FR2661062A2 - Method and device for compatible coding and decoding of television images with different resolutions - Google Patents

Abstract

In this method the signals representing images are transmitted on a transmission channel between at least one transmission coder and one reception decoder linked to a receiver. It consists: - in cutting up the image to be coded into sub-bands according to the same tree structure by filtering and decimating the spatial frequency band of the signals to be transmitted, whatever their resolution family - in independently coding the signals to be transmitted in each sub- band according to an intra-image coding mode or an inter-images mode (95). Application: standardisation of television systems.

Description

Method and device for encoding and decoding
compatible TV images of different resolutions
This certificate of addition is an improvement to the invention described in the main patent application 89 15175 filed in France on November 20, 1989 in the name of the Applicant.

The coding and decoding method as described according to claim 1 of the main patent application consists in:
- cutting each image to be transmitted into sub-bands according to the same tree structure by filtering and decimating the spatial frequency band of the signals to be transmitted whatever their family of belonging,
- to independently code the signals to be transmitted in each sub-band
- multiplexing the coded signals of the sub-bands before transmitting them on the transmission channel
and to be performed in the reception decoder
- a demultiplexing of the coded signals received in each sub-band,
- to decode the signals received relating to each sub-band as a function of the resolution of the signals transmitted from the receiver's own resolution.

 Applied directly to the coding of points located only inside the same image, this process unfortunately quickly loses its effectiveness when the bit rate on the transmission channel decreases. In fact, for bit rates lower than 1.4 bits per pixel, the visual quality of the images observed is no longer sufficient.

 The object of the invention described in the present application for an addition certificate is to take into account the natural temporal correlation of the image sequences in order to reduce the bit rate to approximately 0.8 bit per pixel without apparent degradation of the quality of the image. image, while maintaining compatible coding properties.

 To this end, the subject of the invention is a method of coding and decoding compatible with television images of different resolutions according to claim 1 of the main patent, characterized in that it consists in performing an inter-image coding based on the motion compensation in each of the sub-bands.

Other characteristics and advantages of the invention will become apparent from the following description given with reference to the appended drawings which represent
- Figures 1A and 1B different levels of resolution in spatial-temporal Fourier spaces
- Figure 2 a one-dimensional cell with two channels
- Figure 3A a separable two-dimensional decomposition cell with four bands
- Figure 3b a breakdown of the spectrum obtained by means of the decomposition cell shown in Figure 3A
- Figure 4A an optimal decomposition cell
- Figure 4B a breakdown of the spectrum obtained by means of the decomposition cell shown in Figure 4A
- Figure 5 a frequency division corresponding to a tree with seven branches
- Figure 6 an embodiment of a FIS compatible encoder (INTRA IMAGE)
- Figure 7 an embodiment of a compatible decoder for a receiver in VT format
- Figure 8A a simplified embodiment of an inter-image coder
- Figure 8B a graph to illustrate the operation of the quantizer used in the embodiment of Figure 8A
- Figure 9 an illustration of the positioning of a search window relative to a block of images whose movement is estimated using a line
- Figure 10 a quantization device implemented by the invention to code an intra-intra image with motion compensation
FIG. 11A an embodiment of an HDP sub-band coder according to a configuration in which inter-image coding with motion compensation is selected
- Figure 11B an embodiment of a decoder compatible in EDP format according to an inter-image configuration with motion compensation
- Figure 12 an inter-image compatible encoder
- Figure 13 an embodiment of an inter-image decoder of any sub-band.

 The method according to the invention which is described below is based on the same principle as the method already described in the main patent application, it differs from it by the fact that the tool takes into account the temporal correlation of the image sequences in introducing additional inter-picture coding.

