decimator by two D1 is also connected. The second operator OP2 effects a sum.

The output of the second operator OP2 is the second output S2 of the transformation unit, which delivers a digital signal including low-frequency samples y2i calculated according to the formula y 2 i = x 2 i + APP &Sgr; n k = 1 - n u k y 2 ( i + k ) - 1

where uk is a second filtering coefficient, that is to say the current coefficient of the filter U1, and the parameter n determines the length of the filter U1.

The function APP can be one of those described previously.

Figure 6 depicts a first embodiment of an elementary synthesis unit, that is to say the reverse transformation to that of Figure 5. This transformation unit has a first input E10 to which a first signal to be transformed is applied and a second input E11 to which a second signal to be transformed is applied.

The signals to be transformed include here the samples obtained after analysis filtering of a digital image by the analysis unit of Figure 5. These samples have possibly been modified by another processing between analysis and synthesis.

More precisely, the first signal to be transformed includes the high-frequency samples {y2i+1}, and the second signal to be transformed includes the low-frequency samples {y2i}.

The synthesis unit has a structure similar to that of the analysis unit, and can be derived from it simply. In particular, the synthesis unit uses the same filters P1 and U1 as the analysis unit.

The input E10 is connected to the filter U1, itself connected to an approximation unit A10. The approximation unit A10 is connected to an operator OP10 to which the input E11 is also connected. The operator OP10 effects a subtraction.

The output of the operator OP10 delivers a digital signal including reconstructed samples of even ranking X2i calculated according to the formula x 2 i = y 2 i - APP &Sgr; n k = 1 - n u k y 2 ( i + k ) - 1

where uk is a third filtering coefficient, here identical to the second, that is to say the current coefficient of the filter U1, and the parameter n determines the length of the filter U1 (which is equal to 2.n).

The function APP can be one of those described previously.

The output of the operator OP10 is connected to the filter P1, itself connected to an approximation unit A11. The approximation unit A11 is connected to an operator OP11, to which the input E10 is also connected. The operator OP11 performs an addition.

The output of the operator OP11 delivers a digital signal including reconstructed samples of odd ranking X2i+1 calculated according to the formula x 2 i + 1 = y 2 i + 1 + APP &Sgr; m k = 1 - m p k x 2 i + k

where Pk is a fourth filtering coefficient, here identical to the first, that is to say the current coefficient of the filter P1, and the parameter m determines the length of the filter P1 (which is equal to 2.m).

The function APP can be one of those described previously.

The output of the operator OP10 is connected to an interpolator by two IN10 and the output of the operator OP11 is connected to an interpolator by two IN11 itself connected to a delay R10. The interpolator IN10 and the delay R10 are connected to an operator OP13 which performs an addition in order to deliver at its output S10 the signal including the reconstructed samples xi.

The filter P1 is now detailed with reference to Figure 7. The filter P1 is of the form P 1 z = &Sgr; m k = 1 - m p k Z k

The set of coefficients pk is here chosen from the family of interpolator filters from Deslauriers-Dubuc, as presented in the article " The Lifting Scheme: A custom-design construction of biorthogonal wavelets", by Wim Sweldens, published in Applied and Computational Harmonic Analysis, 3(2) :186-200, 1996 and four examples of which, for four values of the parameter m, are given.

The first set includes the coefficients p0 = S and p1 = S.

The second set includes the coefficients p-1 = -1/16, p0 = 9/16, p1 = 9/16 and P2 = -1/16.

The third set includes the coefficients p-2 = 3/256, p-1 = -25/256, p0 = 150/256, p1 = 150/256, P2 = -25/256 and p3 = 3/256.

The fourth set includes the coefficients p-3 = -5/2048, p-2 = 49/2048, p-1 = -245/2048, p0 = 1225/2048, p1 = 1225/2048, p2 = -245/2048, p3 = 49/2048 and p4 = -5/2048.

The filter U1 is now detailed with reference to Figure 8. The filter U1 is of the form U1(z) =&Sgr;n k=1-n uk . zk

According to the invention, the set of coefficients uk is chosen from amongst three sets. The first set includes the coefficients u0 = 5/16 and u1 = 5/16. The second set includes the coefficients u-1 = -1/16, u0 = 5/16, u1 = 5/16 and u2 = -1/16. The third set includes the coefficients u-2 = 1/256, u-1 = -15/256, u0 = 78/256, u1 = 78/256, u2 = -15/256 and u3 = 1/256.

It is possible to combine any one of the filters P1 with any one of the filters U1. However, certain combinations have superior performance, both theoretical and experimental.

The preferred combination are presented in the table in Figure 9. The first preferred combination includes the filter P1 having the coefficients p-1 = -1/16, p0 = 9/16, p1 = 9/16 and p2 = -1/16 and the filter U1 having the coefficients u0 = 5/16 and u1 = 5/16. The elementary analysis and synthesis units thus produced are 9/7 filters, where 9 is the length of the low-pass filter and 7 is the length of the high-pass filter.

The second preferred combination includes the filter P1 having the coefficients p-1 = -1/16, p0 = 9/16, p1 = 9/16 and p2 = -1/16 and the filter U1 having the coefficients u-1 = -1/16, u0 = 5/16, u1 = 5/16 and u2 = -1/16. The elementary analysis and synthesis units thus produced are 13/7 filters where 13 is the length of the low-pass filter and 7 is the length of the high-pass filter.

The third preferred combination includes the filter P1 having the coefficients P-2 = 3/256, p-1 = -25/256, p0 = 150/16, p1 = 150/16, p2 = -25/256 and p3 = 3/256 and the filter U1 having the coefficients u-1 = -1/16, u0 = 5/16, u1 = 5/16 and u2 = -1/16. The elementary analysis and synthesis units thus produced are 17/11 filters where 17 is the length of the low-pass filter and 11 is the length of the high-pass filter.

The transformation units thus produced are filters of odd length. According to a balancing technique described in the book " Wavelets and Filters Banks " by Gilbert Strang and Truong Nguyen, Wellesley-Cambridge Press, 1996, pages 216 to 218 , it is possible to construct a filter of even length corresponding to a filter of odd length, and vice versa.

Thus, if the analysis filters, for example of odd length, of a unit BE (Figure 4) are denoted H0(z) and H1(z), the "twin" units H0twin(Z) and H1twin(z), of even length, are calculated by means of the formulae H 0 twin z = H 1 - z / ( 1 - z - 1 ) / 2 H 1 twin z = H 0 - z . ( 1 - z - 1 ) / 2

Applying the above formulae to the filters H0twin(Z) and H1twin(Z) gives the filters H0(z) and H1(z).

There is therefore a biunique correspondence between a filtering of even length and a filtering of odd length. Hereinafter the construction of filters of even length will be dealt with.

Figure 10 depicts another embodiment of an elementary transformation unit here performing an analysis filtering of a digital signal. This embodiment is equivalent to that of the unit BE, as disclosed in the article " The Lifting Scheme : A construction of second generation wavelets", by Wim Sweldens, Siam J. Math. Anal., Vol. 29, No. 2, pp 511-546, 1997 .

Compared with the elementary unit of Figure 5, this embodiment is differentiated mainly by the fact that it has three filters.

This unit has an input E20 to which the signal to be transformed is applied. The signal to be transformed includes a series of samples {xi}, where i is a sample ranking index.

The input E20 is connected to a first decimator by two D20 which delivers the samples {x2i} of even ranking.

The input E20 is also connected to an advance AV21 followed by a second decimator D21, which delivers the samples {x2i+1} of odd ranking.

The first decimator D20 is connected to a first filter P2 which filters the samples of even ranking and supplies them to a first approximation unit A20.

The output of the approximation unit A20 and that of the second decimator D21 are connected to a first operator OP20, which performs a subtraction. The filter P2 is chosen so that P2(z) = 1, where 1 is a fifth filtering coefficient, and the approximation unit delivers an approximate value of a variable which is equal to the variable.

Consequently the operator OP20 delivers intermediate samples y2i+1 calculated according to the formula y 2 i + 1 = x 2 i + 1 - x 2 i

This formula is a special case of formula (1), for the chosen filter P2 and approximation.

The output of the first operator OP20 is also connected to a second filter U2 connected to a second approximation unit A21.

The second approximation unit A21 is connected to a second operator OP21, to which the first decimator by two D20 is also connected. The second operator OP21 effects a sum.

The output of the second operator OP21 is the second output S21 of the transformation unit, which delivers a digital signal including low-frequency samples y2i calculated according to the formula y 2 i = x 2 i + APP &Sgr; n 1 k = 1 - n 1 v k y 2 ( i + k ) - 1

where vk is a sixth filtering coefficient, that is to say the current coefficient of the filter U2, and the parameter n1 determines the length of the filter U2.

The function APP can be one of those described previously.

The output of the second operator OP21 is also connected to a third filter Q2 which filters the samples y2i and supplies them to a third approximation unit A22. The operator OP20 is also connected to the third operator OP22, which effects a subtraction.

The operator OP22 is connected to a first output S20 of the elementary unit, which delivers a digital signal including high-frequency samples Z2i+1 calculated according to the formula z 2 i + 1 = y 2 i + 1 - APP &Sgr; m 1 k = 1 - m 1 q k y 2 i + k

where the parameter qk is a seventh filtering coefficient, that is to say the current coefficient of the filter Q2, and the parameter m1 determines the length of the filter Q2.

The function APP can be one of those described previously.

Figure 11 depicts another embodiment of an elementary synthesis unit, that is to say the reverse transformation to that of Figure 10. This transformation unit has a first input E30 to which a first signal to be transformed is applied and a second input E31 to which a second signal to be transformed is applied.

The signals to be transformed include here the samples obtained after analysis filtering of a digital image by the analysis unit of Figure 10. These samples have possibly been modified by another processing between analysis and synthesis.

More precisely, the first signal to be transformed includes the high-frequency samples {Z2i+1}, and the second signal to be transformed includes the low-frequency samples {y2i}.

The synthesis unit has a structure similar to that of the analysis unit, and can be derived therefrom simply. In particular, the synthesis unit uses the same filters P2, U2 and Q2 as the analysis unit.

The input E30 is connected to an operator OP30, which effects a sum.

The input E31 is connected to the filter Q2, itself connected to a first approximation unit A30. The approximation unit A30 is connected to the first operator OP30.

The output of the operator OP30 delivers a digital signal including intermediate samples y2i+1 calculated according to the formula y 2 i + 1 = z 2 i + 1 + APP &Sgr; m 1 - 1 k = 1 - m 1 q k y 2 i + k

where qk is an eighth filtering coefficient, here identical to the seventh, that is to say the current coefficient of the filter Q2, and the parameter m1 determines the length of the filter Q2.

The function APP can be one of those described previously.

The output of the first operator OP30 is connected to the filter U2, itself connected to a second approximation unit A31.

The second approximation unit A31 is connected to a second operator OP31, to which the input E31 is also connected. The operator OP31 effects a subtraction.

The output of the operator OP31 delivers a digital signal including reconstructed samples of even ranking x2i calculated according to the formula x 2 i = y 2 i - APP &Sgr; n 1 - 1 k = 1 - n 1 v k y 2 ( i + k ) - 1

Where vk is a ninth filtering coefficient, here identical to the sixth, that is to say the current coefficient of the filter U2, and the parameter n1 determines the length of the filter U2.

The function APP can be one of those previously described.

The output of the operator OP31 is connected to the filter P2, itself connected to an approximation unit A32. The approximation unit A32 is connected to a third operator OP32, to which the output of the operator OP30 is also connected.

As with the corresponding analysis unit, the filter P2 is chosen so that P2(z) = 1, and the approximation unit delivers an approximate value of a variable which is equal to the variable.

The operator OP32 effects an addition and delivers at its output a digital signal including reconstructed samples of odd ranking x2i+1 calculated according to the formula x 2 i + 1 = x 2 i + 1 + x 2 i

This formula includes a tenth filtering coefficient equal to 1. The output of the operator OP31 is connected to an interpolator by two IN31 and the output of the operator OP32 is connected to an interpolator by two IN32 itself connected to a delay R32. The interpolator IN31 and the delay R32 are connected to an operator OP33 which effects an addition in order to deliver at its output S30 the signal including the reconstructed samples xi.

The filter U2 is now detailed with reference to Figure 12. The filter U2 is of the form U2(z) = &Sgr;n1-1 k=1-n1 vk . zk

The set of coefficients {vk} is constructed from the family of interpolator filters P1(z) from Deslauriers-Dubuc, as presented in the article " The Lifting Scheme: A custom-design construction of biorthogonal wavelets", by Wim Sweldens, published in Applied and Computational Harmonic Analysis, 3(2) :186-200, 1996 . The relationship between the filter U2(z) and the filter P1(z) is : U2(z) = (1-P1(z))/(1-z-1).

Three examples of filters U2, for three values of the parameters n1, are given.

For n1 = 2, the filter U2 includes the coefficients v-1 = 1/16, v0 = S and v1 = -1/16.

For n1 = 3, the filter U2 includes the coefficients v-2 = -3/256, v-1 = 22/256, v0 = S, v1 = -22/256 and v2 = 3/256.

For n1 = 4, the filter U2 includes the coefficients v-3 = 5/2048, V-2 = -44/2048, v-1 = 201/2048, v0 = S, v1= -201/2048, v2 = 44/2048 and v3 = -5/2048.

The filter Q2 is now detailed with reference to Figure 13. The filter Q2 is of the form Q2(z) = &Sgr;m1 k=1-m1 qk . zk

According to the invention, the set of coefficients qk is chosen from amongst three sets. The first set, for m1 = 2, includes the coefficients q-1 = 5/16, q0 = 0 and q1 = -5/16. The second set, for m1 = 3, includes the coefficients q-2 = -1/16, q-1 = 6/16, q0 = 0, q1 = -6/16 and q2 = 1/16. The third set, for m1 = 4, includes the coefficients q-3 = 1/256, q-2 =-16/256, q-1 = 93/256, q0 = 0, q1 = -93/256, q2 = 16/256 and q3 = -1/256.

It is possible to combine any one of the filters U2 with any one of the filters Q2. However, certain combinations have superior performance, both theoretical and experimental.

The preferred combinations are presented in the table in Figure 14. The first preferred combination includes the filter U2 corresponding to n1 = 2 and the filter Q2 corresponding to m1 = 2. The elementary analysis and synthesis units thus produced are 6/10 filters.

The second preferred combination includes the filter U2 corresponding to n1 = 2 and the filter Q2 corresponding to m1 = 3. The elementary analysis and synthesis units thus produced are 6/14 filters.

The third preferred combination includes the filter U2 corresponding to n1 = 3 and the filter Q2 corresponding to m1 = 3. The elementary analysis and synthesis units thus produced are 10/18 filters.

For all the calculation formulae presented above, the edge problems are resolved in a conventional fashion, that is to say any non-existing samples are replaced by other samples determined for example by the principle of symmetrical reflection.

In addition, all the calculation formulae presented above can be multiplied by factors, so as to form intermediate formula, which are then standardised in order to obtain the results presented. This makes it possible to reduce the approximation errors and/or to reduce the number of multiplications.

It should be noted that all the filters constructed according to the invention have coefficients whose denominators are powers of two, in order to simplify implementation. This is because a division by a power of two can be effected by a shifting of bits.

Moreover, the transformation units of Figures 5, 6, 10 and 11 can be implemented in the device 10 (Figure 3), by programs performing the operations of each of the components of the units described. In addition, implementation can be effected by calculations in place, that is to say where a calculated sample is stored in place of another sample which was used for calculating it.