 As already indicated in the main patent application, a decomposition into sub-bands of the two-dimensional spectrum of the image, considering only the progressive formats, can be envisaged in many ways depending on the desired form of the sub-bands obtained by rectangular, staggered, hexagonal, etc. decimations, according to the hierarchical or parallel structure used of the filter bank and finally according to the decomposition filters used, whether these filters are with finite or infinite impulse response, separable or not. , for both practical and theoretical reasons, notably those caused by the need to be able to easily draw sub-band filters ensuring perfect, or almost-perfect reconstruction of the original signal in the absence of any quantization of sub-band signals, but also those due to the desire to obtain a modular system allowing an easier search for subband decomposition optimal as a function of the desired bit rate, it is preferable to choose a system with a hierarchical structure which alone makes it possible to show resolution levels nested one inside the other. This allows the spectrum of the image to be cut out at will from a single specification of an elementary synthesis analysis cell. The advantage is that several elementary cells can then be arranged in a tree structure and that certain parts of the spectrum can thus be more or less broken down according to their informational content. Finally, to reveal rectangular frequency bands required by the constraints of coding compatibility between the various video standards currently considered HDP, HDI, EDP, TV and VT and the spectra of which are shown in FIGS. 1A and 1B, an approach using separable filters is preferable to any other because it allows the rows and columns of the image to be treated separately using one-dimensional filters. As already described in the main patent application, the interlaced images can be associated with equivalent progressive images having a rectangular band spectrum, and the spatial spectrum of the image can then be cut according to a tree structure with four frequency bands and the design of the analysis-synthesis system can then be based on that clearly simpler of a system two-channel synthesis analysis analysis. A one-dimensional cell comprises, as shown in FIG. 2, two low-pass and high-pass filters 11 and 12 respectively of impulse responses HOn and Hln and two interpolation filters 17 and 18 of impulse responses GOn and Gln coupled together by devices 13 and 14 for undersampling and 15 and 16 for oversampling.

 The separable two-dimensional four-band decomposition cell which is represented in FIG. 3A comprises an assembly formed by the cascading of three one-dimensional cells according to a two-stage tree. The first stage formed by filters 20 and 21 of transfer functions HOn and Hln filters the columns of an image and the second stage formed by filters 240 and 243 of transfer functions HOn and Hln filters the lines of two sub-images bands obtained from the first stage. The cascading of such cells therefore makes it possible to cut the spatial image spectrum according to a "quadtree" tree structure with four frequency bands, according to which the hierarchy of the initial image is first cut in 4 sub-bands then some of the sub-bands are cut and so on with the possibility of using truncated trees to cut for example more the low frequencies than the high frequencies. In the embodiment of FIG. 2, the filters 11 and 12 must give a perfect reconstruction or at least greater than 50 decibels of the signal Xn applied to their input. For their realization the choice of filters with finite impulse response is preferable because they can have a strictly linear phase, and they make it possible to avoid the numerical problems (stability in particular) in finite arithmetic. Under such conditions a good choice of filters 17 and 18 eliminates inter-band aliasing and any pair of solutions on the choice of filters 11 and 12 can be obtained by factorization of a half-band filter. However, only part of this set of solutions is really usable for an application such as coding which imposes an imperative condition concerning the quantization of the sub-band signals y0, Y1 obtained at the outputs of the sub-sampling devices 13 and 14 namely that the variance of the reconstruction error is equal to the sum of the variances of the quantization noises of each channel. This condition, obviously added to the requirement of perfect reconstruction of the original signal without quantization, is only satisfied with filters known by the abbreviation "CQF" from the English term "conjugate quadrature filters", according to which the impulse response of the filter H1 is obtained from the impulse response of the filter HO by time inversion and change of sign of a coefficient on two. In addition, the condition for perfect reconstruction requires that the filter 20 HO has an auto-correlation function which is a half-band filter. These filters necessarily have an even length and can be determined for practically any spectral mask. Their main disadvantage is that they cannot have a linear phase as soon as their length is greater than 2, nevertheless a good pairing of zeros, during factorization of the resulting product filter, can give an almost linear phase. Another condition may be a little less fundamental to check during the production of the filters is that of the decorrelation of the sub-band signals YO and Y1 obtained at the output of the sub-samplers 13 and 14, so as to be able to code the sub- bands independently of each other with performances close to the optimal.Nevertheless, this condition is too strong to satisfy however if the transfer functions HO and H1 of filters 11 and 12 form a pair of filters CQF, i.e. - say if the transfer function H1 is obtained from the transfer function HO by time inversion and change of the sign of a coefficient on two and if the filters 11 and 12 have even lengths and are in linear phase then we obtain a point-to-point correlation of the signals YO and Y1. The resulting filters are known by the abbreviation QMF "quadrature mirror filter" but can however only make an approximate reconstruction of the original signal. However, the signal-to-noise reconstruction ratios obtained are greater than 50 decibels as soon as the filters have sufficient lengths. Furthermore, no phase distortion is provided since the transfer function of the overall synthesis analysis system is then in phase linear. The choice of a filter of one or the other type can only result from a compromise between good reconstruction and inter-band decorrelation. The intra-band decorrelation is determined by the choice of the decomposition tree, by an arrangement of four-band two-dimensional cells. Naturally, the more a band is cut, the more the correlation inside it decreases, but moreover the longer the corresponding reconstruction filter, which involves a risk for the quantization noise to propagate in the reconstructed image. A good compromise is represented by the 16-band decomposition tree described in FIG. 4. Whatever the image to be cut out, this tree gives good intra-band decorrelation. Once decomposed, each band is linearly quantified, the quantization step of each band having to be a function of the number of cuts it has undergone in order to obtain an optimal quantification in the sense of the minimum variance of reconstruction terror: more precisely, the step of quantization of the band i must be proportional to 4-ki where ki is the number of times that the band i has been cut into four bands. Thus the bands located on the same floor of the decomposition tree are quantified with the same step and an additional cut into four bands divides the step by two. After quantization each band is coded using variable length codes defined as described in patent application 2 627 337 filed in the name of the Applicant, with a description in the range of zero. However in the baseband (the one that has always been filtered by a low-pass filter) the use of fixed length codes is preferable since its statistics are very dependent on the original image.