Anspruch[de]
Digitales Filter, das angepasst ist zum Umsetzen von einem digitalen Eingangssignal in zwei digitale Ausgangssignale, wobei

das digitale Eingangssignal Eingangsabtastwerte (xi) umfasst,

das erste digitale Ausgangssignal Ausgangsabtastwerte (y2i+1) mit ungerader Rangordnung umfasst, wobei jeder Ausgangsabtastwert mit ungerader Rangordnung gleich dem Eingangsabtastwert (x2i+1) mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Eingangsabtastwerten mit gerader Rangordnung, die jeweils mit ersten Filterkoeffizienten (pk) multipliziert sind, verringert ist,

das zweite digitale Ausgangssignal Ausgangsabtastwerte (y2i) mit gerader Rangordnung umfasst, wobei jeder Ausgangsabtastwert mit gerader Rangordnung gleich dem Eingangsabtastwert (x2i) mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Ausgangsabtastwerten mit ungerader Rangordnung, die jeweils mit zweiten Filterkoeffizienten (uk) multipliziert sind, erhöht ist,

dadurch gekennzeichnet, dass

die zweiten Filterkoeffizienten (uk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Mengen von Koeffizienten ausgewählt ist: [5/16, 5/16], [-1/16, 5/16, 5/16, -1/16] und [1/256, -15/256, 78/256, 78/256, -15/256, 1/256].
Filter gemäß Anspruch 1, dadurch gekennzeichnet, dass die ersten Filterkoeffizienten (pk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Mengen von Koeffizienten ausgewählt ist: [1/2, 1/2], [-1/16, 9/16, 9/16, -1/16], [3/256, -25/256, 150/256, 150/256, -25/256, 3/256], [-5/2048, 49/2048, -245/2048, 1225/2048, 1225/2048, -245/2048, 49/2048, -5/2048]. Digitales Filter, das angepasst ist zum Umsetzen von zwei digitalen Eingangssignalen in ein digitales Ausgangssignal, wobei

ein erstes Eingangssignal Eingangsabtastwerte (y2i+1) mit ungerader Rangordnung umfasst und ein zweites Eingangssignal Eingangssignale (y2i) mit gerader Rangordnung umfasst,

das digitale Ausgangssignal Ausgangsabtastwerte mit gerader Rangordnung und ungerader Rangordnung (xi)

umfasst;

jeder Ausgangsabtastwert mit gerader Rangordnung gleich dem Eingangsabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Eingangsabtastwerten mit ungerader Rangordnung, die jeweils mit dritten Filterkoeffizienten (uk) multipliziert sind, verringert ist,

jeder Ausgangsabtastwert mit ungerader Rangordnung gleich dem Eingangsabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Ausgangsabtastwerten mit gerader Rangordnung, die jeweils mit vierten Filterkoeffizienten (pk) multipliziert sind, erhöht ist,

dadurch gekennzeichnet, dass

die dritten Filterkoeffizienten (uk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Mengen von Koeffizienten ausgewählt ist: [5/16, 5/16], [-1/16, 5/16, 5/16, -1/16] und [1/256, -15/256, 78/256, 78/256, -15/256, 1/256].
Filter gemäß Anspruch 3, dadurch gekennzeichnet, dass die vierten Filterkoeffizienten (pk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Mengen von Koeffizienten ausgewählt ist: [1/2, 1/2], [-1/16, 9/16, 9/16, -1/16], [3/256, -25/256, 150/256, 150/256, -25/256, 3/256], [-5/2048, 49/2048, -245/2048, 1225/2048, 1225/2048, -245/2048, 49/2048, -5/2048]. Digitales Filter, das angepasst ist zum Umsetzen von einem digitalen Eingangssignal in zwei digitale Ausgangssignale, wobei

das digitale Eingangssignal Eingangsabtastwerte (xi) umfasst,

ein digitales Zwischensignal Zwischenabtastwerte mit ungerader Rangordnung (y2i+1) umfasst, wobei jeder Zwischenabtastwert mit ungerader Rangordnung gleich dem Eingangsabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Eingangsabtastwerten mit gerader Rangordnung, die jeweils mit fünften Filterkoeffizienten multipliziert sind, verringert ist,

ein digitales Ausgangssignal Ausgangsabtastwerte (y2i) mit gerader Rangordnung umfasst, wobei jeder Ausgangsabtastwert mit gerader Rangordnung gleich dem Eingangsabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Zwischenabtastwerten mit ungerader Rangordnung, die jeweils mit sechsten Filterkoeffizienten (vk) multipliziert sind, erhöht ist,

das andere digitale Ausgangssignal Ausgangsabtastwerte (z2i+1) mit ungerader Rangordnung umfasst, wobei jeder Ausgangsabtastwert (z2i+1) mit ungerader Rangordnung gleich dem Zwischenabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Zwischenabtastwerten mit gerader Rangordnung, die jeweils mit siebten Filterkoeffizienten (qk) multipliziert sind, verringert ist,

dadurch gekennzeichnet, dass

die siebten Filterkoeffizienten (qk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Koeffizienten ausgewählt ist: [5/16, 0, -5/16], [-1/16, 6/16, 0, -6/16, 1/16] und [1/256, -16/256, 93/256, 0, -93/256, 16/256, -1/256].
Filter gemäß Anspruch 5, dadurch gekennzeichnet, dass die sechsten Filterkoeffizienten (Vk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Mengen von Koeffizienten ausgewählt ist: [1/16, 1/2, -1/16], [-3/256, 22/256, 1/2, -22/256, 3/256], [5/2048, -44/2048, 201/2048, 1/2, -201/2048, 44/2048, -5/2048]. Filter gemäß Anspruch 5 oder 6, dadurch gekennzeichnet, dass es einen einzigen fünften Filterkoeffizienten aufweist, und dass dieser gleich eins ist. Digitales Filter, das angepasst ist zum Umsetzen von zwei digitalen Eingangssignalen in ein digitales Ausgangssignal, wobei

ein Eingangssignal Eingangsabtastwerte (z2i+1) mit ungerader Rangordnung umfasst und das andere Eingangssignal Eingangssignale mit gerader Rangordnung (y2i) umfasst,

das digitale Ausgangssignal Ausgangsabtastwerte (xi) mit gerader Rangordnung und ungerader Rangordnung umfasst,

ein digitales Zwischensignal Zwischenabtastwerte (y2i+1) mit ungerader Rangordnung umfasst, wobei jeder Zwischenabtastwert mit ungerader Rangordnung gleich dem Eingangsabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Eingangsabtastwerten mit gerader Rangordnung, die jeweils mit achten Filterkoeffizienten (qk) multipliziert sind, erhöht ist (OP30),

jeder Ausgangsabtastwert (X2i) mit gerader Rangordnung gleich dem Eingangsabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Zwischenabtastwerten mit ungerader Rangordnung, die jeweils mit neunten Filterkoeffizienten (vk) multipliziert sind, verringert ist (OP31),

jeder Ausgangsabtastwert (x2i+1) mit ungerader Rangordnung gleich dem Zwischenabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Ausgangsabtastwerten mit gerader Rangordnung, die jeweils mit zehnten Filterkoeffizienten multipliziert sind, erhöht ist (OP32),

dadurch gekennzeichnet, dass

die achten Filterkoeffizienten (qk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Mengen von Koeffizienten ausgewählt ist: [5/16, 0, -5/16], [-1/16, 6/16, 0, -6/16, 1/16] und [1/256, -16/256, 93/256, 0, -93/256, 16/256, -1/256].
Filter gemäß Anspruch 8, dadurch gekennzeichnet, dass die neunten Filterkoeffizienten (vk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Mengen von Koeffizienten ausgewählt ist: [1/16, 1/2, -1/16], [-3/256, 22/256, 1/2, -22/256, 3/256], [5/2048, -44/2048, 201/2048, 1/2, -201/2048, 44/2048, -5/2048]. Filter gemäß Anspruch 8 oder 9, dadurch gekennzeichnet, dass es einen einzigen zehnten Filterkoeffizienten aufweist, und dass dieser gleich eins ist. Filter gemäß einem der Ansprüche 1 bis 10, dadurch gekennzeichnet, dass der Näherungswert die Identitätsfunktion ist. Filter gemäß einem der Ansprüche 1 bis 10, dadurch gekennzeichnet, dass der Näherungswert eine Funktion einer reellen Variablen ist, die die Ganzzahl liefert, die am nächsten an der Variablen liegt. Filter gemäß einem der Ansprüche 1 bis 10, dadurch gekennzeichnet, dass der Näherungswert eine Funktion einer reellen Variablen ist, die die Ganzzahl liefert, die unter der Variablen liegt. Filter gemäß einem der Ansprüche 1 bis 10, dadurch gekennzeichnet, dass der Näherungswert eine Funktion der reellen Variablen ist, die die Ganzzahl liefert, die über der Variablen liegt. Filter gemäß einem der Ansprüche 1 bis 10, dadurch gekennzeichnet, dass der Näherungswert eine Funktion einer reellen Variablen ist, die in Untervariable zerlegt wird, deren Summe gleich der Variablen ist, wobei die Funktion eine Summe von Näherungswerten der Untervariablen liefert, wobei jeder der Näherungswerte der Untervariablen im Einklang mit einem der Ansprüche 12 bis 14 steht. Signalverarbeitungsvorrichtung (2,5) umfassend das Filter gemäß einem der Ansprüche 1 bis 15. Signalverarbeitungsvorrichtung (2,5) umfassend zumindest zwei Filter gemäß einem der Ansprüche 1 bis 15, wobei das Ausgangssignal von einem der Filter das Eingangssignal von dem anderen Filter ist. Filterverfahren, das angepasst ist zum Umsetzen von einem digitalen Eingangssignal in zwei digitale Ausgangssignale, wobei