In the intra-compatible coding scheme described in the main patent application, the interleaved formats are taken into account by using their equivalent progressive signal. Any interlaced signal is first transformed into an equivalent progressive signal by a deinterlacer and then decomposed like any progressive signal, apart from the first decomposition into two horizontal bands which is not performed. In what follows, therefore, interlaced signals are no longer considered and only the progressive formats HDP, EDP and VT are taken into account, however the same process can easily be extended to the case of interlaced signals HDI and TV by considering their associated progressive signal and by proceeding as in the main patent application by incorporating a deinterlacer in the coder and an interleaver in the decoder in the case of an interlaced receiver. The intra-compatible coding which is carried out is a type scheme known under the designation FIS (independent format spliting) according to which the signal to be coded is always decomposed in the same way whatever its resolution. Therefore, for example to ensure compatibility at three levels, cutting into at least seven strips according to the representation in FIG. 5 is necessary. The corresponding tree also allows compatibility with HDI interlaced formats and
TV. Naturally, coding and independent transmission must be ensured, that is to say separated by inimitable synchronization words, groups of bands (1, 2, 3) [part
HFC], (5, 6, 7) [part IFCj and band 4 [part BB].

An embodiment of a corresponding compatible encoder is shown in Figure 6A. This coder comprises a BBC coding block 29, an IFCC coding block 30 and an HFCC coding block 31. The outputs of coding blocks 29, 30 and 31 are respectively connected to the corresponding inputs of a multiplexer circuit 32. The chopping in sub-bands is obtained by means of low-pass filters referenced respectively 32 to 37 coupled to sub-sampling circuits not shown and by means of high-pass filters referenced respectively from 38 to 43 also associated with operators of sub- sampling not shown. The coding device of FIG. 6A allows, in addition to these good performances, to ensure complete transparency between the different HDP sampling formats,
HDI, EDP, TV and VT described above, regardless of the BBC, IFCC and HFCC coding blocks. I1 also makes it possible to obtain a sixth format which corresponds to the definition of the interlaced telephone video signal. The signal to be coded whatever its HDP, EDP or VT format is applied to the input marked IP of the device and is divided into seven sub-bands referenced from S1 to S7 according to a decomposition tree similar to the diagram in FIG. 5. The sub-band S1 is obtained through the series of filters 42 and 43. The sub-band S2 is obtained through the series of filters 37 and 43, the sub-band S3 is obtained through the series of filters 41 and 32, the sub-band S4 is obtained through the series of low-pass filters 32 to 35, the sub-band S5 is obtained through filters 36, 40, 33 and 32, the sub-band
S6 is obtained through the series of filters 38, 34, 33 and 32 and the S7 sub-band is obtained through filters 39, 40, 33 and 32. Whatever the type of coders BBC, IFCC and
HFCC, compatibility properties are checked. In fact they constitute additional cuts of the sub-bands S1 to S7 so as to obtain the optimal tree with 16 bands represented in FIG. 4. Each of the bands which is then obtained is followed by a linear quantizer Qi and a VLC variable length encoder. as shown in FIG. 6B which represents an embodiment of the BBC encoder 29. Thus in FIG. 6B the sub-band S4 is cut again into two sub-bands by a low-pass filter 45 and a pass-filter high 46. The sub-band obtained at the output of the low-pass filter 45 is again divided into two sub-bands respectively by a low-pass filter 47 and a high-pass filter 48. The resulting sub-bands are then quantified respectively by quantization circuits 49 and 50, the quantized signals obtained are then coded using variable length coding devices 51 and 52. Similarly, the sub-band obtained at the output of the filter 46 is itself divided into two sub-bands by a low-pass filter 53 and a high-pass filter 54. The sub-bands thus obtained are also quantized by two quantization circuits 55 and 56 and the quantized signals obtained are coded by length coding devices variable 57 and 58. The outputs of the variable-length coding devices 51, 52, 57 and 58 are respectively connected to the inputs of a multiplexing circuit 59 whose output is connected to a first input of the multiplexer circuit 32 of FIG. 6A. Similarly, the IFCC 30 coder only performs quantization and variable length coding of the bands S5,
S6 and S7. Finally, the HFCC 31 encoder cuts each of the bands
S3 and S2 in 4 additional bands. The S1 band remains as it is. For interlaced signals the decomposition procedure is not quite the same. Indeed, to avoid having to carry out a three-dimensional decomposition involving a staggered decomposition in the plane of the temporal and vertical frequencies which could seriously compromise the compatibility properties, an interlaced signal is first transformed into an equivalent progressive signal. The equivalent progressive signal is obtained by placing the interlaced signal which is applied to the input IE of the device in a deinterlacer 44 in FIG. 6A, with the aim of vertically phase-shifting by half a pixel each odd frame, this is done simply at by means of a vertical one-dimensional interpolator filter. This produces a progressive signal with twice as few vertical points as the progressive signal of the same horizontal definition, i.e. corresponding to the low frequency band of the first cut. into two bands of the encoder (cut into two horizontal bands by filtering along the columns), in this way the signal obtained at the output of the deinterleaver 44 is applied to the point of connection common to the low-pass filters 32 and 33. This signal is then decomposed in the same way as all progressive signals but naturally in this case it only leads to five sub-bands since the sub-bands S1 and S2 cannot be formed . In the embodiments which follow, the optimal 16-band coder which has just been described will be used.

An embodiment of a corresponding decoder is shown in Figure 7 in the case of a VT receiver. This decoder performs the inverse functions of the coder of FIG. 6A. The complexity of this decoder depends essentially on the speed of the associated work. It is formed similarly to the decoding device corresponding to the decoding tables in FIG. 13 of the main patent, by a set of decoding blocks numbered from 60 to 65 receiving from a transmission channel 66 the digital frame of the video signals at through a multiplexer circuit 67 and a logic switching block 68 connected in series. Decoding blocks 60 to 62 are blocks
BBC 1, the decoding blocks 63 and 65 are IFCC 1 decoding blocks and the block 64 is an HFCC 1 decoding block.

A set of summing circuits and low-pass and high-pass filters referenced from 69 to 86 ensures, as represented by the tables in FIG. 13 of the main patent application, the reconstruction of the telephone video signal and its application to the corresponding receiver referenced 87 in FIG. 7. In FIG. 7 the logic block 68 applies the useful part of the binary train received on channel 66 to the appropriate inputs of the decoder 60 to 65. Under these conditions the decoders associated with the receiver in EDP format and HDP is easily deduced. In the case of an EDP decoder the decoder 60 for example must be removed and in the case of a decoder
HDP it is necessary in addition to remove the decoders IFCC-1 63 and 65.For interlaced decoders the principle remains the same and it is necessary in this case to remove from the VT decoder the inputs which are never used according to table 13 of the patent application main. Finally, at the level of the decoder for a given receiver format, some of the hardware is not used if the transmitted image has a higher format, however the decoders will always rotate at the same speed whatever the resolution of the transmitted image. . This speed is that of the receiver associated with the last reconstruction stage and decreases towards the input stages since it is halved on each stage.

The intra-image coding scheme which has just been described unfortunately loses its effectiveness as soon as the transmitted bit rate decreases, the visual quality obtained with this type of coding no longer being sufficient. The device which is described below makes it possible to take into account the temporal correlation of the image sequences so as to be able to descend to still much lower bit rates of the order of 0.8 bit per pixel. The problem is solved by the use of a hybrid diagram by means of a temporal prediction allied with a compensation of the motion estimated in the image. Although this scheme is conventional and is already applied for example for cosine transform coders, the originality here resides in its use in the sub-band switching scheme described above while ensuring the properties of compatible encoding. simpler that is (INTER PUR) is shown in Figure 8A. It consists according to a differential coding method to be removed from the pixel to be coded x (k, l) of the image at time n the coded value of the pixel opposite in the previous image at time nl and at then code this difference. It comprises, as shown in FIG. 8A, a quantizer 88 coupled on its input to the output of a subtractor circuit 89. The output of the quantizer circuit 88 is coupled to the transmission channel and to a dequantifier circuit 90 whose output is connected to an adder circuit 91. A delay circuit 92 connects the output of the circuit 91 to a first input of the subtractor circuit 89 and to a second input of the adder circuit 91. Following this diagram the pixel x (k, 1, n) of address k, l in image n is applied to the encoder input on the second input of subtractor circuit 89. Circuit 89 subtracts from the value of the pixel x (k, l, n) in the image at instant n the decoded value of pixel x (k, l, n-1) at instant nl. The difference d (k, l, n) obtained is under these conditions equal to
d (k, l, n) = x (k, l, n) -x (k, l, nl)
In this diagram it is the coded value of the corresponding pixel in the previous image which is taken into account to avoid any phenomenon of drift of the decoder, that is to say so as to do temporally exactly the same thing to the coder as to the decoder.