das digitale Eingangssignal Eingangsabtastwerte (xi) umfasst,

das erste digitale Ausgangssignal Ausgangsabtastwerte (y2i+1) mit ungerader Rangordnung umfasst, wobei jeder Ausgangsabtastwert mit ungerader Rangordnung gleich dem Eingangsabtastwert (x2i+1) mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Eingangsabtastwerten mit gerader Rangordnung, die jeweils mit ersten Filterkoeffizienten (pk) multipliziert sind, verringert ist,

das zweite digitale Ausgangssignal Ausgangsabtastwerte (y2i) mit gerader Rangordnung umfasst, wobei jeder Ausgangsabtastwert mit gerader Rangordnung (y2i) gleich dem Eingangsabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Ausgangsabtastwerten mit ungerader Rangordnung, die jeweils mit zweiten Filterkoeffizienten (uk) multipliziert sind, erhöht ist,

dadurch gekennzeichnet, dass

die zweiten Filterkoeffizienten (uk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Mengen von Koeffizienten ausgewählt ist: [5/16, 5/16], [-1/16, 5/16, 5/16, -1/16] und [1/256, -15/256, 78/256, 78/256, -15/256, 1/256].
Filterverfahren gemäß Anspruch 18, dadurch gekennzeichnet, dass die ersten Filterkoeffizienten (pk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Mengen von Koeffizienten ausgewählt ist: [1/2, 1/2], [-1/16, 9/16, 9/16, -1/16], [3/256, -25/256, 150/256, 150/256, -25/256, 3/256], [-5/2048, 49/2048, -245/2048, 1225/2048, 1225/2048, -245/2048, 49/2048, -5/2048]. Filterverfahren, das angepasst ist zum Umsetzen von zwei digitalen Eingangssignalen in ein digitales Ausgangssignal, wobei

ein erstes Eingangssignal Eingangsabtastwerte (y2i+1) mit ungerader Rangordnung umfasst und ein zweites Eingangssignal Eingangssignale (y2i) mit gerader Rangordnung umfasst,

das digitale Ausgangssignal Ausgangsabtastwerte (xi) mit gerader Rangordnung und ungerader Rangordnung

umfasst;

jeder Ausgangsabtastwert mit gerader Rangordnung gleich dem Eingangsabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Eingangsabtastwerten mit ungerader Rangordnung, die jeweils mit dritten Filterkoeffizienten (uk) multipliziert sind, verringert ist,

jeder Ausgangsabtastwert mit ungerader Rangordnung gleich dem Eingangsabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Ausgangsabtastwerten mit gerader Rangordnung, die jeweils mit vierten Filterkoeffizienten (pk) multipliziert sind, erhöht ist,

dadurch gekennzeichnet, dass

die dritten Filterkoeffizienten (uk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Mengen von Koeffizienten ausgewählt ist: [5/16, 5/16], [-1/16, 5/16, 5/16, -1/16] und [1/256, -15/256, 78/256, 78/256, -15/256, 1/256].
Filterverfahren gemäß Anspruch 20, dadurch gekennzeichnet, dass die vierten Filterkoeffizienten (pk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Mengen von Koeffizienten ausgewählt ist: [1/2, 1/2], [-1/16, 9/16, 9/16, -1/16], [3/256, -25/256, 150/256, 150/256, -25/256, 3/256], [-5/2048, 49/2048, -245/2048, 1225/2048, 1225/2048, -245/2048, 49/2048, -5/2048]. Filterverfahren, das angepasst ist zum Umsetzen von einem digitalen Eingangssignal in zwei digitale Ausgangssignale, wobei

das digitale Eingangssignal Eingangsabtastwerte (xi) umfasst,

ein digitales Zwischensignal Zwischenabtastwerte (y2i+1) mit ungerader Rangordnung umfasst, wobei jeder Zwischenabtastwert mit ungerader Rangordnung gleich dem Eingangsabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Eingangsabtastwerten mit gerader Rangordnung, die jeweils mit fünften Filterkoeffizienten multipliziert sind, verringert ist,

ein digitales Ausgangssignal Ausgangsabtastwerte (y2i) mit gerader Rangordnung umfasst, wobei jeder Ausgangsabtastwert mit gerader Rangordnung gleich dem Eingangsabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Zwischenabtastwerten mit ungerader Rangordnung, die jeweils mit sechsten Filterkoeffizienten (vk) multipliziert sind, erhöht ist,

das andere digitale Ausgangssignal Ausgangsabtastwerte (z2i+1) mit ungerader Rangordnung umfasst, wobei jeder Ausgangsabtastwert (z2i+1) mit ungerader Rangordnung gleich dem Zwischenabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Zwischenabtastwerten mit gerader Rangordnung, die jeweils mit siebten Filterkoeffizienten (qk) multipliziert sind, verringert ist,

dadurch gekennzeichnet, dass

die siebten Filterkoeffizienten (qk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Koeffizienten ausgewählt ist: [5/16, 0, -5/16], [-1/16, 6/16, 0, -6/16, 1/16] und [1/256, -16/256, 93/256, 0, -93/256, 16/256, -1/256].
Filterverfahren gemäß Anspruch 22, dadurch gekennzeichnet, dass die sechsten Filterkoeffizienten (vk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Mengen von Koeffizienten ausgewählt ist: [1/16, 1/2, -1/16], [-3/256, 22/256, 1/2, -22/256, 3/256], [5/2048, -44/2048, 201/2048, 1/2, -201/2048, 44/2048, -5/2048]. Filterverfahren gemäß Anspruch 22 oder 23, dadurch gekennzeichnet, dass es einen einzigen fünften Filterkoeffizienten aufweist, und dass dieser gleich eins ist. Filterverfahren, das angepasst ist zum Umsetzen von zwei digitalen Eingangssignalen in ein digitales Ausgangssignal, wobei

ein Eingangssignal Eingangsabtastwerte (Z2i+1) mit ungerader Rangordnung umfasst und das andere Eingangssignal Eingangssignale mit gerader Rangordnung (y2i) umfasst,

das digitale Ausgangssignal Ausgangsabtastwerte (xi) mit gerader Rangordnung und ungerader Rangordnung umfasst,

ein digitales Zwischensignal Zwischenabtastwerte (y2i+1) mit ungerader Rangordnung umfasst, wobei jeder Zwischenabtastwert mit ungerader Rangordnung gleich dem Eingangsabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Eingangsabtastwerten mit gerader Rangordnung, die jeweils mit achten Filterkoeffizienten (qk) multipliziert sind, erhöht ist,

jeder Ausgangsabtastwert (x2i) mit gerader Rangordnung gleich dem Eingangsabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Zwischenabtastwerten mit ungerader Rangordnung, die jeweils mit neunten Filterkoeffizienten (Vk) multipliziert sind, verringert ist,

jeder Ausgangsabtastwert (x2i+1) mit ungerader Rangordnung gleich dem Zwischenabtastwert mit der gleichen Rangordnung ist, der um den Näherungswert einer Summe von Ausgangsabtastwerten mit gerader Rangordnung, die jeweils mit zehnten Filterkoeffizienten multipliziert sind, erhöht ist,

dadurch gekennzeichnet, dass

die achten Filterkoeffizienten (qk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Mengen von Koeffizienten ausgewählt ist: [5/16, 0, -5/16], [-1/16, 6/16, 0, -6/16, 1/16] und [1/256, -16/256, 93/256, 0, -93/256, 16/256, -1/256].
Filterverfahren gemäß Anspruch 25, dadurch gekennzeichnet, dass die neunten Filterkoeffizienten (vk) eine Menge darstellen, die gemäß der Länge des Filters aus all den folgenden Mengen von Koeffizienten ausgewählt ist: [1/16, 1/2, -1/16], [-3/256, 22/256, 1/2, -22/256, 3/256], [5/2048, -44/2048, 201/2048, 1/2, -201/2048, 44/2048, -5/2048]. Filterverfahren gemäß Anspruch 25 oder 26, dadurch gekennzeichnet, dass es einen einzigen zehnten Filterkoeffizienten aufweist, und dass dieser gleich eins ist. Filterverfahren gemäß einem der Ansprüche 18 bis 27, dadurch gekennzeichnet, dass der Näherungswert die Identitätsfunktion ist. Filterverfahren gemäß einem der Ansprüche 18 bis 27, dadurch gekennzeichnet, dass der Näherungswert eine Funktion einer reellen Variablen ist, die die Ganzzahl liefert, die am nächsten an der Variablen liegt. Filterverfahren gemäß einem der Ansprüche 18 bis 27, dadurch gekennzeichnet, dass der Näherungswert eine Funktion einer reellen Variablen ist, die die Ganzzahl liefert, die unter der Variablen liegt. Filterverfahren gemäß einem der Ansprüche 18 bis 27, dadurch gekennzeichnet, dass der Näherungswert eine Funktion einer reellen Variablen ist, die die Ganzzahl liefert, die über der Variablen liegt. Filterverfahren gemäß einem der Ansprüche 18 bis 27, dadurch gekennzeichnet, dass der Näherungswert eine Funktion einer reellen Variablen ist, die in Untervariable zerlegt wird, deren Summe gleich der Variablen ist, wobei die Funktion eine Summe von Näherungswerten der Untervariablen liefert, wobei jeder der Näherungswerte der Untervariablen im Einklang mit einem der Ansprüche 29 bis 31 steht. Signalverarbeitungsvorrichtung (10) umfassend eine Einrichtung zum Implementieren des Filterverfahrens gemäß einem der Ansprüche 18 bis 32. Digitalvorrichtung umfassend die Signalverarbeitungsvorrichtung gemäß einem der Ansprüche 16, 17 oder 33. Digitalfotografievorrichtung umfassend die Signalverarbeitungsvorrichtung gemäß einem der Ansprüche 16, 17 oder 33.
Anspruch[en]
Digital filter adapted to transform an input digital signal into two output digital signals, - the input digital signal including input samples (xi), - the first output digital signal including output samples (y2i+1) of odd ranking, each output sample of odd ranking being equal to the input sample (x2i+1) of the same ranking decreased by the approximate value of a sum of input samples of even ranking respectively multiplied by first filtering coefficients (pk), - the second output digital signal including output samples (y2i) of even ranking, each output sample of even ranking being equal to the input sample (x2i) of the same ranking increased by the approximate value of a sum of output samples of odd ranking respectively multiplied by second filtering coefficients (uk) characterised in that - the second filtering coefficients (uk) are one set chosen according to the length of the filter from all the following sets of coefficients: [5/16, 5/16], [-1/16, 5/16, 5/16, -1/16] and [1/256, -15/256, 78/256, 78/256, -15/256, 1/256]. Filter according to Claim 1, characterised in that the first filtering coefficients (pk) are one set chosen according to the length of the filter from all the following sets of coefficients: [S, S], [-1/16, 9/16, 9/16, -1/16], [3/256, -25/256, 150/256, 150/256, -25/256, 3/256], [-5/2048, 4912048, -245/2048, 1225/2048, 1225/2048, -245/2048, 4912048, -512048]. Digital filter adapted to transform two input digital signals into one output digital signal, - a first input signal including input samples (y2i+1) of odd ranking and a second input signal including input signals (y2i) of even ranking, - the output digital signal including output samples of even ranking and odd ranking (xi), - each output sample of even ranking being equal to the input sample of the same ranking decreased by the approximate value of a sum of input samples of odd ranking respectively multiplied by third filtering coefficients (uk), - each output sample of odd ranking being equal to the input sample of the same ranking increased by the approximate value of a sum of output samples of even ranking respectively multiplied by fourth filtering coefficients (pk), characterised in that - the third filtering coefficients (uk) are one set chosen according to the length of the filter from all the following sets of coefficients: [5/16, 5/16], [-1/16, 5/16, 5/16, -1/16] and [1/256, -15/256, 78/256, 78/256, -15/256, 1/256]. Filter according to Claim 3, characterised in that the fourth filtering coefficients (pk) are one set chosen according to the length of the filter from all the following sets of coefficients: [1/2, 1/2], [-1/16, 9/16, 9/16 -1/16], [3/256, -25/256, 150/256, 150/256, -25/256, 3/256], [-5/2048, 49/2048, -245/2048, 1225/2048, 1225/2048, -245/2048, 49/2048, -5/2048]. Digital filter adapted to transform an input digital signal into two output digital signals, - the input digital signal including input samples (xi). - an intermediate digital signal including intermediate samples of odd ranking (y2i+1), each intermediate sample of odd ranking being equal to the input sample of the same ranking decreased by the approximate value of a sum of input samples of even ranking respectively multiplied by fifth filtering coefficients, - one output digital signal including output samples (y2i) of even ranking, each output sample of even ranking being equal to the input sample of the same ranking increased by the approximate value of a sum of intermediate samples of odd ranking respectively multiplied by sixth filtering coefficients (vk), - the other output digital signal including output samples (Z2i+1) of odd ranking, each output sample (Z2i+1) of odd ranking being equal to the intermediate sample of the same ranking decreased by the approximate value of a sum of intermediate samples of even ranking respectively multiplied by seventh filtering coefficients (qk), characterised in that - the seventh filtering coefficients (qk) are one set chosen according to the length of the filter from all the following sets of coefficients: [5/16, 0, -5/16], [-1/16, 6/16, 0, -6/16, 1/16], and [1/256, -16/256, 93/256, 0, -93/256, 16/256, -1/256]. Filter according to Claim 5, characterised in that the sixth filtering coefficients (vk) are one set chosen according to the length of the filter from all the following sets of coefficients: [1/16, 1/2, -1/16], [-3/256, 22/256, S, -22/256, 3/256], [5/2048, -44/2048, 201/2048, 1/2, -201/2048, 44/2048, -5/2048]. Filter according to Claim 5 or 6, characterised in that it has a single fifth filtering coefficient and in that it is equal to one. Digital filter adapted to transform two input digital signals into one output digital signal, - one input signal including input samples (z2i+1) of odd ranking and the other input signal including input signals of even ranking (y2i), - the output digital signal including output samples (xi) of even ranking and odd ranking, - an intermediate digital signal including intermediate samples (y2i+1) of odd ranking, each intermediate sample of odd ranking being equal to the input sample of the same ranking increased (OP30) by the approximate value of a sum of input samples of even ranking respectively multiplied by eighth filtering coefficients (qk), - each output sample (X2i) of even ranking being equal to the input sample of the same ranking decreased (OP31) by the approximate value of a sum of intermediate samples of odd ranking respectively multiplied by ninth filtering coefficients (vk), - each output sample (X2i+1) of odd ranking being equal to the intermediate sample of the same ranking increased (OP32) by the approximate value of a sum of output samples of even ranking respectively multiplied by tenth filtering coefficients, characterised in that - the eighth filtering coefficients (qk) are one set chosen according to the length of the filter from all the following sets of coefficients: [5/16, 0, -5/16], [-1/16, 6/16, 0, -6/16, 1/16], and [1/256, -16/256, 93/256, 0, -93/256, 16/256, -1/256]. Filter according to Claim 8, characterised in that the ninth filtering coefficients (vk) are one set chosen according to the length of the filter from all the following sets of coefficients: [1/16, S, -1/16], [-3/256, 22/256, S, -22/256, 3/256], [5/2048, -44/2048, 201/2048, S, -201/2048, 44/2048, -5/2048]. Filter according to Claim 8 or 9, characterised in that it has a single tenth filtering coefficient and in that it is equal