The quantizer 88 associates with the difference d (k, l, n) a value d (k, 1, n) which is in fact a quantization level number. The quantization is carried out using the dequantifier 90 associates with this level number its corresponding reconstruction value which is an approximate value of the difference d (k, 1, n) marred by the quantization noise. The reconstruction value d (k, l, n) which is obtained at the output of the dequantizer circuit 90 is represented in FIG. 8B as a function of the difference d (k, l, n) calculated by the subtractor circuit 89. is clear that it is the prediction value x (k, l, nl) which must be used and not the value x (k, l, n-1) in order to be able to recalculate the prediction value x (k, l, n ) at the decoder. In this case at the level of the decoder the recalculated value x (k, l, n) is equal to x (k, l, nl) + d (k, l, n) and only the quantization error made on the difference d (k, l, n) intervenes. Indeed, the coding error is equal to x (k, l, n) -x (k, l, n) or even e (k, l, n) = d (k, 1, n) + x ( k, l, n-1) -x (k, l, n) or d (k, l, n) -d (k, l, n-1)
By using the value of the pixel x (k, l, n-1) as a predictor the previous coding errors of the pixels x (k, l, i) for i <ns would accumulate creating a drift phenomenon.

In fact, a local decoder is obtained here at the very level of the coder. The diagram which has just been described being very unsophisticated there is a risk that this diagram will lose its effectiveness when there is a little movement in the sequences of images, the prediction becoming very bad and therefore giving a very large prediction error much more difficult to code. This risk can however be minimized by improving prediction by introducing motion estimation and image registration. Thus to improve the prediction the method according to the invention consists in introducing an estimation of movement and a registration of the images. This method consists instead of removing the opposite pixel decoded from the previous image (x (k, l, n-1)), to remove the pixel x (k ', l', n-1) of coordinates k 'and l calculated according to the local movement.

A corresponding device making it possible to obtain this result is represented in FIG. 10. It comprises, in addition to the elements of FIG. 8A which are represented with the same references, a device 93 for forming blocks, a motion estimator 94, a device d inter-intra image referral and a motion compensation device 96. The device allows a simple quantification of an image associated with an inter scheme or an intra scheme with motion compensation. The motion estimator device 94 makes it possible to estimate the movement between two successive images by executing the known "block-matching" algorithm which makes it possible to calculate an intercorrelation between two successive images and estimates the corresponding motion vector from the correlation peak obtained. Each image is divided by the blocking device 93 into disjoint blocks of dimension NxN and completely covering the image and a motion vector is estimated per block. In this calculation the number N is assumed to be even. For each block B (k, 1, n) as shown in FIG. 9, the nearest block B (k ', l', nl) within the meaning of a certain criterion and the motion vector estimated for each of the points of the block B (k, i, n) is given the coordinates k'-k, l'-l. The search window used has a range of 2M. 2L and is centered on block B (k, i, n). The criterion used is an energy criterion consisting in carrying out a sum of difference squared or a sum of difference in absolute value. The registration of the movement which is carried out by block 96 is obtained by searching for each block of the current image the corresponding block which in the previous decoded image must be used as prediction. When the components of the motion vector are whole, that is to say correspond to an integer number of pixels, the problem is reduced only to a memory addressing problem. On the other hand, when the motion vector is not integer, its values must be interpolated as a function of the neighboring pixels (bilinear interpolations can be used). The choice at the level of each block between the intra mode and the inter mode is made by block 95. This choice is based on an energy criterion. It consists in calculating the energy of the block to be coded intra and the energy of the prediction error block (inter block) supplied by the subtractor 89. This choice consists in selecting the mode (inter or intra) which gives the minimum energy. In the embodiment of FIG. 10, the delay element 92 is constituted by an image memory. This memory contains the previous decoded image and it is this image which is generally used to carry out the motion estimation. But it is also possible to use the previous image without coding.