to one. Filter according to any one of Claims 1 to 10, characterised in that the approximate value is the identity function. Filter according to any one of Claims 1 to 10, characterised in that the approximate value is a function of a real variable which supplies the integer which is closest to the variable. Filter according to any one of Claims 1 to 10, characterised in that the approximate value is a function of a real variable which supplies the integer which is lower than the variable. Filter according to any one of Claims 1 to 10, characterised in that the approximate value is a function of a real variable which supplies the integer which is higher than the variable. Filter according to any one of Claims 1 to 10, characterised in that the approximate value is a function of a real variable broken down into sub-variables whose sum is equal to the variable, which supplies a sum of approximate values of the sub-variables, each of the approximate values of the sub-variables being in accordance with any one of Claims 12 to 14. Signal processing device (2,5) including the filter according to any one of Claims 1 to 15. Signal processing device (2,5) including at least two filters according to any one of Claims 1 to 15, the output signal of one of the filters being the input signal of the other filter. Filtering method adapted to transform an input digital signal into two output digital signals, - the input digital signal including input samples (xi). - the first output digital signal including output samples (y2i+1) of odd ranking, each output sample of odd ranking being equal to the input sample (x2i+1) of the same ranking decreased by the approximate value of a sum of input samples of even ranking respectively multiplied by first filtering coefficients (pk), - the second output digital signal including output samples (y2i) of even ranking, each output (y2i) sample of even ranking being equal to the input sample of the same ranking increased by the approximate value of a sum of output samples of odd ranking respectively multiplied by second filtering coefficients (uk), characterised in that - the second filtering coefficients (uk) are one set chosen according to the length of the filter from all the following sets of coefficients: [5/16, 5/16], [-1/16, 5/16, 5/16, -1/16] and [1/256, -15/256, 78/256, 78/256, -15/256, 1/256]. Filtering method according to Claim 18, characterised in that the first filtering coefficients (pk) are one set chosen according to the length of the filter from all the following sets of coefficients: [1/2, 1/2], [-1/16, 9/16, 9/16, -1/16], [3/256, -25/256, 150/256, 150/256, -25/256, 3/256], [-5/2048, 49/2048, -245/2048, 1225/2048, 1225/2048, -245/2048, 49/2048, -5/2048]. Filtering method adapted to transform two input digital signals into one output digital signal, - a first input signal including input samples (y2i+1)of odd ranking and a second input signal including input signals (y2i) of even ranking, - the output digital signal including output samples (xi) of even ranking and odd ranking, - each output sample ot even ranking being equal to the input sample of the same ranking decreased by the approximate value of a sum of input samples of odd ranking respectively multiplied by third filtering coefficients (uk), - each output sample of odd ranking being equal to the input sample of the same ranking increased by the approximate value of a sum of output samples of even ranking respectively multiplied by fourth filtering coefficients (pk), characterised in that - the third filtering coefficients (uk) are one set chosen according to the length of the filter from all the following sets of coefficients: [5/16. 5/16]. [-1/16, 5/16, 5/16, -1/16] and [1/256, -15/256, 78/256, 78/256, -15/256, 1/256]. Filtering method according to Claim 20, characterised in that the fourth filtering coefficients (pk) are one set chosen according to the length of the filter from all the following sets of coefficients: [S, S], [-1/16, 9/16, 9/16 -1/16], [3/256, -25/256, 150/256, 150/256, -25/256, 3/256], [-5/2048, 49/2048, -245/2048, 1225/2048, 1225/2048, -245/2048, 49/2048, -5/2048]. Filtering method adapted to transform an input digital signal into two output digital signals, - the input digital signal including input samples (Xi), - an intermediate digital signal including intermediate samples (y2i+1) of odd ranking, each intermediate sample of odd ranking being equal to the input sample of the same ranking decreased by the approximate value of a sum of input samples of even ranking respectively multiplied by fifth filtering coefficients, - one output digital signal including output samples (y2i) of even ranking, each output sample of even ranking being equal to the input sample of the same ranking increased by the approximate value of a sum of intermediate samples of odd ranking respectively multiplied by sixth filtering coefficients (vk), - the other output digital signal including output samples (z2i+1) of odd ranking, each output sample of odd ranking being equal to the intermediate sample of the same ranking decreased by the approximate value of a sum of intermediate samples of even ranking respectively multiplied by seventh filtering coefficients (qk), characterised in that - the seventh filtering coefficients (qk) are one set chosen according to the length of the filter from all the following sets of coefficients: [5/16, 0, -5/16], [-1/16, 6/16, 0, -6/16, 1/16], and [1/256, -16/256, 93/256, 0, -93/256, 16/256, -1/256]. Filtering method according to Claim 22, characterised in that the sixth filtering coefficients (vk) are one set chosen according to the length of the filter from all the following sets of coefficients: [1/16, S, -1/16], [-3/256, 22/256, S, -22/256, 3/256], [5/2048, -44/2048, 201/2048, S, -201/2048, 44/2048, -5/2048]. Filtering method according to Claim 22 or 23, characterised in that it has a single fifth filtering coefficient and in that it is equal to one. Filtering method adapted to transform two input digital signals into one output digital signal, - one input signal including input samples (Z2i+1) of odd ranking and the other input signal including input signals (y2i) of even ranking, - the output digital signal including output samples (Xi) of even ranking and odd ranking, - an intermediate digital signal including intermediate samples (y2i+1) of odd ranking, each intermediate sample of odd ranking being equal to the input sample of the same ranking increased by the approximate value of a sum of input samples of even ranking respectively multiplied by eighth filtering coefficients (qk), - each output sample of even ranking being equal to the input sample of the same ranking decreased by the approximate value of a sum of intermediate samples of odd ranking respectively multiplied by ninth filtering coefficients (vk). - each output sample (x2i+1) of odd ranking being equal to the intermediate sample of the same ranking increased by the approximate value of a sum of output samples of even ranking respectively multiplied by tenth filtering coefficients, characterised in that - the eighth filtering coefficients (qk) are one set chosen according to the length of the filter from all the following sets of coefficients: [5/16, 0, -5/16], [-1/16, 6/16, 0, -6/16, 1/16], and [1/256, -16/256, 93/256, 0, -93/256, 16/256, -1/256]. Filtering method according to Claim 25, characterised in that the ninth filtering coefficients (vk) are one set chosen according to the length of the filter from all the following sets of coefficients: [1/16, S, -1/16], [-3/256,22/256, S, -22/256,3/256], [5/2048, -44/2048, 201/2048, S, -201/2048, 44/2048, -5/2048]. Filtering method according to Claim 25 or 26, characterised in that it has a single tenth filtering coefficient and in that it is equal to one. Filtering method according to any one of Claims 18 to 27, characterised in that the approximate value is the identity function. Filtering method according to any one of Claims 18 to 27, characterised in that the approximate value is a function of a real variable which supplies the integer which is closest to the variable. Filtering method according to any one of Claims 18 to 27, characterised in that the approximate value is a function of a real variable which supplies the integer which is lower than the variable. Filtering method according to any one of Claims 18 to 27, characterised in that the approximate value is a function of a real variable which supplies the integer which is higher than the variable. Filtering method according to any one of Claims 18 to 27, characterised in that the approximate value is a function of a real variable broken down into sub-variables whose sum is equal to the variable, which supplies a sum of approximate values of the sub-variables, each of the approximate values of the sub-variables being in accordance with any one of Claims 29 to 31. Signal processing device (10) including means of implementing the filtering method according to any one of Claims 18 to 32. Digital apparatus including the signal processing device according to any one of Claims 16, 17 or 33. Digital photographic apparatus including the signal processing device according to any one of Claims 16, 17 or 33