 The subband decomposition is very well suited to the context of motion estimation and compensation since the geometric properties are preserved in the subband images and the motion estimation and compensation can be done as well level of the full-band image, that is to say before the decomposition into sub-bands, than at the level of the sub-bands. However, the constraints of compatible coding require first of all that the motion compensation be carried out at the level of the sub-bands and not at the level of the non-decomposed image. Indeed, in the case of motion compensation carried out at the level of the non-decomposed image, the processing operations at the level of the coder and of an associated compatible decoder (i.e. decoding only part of the transmitted image) would be different and a phenomenon of drift would inevitably occur. This case of compensation of the movement at the level of the non-decomposed image is illustrated in FIGS. 11A and 11B and we will highlight below the phenomenon of drift that appears. To simplify the description, the operation of the coder and of the decoder represented in FIGS. 1iA and 11E is explained below, assuming that the decoder represented in FIG. 11E is the associated EDP compatible decoder which decodes the EDP part of the HDP signal supplied. by the coder of FIG. 11A. In FIG. 1iA, the elements homologous to the coding device represented in FIG. 8A are represented with the same references. Unlike the elements represented in FIG. 8A, it comprises two quantizers 88a and 88b for quantifying respectively the parts relating to the EDP signal and the complement of the HDP signal and two corresponding decantifiers 90a and 90b. They also include an HDP 97 sub-band decomposition device and an HDP 98 sub-band reconstruction device. The delay device 92 consists of an image memory.

A motion compensation device 96 is also introduced as in FIG. 10 between the image memory 92 and the subtraction circuit 89. The motion compensation device 96 receives motion vectors vx and vy from an estimation device motion shown in Figure liA having characteristics identical to the device 94 shown in Figure 10. The HDP sub-band decomposition device 97 is interposed between the output of the subtractor circuit 89 and the inputs of the quantization circuits 88a, 88b. The HDP 98 sub-band reconstruction device is interposed between the outputs of the dequantization circuits 90a, 90b and the input of the summing circuit 91. Following this embodiment, the signal d (k, l, n) obtained at the output of the subtractor circuit 89 is equal to the difference of the signal x (k, l, u) applied to input "+" of the subtractor circuit 89 and of the reconstructed signal x (k + vx, l + vy, n-1) supplied by the motion compensator block 96, vx and vy being the components of the estimated motion vector provided by the motion estimator 94. After reconstruction in sub-bands performed by the reconstruction block 98 the signal d (k, l, n is recovered) ) which in fact represents the signal d (k, l, n) affected by the quantization errors of the sub-band signals. The reconstructed signal x (k, l, n) equal to the signal x (k + vx, l + vy, nl) + d (k, l, n) is applied to the input of memory 92 by the output of the circuit adder 91.

The global quantification error is given by the relation e (k, l, n) = x (k, l, n) -x (k, l, n) let again be
e (k, l, n) = d (k, l, n) + x (k + vx, l + vy, n-1) -x (k, l, n) let e (k, l, n ) = d (k, 1, n) -d (k, l, n)
The previous relation shows that the global quantification error is identical to the quantization error committed on the prediction error and that there is no drift phenomenon. The corresponding EDP compatible decoder which is represented in FIG. 11B comprises connected to the transmission channel 99, a dequantifier 100 coupled on its output to an EDP sub-band reconstruction device 101. It also comprises a motion compensation device 102 and an image memory 104. An adder circuit 103 is coupled by a first operand input to the output of the EDP sub-band reconstruction device 101 and by a second operand input to the output of the motion compensation device 102. The output of the adder circuit 103 is connected to the input of the image memory 104 and provides the reconstituted EDP compatible signal xc (k, 1, n). In the decoder of FIG. 11B, the device 101 for reconstructing sub-bands
EDP provides the filtered, sub-sampled and quantified part dc (k, l, n) of the signal d (k, l, n). The reconstructed signal xc (k, l, n) is obtained at the output of the adder circuit 103 by adding the signal dc (k, l, n) and the signal xc (k + vx / 2,1 + vy / 2, n -1) supplied by the output of the motion compensation device 102.

In this example it is assumed that the motion vectors vx / 2 and vy / 2 are integers so as not to take account of the interpolation which would be necessary in the opposite case.

The coding error committed with respect to the signal xc (k, l, n) (which represents the EDP part of the signal x (l, l, s)) is equal in these conditions to ec (k, i, n) = xc (k, l, n) ~ XC (k, l, n) let ec (k, l, n) =
Xc (k, l, n) -x
c (k + vx / 2, l + vy / 2, n- 1) -dc (k, I, n)
As soon as the motion vector is non-zero the preceding relation shows that the error is not only due to the quantification of the error of prediction of the signal dc (k, i, n) since the motion estimation is not not done at signal level
EDP. And there is therefore the introduction of a recursive error term which creates the drift. The previous example gives a major reason to perform motion compensation at the level of the sub-bands rather than at the level of the full-band image. .

A second reason comes from the fact that a choice of inter or intra mode is taken for each block. So if the motion compensation and the inter / intra choice were made on the non-decomposed image, it is a set of inter and intra disparate blocks that should be cut into sub-bands. However, the artificial high frequencies introduced at the borders between inter and intra blocks would create rebounds in the sub-band images thus making them much more difficult to code and under these conditions the diagram would lose its effectiveness.

Of course a possible solution could consist in making the inter-intra choice at the level of the sub-bands, while making the motion compensation on the full band image, by compensating in motion the whole image and by decomposing into sub-bands. both the intra image and the compensated image. The inter-intra decision would then be taken in the sub-bands on the homothetic blocks corresponding to the motion estimation blocks on the full band image. But such an approach proves to be very costly in terms of the hardware implementation of the coders and decoders since two decompositions into sub-bands must be carried out. On the other hand, motion compensation by block would risk creating certain artificial boundaries in the compensated image, due in particular to the non-uniformity of the field of motion estimated, thus bringing problems of rebound in the sub-bands, and the inter-intra choice could thus be biased in favor of the intra. It is therefore preferable to carry out the motion compensation at the level of the sub-bands and this is what has been retained at the level of the invention. In this case, the motion estimation can be made both at the level of the full-band image and at the level of the sub-bands.

However, since the physical movement is the same in each subband image it is better to use a single set of motion vectors rather than to estimate the motion in each band. The latter solution moreover brings a certain redundancy between the motion vectors of each band and a fairly sophisticated coding of these vectors is necessary so as not to lose too much bit rate during their transmission. By carrying out the motion estimation at the level of the full band image, this can be done either between two successive source images where a source image and the previous decoded, the first solution giving a more physical movement but may be less well suited for coding as the second. Furthermore, estimation at the full band level makes it possible to obtain better precision at the level of the motion vectors. An embodiment of corresponding coders is represented in FIG. 12. This coder contains as many coding devices homologous to that represented at Figure 10 that there are sub-bands. In FIG. 12, each homologous coding device comprises the same elements as those in FIG. 10 assigned the same references with an index i corresponding to the sub-band. After decomposition of the original image into sub-bands (filter bank 93). The sub-bands are transmitted to each coding device using a block forming device B. and an estimation device of motion 94 is coupled to a device for scaling motion vectors 105 to apply the motion parameters respectively to each motion compensation device 96. The reconstruction device 106 is connected to the outputs of memories 92. to reconstruct 1 starting image and apply it to the motion estimation device 94. Variable length coding devices 107.

are coupled at the output of the quantization devices 88. and a multiplexer circuit 108 coupled at the output of the variable-length coding devices 107. to transmit the result of the coding on the transmission channel through a buffer memory 109.

 An embodiment of the corresponding decoder for any of the bands is represented in FIG. 13, it comprises connected in this order, a device for decoding variable length codes 110, a decoding device 111, a device for selecting intra modes or inter-images.

In the intra mode, the decoded information is directly transmitted to the reconstruction filter bank, not shown. In contrast, in the inter operating mode, the information is transmitted to a first input of an adder circuit 113. The second input of the adder circuit 113 is connected to the output of a motion compensation device 114 connected to the output of an image memory 115 and controlled by the movement information received from the transmission coder. The memory 115 is connected by its input to the common output of the intra-inter image selection device 112 and to the output of the adder circuits 113. According to the embodiments of the coder and the decoder which have just been described, the images are divided in blocks of NN A motion vector is estimated per block and is transmitted to the reception decoder. Variable length coding can be considered for motion vectors. These vectors are scaled according to the sub-sampling factor, by dividing each component by this factor, in order to be able to use them in each band where the motion compensation is carried out on the blocks corresponding to the estimate. These "corresponding blocks" were scaled by homothety in the same way as the vectors (size divided by the sub-sampling factor). It should however be noted that the phase shifts introduced by the filtering must be taken into account in the localization of the corresponding blocks. In the case of using linear phase filters of length L a length shift (L-1) / 2 is introduced in the filtered image.