Anspruch[fr]
Filtre numérique apte à transformer un signal numérique d'entrée en deux signaux numériques de sortie, - le signal numérique d'entrée comprenant des échantillons d'entrée (xi), - le premier signal numérique de sortie comprenant des échantillons de sortie (y2i+1) de rang impair, chaque échantillon de sortie de rang impair étant égal à l'échantillon d'entrée (x2i+1) du même rang réduit de la valeur approximative d'une somme d'échantillons d'entrée de même rang respectivement multipliés par des premiers coefficients de filtrage (pk), - le deuxième signal numérique de sortie comprenant des échantillons de sortie (y2i) de rang pair, chaque échantillon de sortie de rang pair étant égal à l'échantillon d'entrée (x2i) de même rang augmenté de la valeur approximative d'une somme d'échantillons de sortie de rang impair respectivement multipliés par des deuxièmes coefficients de filtrage (uk), caractérisé en ce que : - les deuxièmes coefficients de filtrage (uk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients : [5/16, 5/16], [-1/16, 5/16, 5/16, -1/16] et [1/256, -15/256, 78/256, 78/256, -15/256, 1/256]. Filtre selon la revendication 1, caractérisé en ce que les premiers coefficients de filtrage (pk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients : [1/2, 1/2], [-1/16, 9/16, 9/16, -1/16], [3/256, -25/256, 150/256, 150/256, -25/256, 3/256], [-5/2048, 49/2048, -245/2048, 1225/2048, 1225/2048, -245/2048, 49/2048, -5/2048]. Filtre numérique apte à transformer deux signaux numériques d'entrée en un signal numérique de sortie, - un premier signal d'entrée comprenant des échantillons d'entrée (y2i+1) de rang impair et un deuxième signal d'entrée comprenant des signaux d'entrée (y2i) de rang pair, - le signal numérique de sortie comprenant des échantillons de sortie de rang pair et de rang impair (xi), - chaque échantillon de sortie de rang pair étant égal à l'échantillon d'entrée de même rang réduit de la valeur approximative d'une somme d'échantillons d'entrée de rang impair respectivement multipliés par des troisièmes coefficients de filtrage (uk), - chaque échantillon de sortie de rang impair étant égal à l'échantillon d'entrée de même rang augmenté de la valeur approximative d'une somme d'échantillons de sortie de rang pair respectivement multipliés par des quatrièmes coefficients de filtrage (pk), caractérisé en ce que : - les troisièmes coefficients de filtrage (uk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients : [5/16, 5/16], [-1/16, 5/16, 5/16, -1/16] et [1/256, -15/256, 78/256, 78/256, -15/256, 1/256]. Filtre selon la revendication 3, caractérisé en ce que les quatrièmes coefficients de filtrage (pk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients : [1/2, 1/2], [-1/16, 9/16, 9/16, -1/16], [3/256, -25/256, 150/256, 150/256, -25/256, 3/256], [-5/2048, 49/2048, -245/2048, 1225/2048, 1225/2048, -245/2048, 49/2048, -5/2048]. Filtre numérique apte à transformer un signal numérique d'entrée en deux signaux numériques de sortie, - le signal numérique d'entrée comprenant des échantillons d'entrée (xi), - un signal numérique intermédiaire comprenant des échantillons intermédiaires de rang impair (y2i+1), chaque échantillon intermédiaire de rang impair étant égal à l'échantillon d'entrée de même rang réduit de la valeur approximative d'une somme d'échantillons d'entrée de rang pair respectivement multipliés par des cinquièmes coefficients de filtrage, - un signal numérique de sortie comprenant des échantillons de sortie (y2i) de rang pair, chaque échantillon de sortie de rang pair étant égal à l'échantillon d'entrée de même rang augmenté de la valeur approximative d'une somme d'échantillons intermédiaires de rang impair respectivement multipliés par des sixièmes coefficients de filtrage (vk), - l'autre signal numérique de sortie comprenant des échantillons de sortie (Z2i+1) de rang impair, chaque échantillon de sortie (Z2i+1) de rang impair étant égal à l'échantillon intermédiaire de même rang réduit de la valeur approximative d'une somme d'échantillons intermédiaires de rang pair respectivement multipliés par des septièmes coefficients de filtrage (qk), caractérisé en ce que : - les septièmes coefficients de filtrage (qk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients : [5/16, 0, -5/16], [-1/16, 6/16, 0, -6/16, 1/16] et [1/256, - 16/256, 93/256, 0, -93/256, 16/256, -1/256]. Filtre selon la revendication 5, caractérisé en ce que les sixièmes coefficients de filtrage (vk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients: [1/16, 1/2, -1/16], [-3/256, 22/256, 1/2, -22/256, 3/256], [5/2048, -44/2048, 201/2048, 1/2, -201/2048, 44/2048, -5/2048]. Filtre selon la revendication 5 ou 6, caractérisé en ce qu'il a un cinquième coefficient de filtrage unique et en ce que celui-ci est égal à un. Filtre numérique apte à transformer deux signaux numériques d'entrée en un signal numérique de sortie, - un signal d'entrée comprenant des échantillons d'entrée (Z2i+1) de rang impair et l'autre signal d'entrée comprenant des signaux d'entrée de rang pair (y2i), - le signal numérique de sortie comprenant des échantillons de sortie (xi) de rang pair et de rang impair, - un signal numérique intermédiaire comprenant des échantillons intermédiaires (y2i+1) de rang impair, chaque échantillon intermédiaire de rang impair étant égal à l'échantillon d'entrée de même rang augmenté (OP30) de la valeur approximative d'une somme d'échantillons d'entrée de rang pair respectivement multipliés par des huitièmes coefficients de filtrage (qk), - chaque échantillon de sortie (x2i) de rang pair étant égal à l'échantillon d'entrée de même rang réduit (OP31) de la valeur approximative d'une somme d'échantillons intermédiaires de rang impair respectivement multipliés par des neuvièmes coefficients de filtrage (vk), - chaque échantillon de sortie (X2i+1) de rang impair étant égal à l'échantillon intermédiaire de même rang augmenté (OP32) de la valeur approximative d'une somme d'échantillons de sortie de rang pair respectivement multipliés par des dixièmes coefficients de filtrage, caractérisé en ce que : - les huitièmes coefficients de filtrage (qk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients: [5/16, 0, -5/16], [-1/16, 6/16, 0, -6/16, 1/16] et [1/256, - 16/256, 93/256, 0, -93/256, 16/256, -1/256]. Filtre selon la revendication 8, caractérisé en ce que les neuvièmes coefficients de filtrage (vk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients: [1/16, 1/2, - 1/16], [-3/256, 22/256, 1/2 -22/256, 3/256], [5/2048, - 44/2048, 201/2048, 1/2, -201/2048, 44/2048, -5/2048]. Filtre selon la revendication 8 ou 9, caractérisé en ce qu'il a un dixième coefficient de filtrage unique et en ce que celui-ci est égal à un. Filtre selon l'une quelconque des revendications 1 à 10, caractérisé en ce que la valeur approximative est la fonction identité. Filtre selon l'une quelconque des revendications 1 à 10, caractérisé en ce que la valeur approximative est une fonction d'une variable réelle qui fournit l'entier le plus proche de la variable. Filtre selon l'une quelconque des revendications 1 à 10, caractérisé en ce que la valeur approximative est une fonction d'une variable réelle qui fournit l'entier inférieur à la variable. Filtre selon l'une quelconque des revendications 1 à 10, caractérisé en ce que la valeur approximative est une fonction d'une variable réelle qui fournit l'entier supérieur à la variable. Filtre selon l'une quelconque des revendications 1 à 10, caractérisé en ce que la valeur approximative est une fonction d'une variable réelle décomposée en des sous-variables dont la somme est égale à la variable, qui fournit une somme de valeurs approximatives des sous-variables, chacune des valeurs approximatives des sous-variables étant conforme à l'une quelconque des revendications 12 à 14. Dispositif de traitement du signal (2, 5) comprenant le filtre selon l'une quelconque des revendications 1 à 15. Dispositif de traitement du signal (2, 5) comprenant au moins deux filtres selon l'une quelconque des revendications 1 à 15, le signal de sortie de l'un des filtres étant le signal d'entrée de l'autre filtre. Procédé de filtrage apte à transformer un signal numérique