This offset must be compensated. However, given that the length L is even there is always a phase shift of + pixel which is negligible all the more since it is reduced to i of pixel with the sub-sampling. This compensation is simply done using a delay line. If the precision of the motion vectors estimated at the level of the full-band image is 1 pixel (components estimated as a whole number of pixels) then the precision of the corresponding vectors is i in band 1, i in bands 2 up to 12, and 1/8 in bands 13 to 16 assuming of course the use of the 16-band decomposition tree shown in Figure 4. A generally bilinear interpolation is therefore necessary to perform the compensation. Furthermore, N, the size of the motion estimation blocks at the level of the full-band image can be of the form k. 21 and 1 must be at least equal to 4 in order to obtain compensation blocks larger than 2x2 at level of the lowest bands. Generally k can be fixed equal to 1. Finally for each block and for each band an inter-intra decision is taken according to an energy criterion.

However it is especially at the low frequency bands that the inter mode seems the most useful. In FIG. 12 the memories 92. contain decoded sub-bands from previous instants and can therefore allow via the reconstruction filter bank 106 to create the previous decoded image (decoder local to the coder) which can be used for motion estimation performed by block 94. The separation of motion compensation blocks 96. by strip is very advantageous from the hardware point of view since it allows processing to be carried out in parallel. Smaller memories running at lower speeds are needed compared to full band image motion compensation. Similarly, the interpolations can be done at reduced speed. With regard to the compatible coding-decoding process, everything takes place as in the case of the intra schemes described in the main patent application and in the table represented in FIG. 13 of the main patent application.

Claims (15)

 1. A method of coding and decoding compatible with television images of different resolutions according to claim 1 of the main patent, characterized in that it consists in also carrying out an inter-image coding (95i '88i, 67i) in each of the sub -bands.  2. Method according to claim 1 characterized in that it consists in carrying out the inter-image coding in carrying out an estimation of the moving points (94) at the level of the full-band image, before carrying out compensation for this movement in each sub-band.  3. Method according to any one of claims 1 and 2 characterized in that it consists in carrying out a differential coding (89i) of the points located opposite the current image and the previous image.  4. Method according to any one of claims 1 to 3 characterized in that it consists before decomposing the image into sub-bands  - to cut the image into blocks (93)  and determining for each current block a motion vector by means of a prediction block obtained by comparing each current block with its counterpart inside a window in the previous image.  5. Method according to claim 4 characterized in that the cutting of the image into sub-bands is obtained by means of filters (20, 21, 24i) associated with sub-sampling circuits (22, 23, 25 .. . 27) and in that the phase shift introduced by the filters is compensated to allow the localization of the blocks before compensating for the movement in each sub-band.  6. Method according to any one of claims 4 and 5 characterized in that it consists in making correspond to each block of the full-band image a homothetic block in each sub-band image.  7. Method according to any one of claims 2 to 6 characterized in that the estimation of the moving points (94) at the level of the full band image provides motion vectors for each block of the full band image and in what the motion vectors are divided according to the sampling factors used to obtain the sub-bands considered, before being applied to the blocks homothetic to the blocks of the full-band image.  8. Method according to any one of claims 4 to 7 characterized in that it consists in choosing (95) the intra-image coding mode or the inter-image coding mode by calculating the energy of the current block and the energy difference between the current block and that of the predicted block to choose the coding mode whose energy is minimal.  9. Method according to any one of claims 1 to 8 characterized in that the coding takes place on interlaced or progressive signals, the coding of the interlaced signals taking place after they have been transformed into equivalent progressive signals.  10. Method according to claim 9 characterized in that it consists in cutting each progressive signal into the same determined number of sub-bands whatever its format.  11. Method according to any one of claims 1 to 10 characterized in that the estimated movement vectors are scaled (105) before effecting the compensation of the movement in each sub-band.  12. Device for implementing the method according to any one of claims 1 to 11 characterized in that it comprises for each sub-band an inter-intra coder (88il 107i) image controlled from a device d motion estimation (94).  13. Device according to claim 12 characterized in that each coder of a sub-band comprises a quantizing device (88) coupled to a differential coding device via a switching device (95) to quantify each pixel supplied by a block directly when the intra-image coding mode has been chosen or quantify its difference with respect to its counterpart corrected by motion vectors in the previous image when the inter-image coding mode has been chosen.  14. Device according to claim 13 characterized in that each differential coding device is coupled to a motion estimation device (94) of the points of the image by means of a motion compensation device (96i) .  15. Device according to claim 14 characterized in that each motion compensation device (96i) is coupled to the motion estimation device (94) by means of a scaling device (105) of the motion vectors provided by the motion estimation device (94), so that motion estimation is made at the level of the non-decomposed image while motion compensation takes place on the sub-bands.