d'entrée en deux signaux numériques de sortie, - le signal numérique d'entrée comprenant des échantillons d'entrée (xi), - le premier signal numérique de sortie comprenant des échantillons de sortie (y2i+1) de rang impair, chaque échantillon de sortie de rang impair étant égal à l'échantillon d'entrée (X2i+1) de même rang réduit de la valeur approximative d'une somme d'échantillons d'entrée de rang pair respectivement multipliés par des premiers coefficients de filtrage (pk), - le deuxième signal numérique de sortie comprenant des échantillons de sortie (y2i) de rang pair, chaque échantillon de sortie (y2i) de rang pair étant égal à l'échantillon d'entrée de même rang augmenté de la valeur approximative d'une somme d'échantillons de sortie de rang impair respectivement multipliés par des deuxièmes coefficients de filtrage (uk), caractérisé en ce que : - les deuxièmes coefficients de filtrage (uk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients : [5/16, 5/16], [-1/16, 5/16, 5/16, -1/16] et [1/256, -15/256, 78/256, 78/256, -15/256, 1/256]. Procédé de filtrage selon la revendication 18, caractérisé en ce que les premiers coefficients de filtrage (pk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients : [1/2, 1/2], [-1/16, 9/16, 9/16, -1/16], [3/256, -25/256, 150/256, 150/256, -25/256, 3/256], [-5/2048, 49/2048, -245/2048, 1225/2048, 1225/2048, -245/2048, 49/2048, -5/2048]. Procédé de filtrage apte à transformer deux signaux numériques d'entrée en un signal numérique de sortie, - un premier signal d'entrée comprenant des échantillons d'entrée (y2i+1) de rang impair et un deuxième signal d'entrée comprenant des signaux d'entrée (y2i) de rang pair, - le signal numérique de sortie comprenant des échantillons de sortie (xi) de rang pair et de rang impair, - chaque échantillon de sortie de rang pair étant égal à l'échantillon d'entrée de même rang réduit de la valeur approximative d'une somme d'échantillons d'entrée de rang impair respectivement multipliés par des troisièmes coefficients de filtrage (uk), - chaque échantillon de sortie de rang impair étant égal à l'échantillon d'entrée de même rang augmenté de la valeur approximative d'une somme d'échantillons de sortie de rang pair respectivement multipliés par des quatrièmes coefficients de filtrage (pk), caractérisé en ce que : - les troisièmes coefficients de filtrage (uk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients: [5/16, 5/16], [-1/16, 5/16, 5/16, -1/16] et [1/256, -15/256, 78/256, 78/256, -15/256, 1/256]. Procédé de filtrage selon la revendication 20, caractérisé en ce que les quatrièmes coefficients de filtrage (pk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients : [1/2, 1/2], [-1/16, 9/16, 9/16, -1/16], [3/256, -25/256, 150/256, 150/256, -25/256, 3/256], [-5/2048, 49/2048, -245/2048, 1225/2048, 1225/2048, -245/2048, 49/2048, -5/2048]. Procédé de filtrage apte à transformer un signal numérique d'entrée en deux signaux numériques de sortie, - le signal numérique d'entrée comprenant des échantillons d'entrée (xi), - un signal numérique intermédiaire comprenant des échantillons intermédiaires (y2i+1) de rang impair, chaque échantillon intermédiaire de rang impair étant égal à l'échantillon d'entrée du même rang réduit de la valeur approximative d'une somme d'échantillons d'entrée de rang pair respectivement multipliés par des cinquièmes coefficients de filtrage, - un signal numérique de sortie comprenant des échantillons de sortie (y2i) de rang pair, chaque échantillon de sortie de rang pair étant égal à l'échantillon d'entrée de même rang augmenté de la valeur approximative d'une somme des échantillons intermédiaires de rang impair respectivement multipliés par des sixièmes coefficients de filtrage (vk), - l'autre signal numérique de sortie comprenant des échantillons de sortie (z2i+1) de rang impair, chaque échantillon de sortie de rang impair étant égal à l'échantillon intermédiaire de même rang réduit de la valeur approximative d'une somme d'échantillons intermédiaires de rang pair respectivement multipliés par des septièmes coefficients de filtrage (qk), caractérisé en ce que : - les septièmes coefficients de filtrage (qk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients : [5/16, 0, -5/16], [-1/16, 6/16, 0, -6/16, 1/16] et [1/256, -16/256, 93/256, 0, -93/256, 16/256, -1/256]. Procédé de filtrage selon la revendication 22, caractérisé en ce que les sixièmes coefficients de filtrage (vk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients : [1/16, 1/2, -1/16], [-3/256, 22/256, 1/2, -22/256, 3/256], [5/2048, -44/2048, 201/2048, 1/2, -201/2048, 44/2048, -5/2048]. Procédé de filtrage selon la revendication 22 ou 23, caractérisé en ce qu'il a un cinquième coefficient de filtrage unique et en ce que ce dernier est égal à un. Procédé de filtrage apte à transformer deux signaux numériques d'entrée en un signal numérique de sortie, - un signal d'entrée comprenant des échantillons d'entrée (Z2i+1) de rang impair et l'autre signal d'entrée comprenant des signaux d'entrée (y2i) de rang pair, - le signal numérique de sortie comprenant des échantillons de sortie (xi) de rang pair et de rang impair, - un signal numérique intermédiaire comprenant des échantillons intermédiaires (y2i+1) de rang impair, chaque échantillon intermédiaire de rang impair étant égal à l'échantillon d'entrée de même rang augmenté de la valeur approximative d'une somme d'échantillons d'entrée de rang pair respectivement multipliés par des huitièmes coefficients de filtrage (qk), - chaque échantillon de sortie de rang pair étant égal à l'échantillon d'entrée de même rang réduit de la valeur approximative d'une somme d'échantillons intermédiaires de rang impair respectivement multipliés par des neuvièmes coefficients de filtrage (vk), - chaque échantillon de sortie (X2i+1) de rang impair étant égal à l'échantillon intermédiaire de même rang augmenté de la valeur approximative d'une somme d'échantillons de sortie de rang pair respectivement multipliés par des dixièmes coefficients de filtrage, caractérisé en ce que : - les huitièmes coefficients de filtrage (qk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients : [5/16, 0, -5/16], [-1/16, 6/16, 0, -6/16, 1/16] et [1/256, -16/256, 93/256, 0, -93/256, 16/256, -1/256]. Procédé de filtrage selon la revendication 25, caractérisé en ce que les neuvièmes coefficients de filtrage (vk) sont un ensemble choisi en fonction de la longueur du filtre parmi tous les ensembles suivants de coefficients: [1/16, 1/2, -1/16], [-3/256, 22/256, 1/2, -22/256, 3/256], [5/2048, -44/2048, 201/2048, 1/2, -201/2048, 44/2048, -5/2048]. Procédé de filtrage selon la revendication 25 ou 26, caractérisé en ce qu'il a un dixième coefficient de filtrage unique et en ce que ce dernier est égal à un. Procédé de filtrage selon l'une quelconque des revendications 18 à 27, caractérisé en ce que la valeur approximative est la fonction identité. Procédé de filtrage selon l'une quelconque des revendications 18 à 27, caractérisé en ce que la valeur approximative est une fonction d'une variable réelle qui fournit l'entier le plus proche de la variable. Procédé de filtrage selon l'une quelconque des revendications 18 à 27, caractérisé en ce que la valeur approximative est une fonction d'une variable réelle qui fournit l'entier inférieur à la variable. Procédé de filtrage selon l'une quelconque des revendications 18 à 27, caractérisé en ce que la valeur approximative est une fonction d'une variable réelle qui fournit l'entier supérieur à la variable. Procédé de filtrage selon l'une quelconque des revendications 18 à 27, caractérisé en ce que la valeur approximative est une fonction d'une variable réelle décomposée en des sous-variables dont la somme est égale à la variable, qui fournit une somme de valeurs approximatives des sous-variables, chacune des valeurs approximatives des sous-variables étant conforme à l'une quelconque des revendications 29 à 31. Dispositif de traitement du signal (10) comprenant un moyen pour mettre en oeuvre le procédé de filtrage selon l'une quelconque des revendications 18 à 32. Appareil numérique comprenant le dispositif de traitement du signal selon l'une quelconque des revendications 16, 17 ou 33. Appareil photographique numérique comprenant le dispositif de traitement du signal selon l'une quelconque des revendications 16, 17 ou 33.






IPC
A Täglicher Lebensbedarf
B Arbeitsverfahren; Transportieren
C Chemie; Hüttenwesen
D Textilien; Papier
E Bauwesen; Erdbohren; Bergbau
F Maschinenbau; Beleuchtung; Heizung; Waffen; Sprengen
G Physik
H Elektrotechnik

Anmelder
Datum

Patentrecherche

Patent Zeichnungen (PDF)

Copyright © 2008 Patent-De Alle Rechte vorbehalten. eMail: info@patent-de.com