Classifications

• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
  • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
  • H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
  • H04N19/177—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a group of pictures [GOP]
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
  • H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
  • H04N19/62—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding by frequency transforming in three dimensions
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
  • H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
  • H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
  • H04N19/61—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding in combination with predictive coding
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
  • H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
  • H04N19/63—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding using sub-band based transform, e.g. wavelets

Definitions

• the present invention relates to a method of coding and / or decoding groups of images.
• the invention applies in the context of video compression based on decomposition by 3D wavelets.
• the movement is described by blocks, 16 pixels wide by 16 pixels high, for example.
• the problem raised above is then returned to the boundaries of the blocks.
• the trajectories of the pixels in the two directions of travel of the image sequence have also been taken into account: in the natural direction of the GOP image sequence and in the opposite direction. But then we double the amount of movement information compared to conventional techniques.
• the present invention consists in choosing configuration data for each sequence and an optimizing criterion making it possible to reduce the number of unconnected pixels in each sequence of GOP images.
• the number of unconnected pixels is significantly reduced without requiring more processing according to a 3D wavelet decomposition.
• the present invention relates to a method of coding and / or decoding image sequences, on the basis of a 3D wavelet processing.
• the method of the invention comprises a coding phase which includes: a step for determining at least one piece of configuration information for the 3D wavelet coding such as the selection of a reference image for a sequence of successive images, based on a predetermined optimizing criterion; then - a step for generating a unit of configuration information comprising the datum of the reference image for said sequence of images; and finally - a step for executing the 3D wavelet coding on the basis of said configuration information, so as to produce a coded data stream comprising the meeting for each sequence of images of a unit of configuration information and / or 3D wavelet coding data of the images in the image sequence.
• the step for determining at least one configuration item of information comprises a wavelet decomposition step in at least one test image as a reference image, then a motion estimation step, then a step determining the image of the sequence of images for which the number of pixels not connected during the motion estimation step is minimal, as a reference image.
• the step for determining at least one configuration item of information comprises a step of determining the amplitude of the movement in at least one test image, then a step of determining the image of the sequence images for which the amplitude of movement in the image is minimal, as a reference image,
• the step for determining at least one configuration item of information comprises a step of determining texture information, by measuring the high-frequency components, in at least one test image, then a step of determination of the image of the sequence of images for which the texture information is the richest, as the reference image.
• the step for determining at least one configuration item of information also includes determining the number of images in said sequence of images. According to one aspect of the invention, the step for determining at least one configuration item of information also comprises the determination of at least one direction of temporal decomposition for a sequence of images.
• the step for determining at least one configuration item of information consists of:
• the step for determining at least one configuration item of information consists of:
• - in a first step to constitute a sequence of GOP images of a predetermined size
• - in a second step to execute the choice of a reference image, and in that the direction of temporal decomposition is determined in the forward direction for the images of the sequence of images which follow the selected reference image and in the opposite direction for the images of the sequence of images which precede the selected reference image.
• the step for determining at least one configuration item of information comprises:
• a step of estimating the movement a step of constructing the trajectories of the points of movement in the estimation of direct movement and a step of calculating the number of pixels not connected during the step of construction of trajectories;
• a step of estimation of the movement in the opposite direction between images a step of construction of the trajectories of the points of motion in the estimation of direct motion and a step of calculation of the number of unconnected pixels during the stage of construction of the trajectories; a step of choosing the best reference image on the basis of the lowest number of unconnected pixels.
• the step for determining at least one configuration item of information comprises:
• a loop is made for each image of the GOP sequence being adapted during which successively a step of constructing the trajectories is executed, then a step of calculating the number of unconnected pixels; then
• a step of selecting the best reference image in the sequence of GOP images being adapted is carried out on the basis of the lowest number of unconnected points.
• the method comprises a decoding phase which consists:
• the method is integrated into a 3D wavelet video compressor-decompressor scheme.
• the method is integrated into an MPEG TM coding-decoding scheme.
• the selection of a reference image, from which the movement is calculated in a GOP, the direction of the temporal decomposition are carried out according to the number of pixels connected by this movement, information which can be combined with the data of texture or motion of the motion vector field.
• the efficiency of the 2D + T wavelet decomposition is improved with the number of pixels connected in time.
• FIG 1 there is shown an embodiment of a 3D wavelet compression scheme according to the prior art. It will be noted that this 3D wavelet compression scheme can also be used once the method of the invention has been executed. It will also be noted that what will be described in terms of method, could be implemented in the form of processing blocks in a computer for processing image sequences.
• a step 1 the acquisition or the synthesis of a sequence of GOP images, comprising N successive images, is carried out.
• the GOP sequence is then transmitted to a decomposition step by wavelets 2 and in parallel to a step 3 for estimating the movement between images.
• the wavelet decomposition step includes a step 2a of analysis in the time domain and a step 2b of analysis in the space domain (pattern repetition frequencies, etc.)
• the step of estimating the movement 3 begins with the choice of a basic image in the sequence of GOP images which is the last image of the sequence. By differences between images, between image_2 and image_1, then between image_3 and image_2, etc. an estimate of the movement is then made.
• motion estimation information is transmitted by motion estimation information 8 in step 2a of analysis in the time domain for the wavelet decomposition to perform a one-dimensional decomposition in wavelets in the direction of movement.
• a sequence of transformed images corresponding to different time frequencies is produced.
• the sequence of images transformed into time frequencies is then processed during step 2b of spatial decomposition by a 2D wavelet transform.
• the movement coefficients representative of the motion vectors in the sequence of images GOP is transmitted to a step of coding the vector fields representative of the movements in the sequence d 'GOP images.
• coefficients representative of the spatial frequencies in each image of the sequence of images are transmitted to an input of an entropy coding circuit.
• the entropy coding carried out in step 4 comprises two phases: preliminary modeling and coding proper. Modeling allows you to assign probabilities to each symbol in the data stream to be processed. The coding assigns to each symbol a binary word whose length depends on the associated probability. To this end, an encoding scheme such as the Huffman code or an arithmetic encoding is used.
• the coding coefficients from the entropy coding 4 are transmitted to a first input of a data flow conformation processing circuit 6, a second input of which receives the components of the motion vector fields from step 3.
• the set thus constitutes the data stream of the 3D wavelet coefficients, serialized so as to produce a binary data stream 7 to which any subsequent suitable processing can be applied, such as encryption, storage or transmission processing.
• FIG. 2 a diagram has been shown which illustrates a one-dimensional transformation technique, in the case a sequence of GOP images of four images.
• the temporal filtering is carried out along the trajectory of each point of the basic image which was chosen during step 3 of estimation of the movement.
• Each image referenced in FIG. 2 in image_1, image_2 is represented in the serialized form of its pixels, which corresponds to the classic one-dimensional wavelet transformation scheme.
• the trajectory considered therefore passes through the positions marked respectively x1 on image_1, x2 on image_2, etc.
• the arrows F1, F2 and F3 represent the successive inter-image displacements and make it possible to identify the wavelet coefficients corresponding to the trajectory considered for the entire sequence of GOP images.
• step of decomposing into one-dimensional wavelets is shown, mentioned in step 2 of FIG. 1.
• a high-pass filtering step 11 and a sub-sampling by 2 are applied which produces a pair of high frequency coefficients [H1, H2].
• a step 12 of low pass filtering and subsampling by 2 is again applied which produces a coefficient [LL] which represents the lowest time frequency component and in parallel, a step 13 of high-pass filtering and of sub-sampling by 2 which produces a coefficient [LH] which represents an intermediate frequency component.
• the method of the invention mainly to associate a unit of configuration information for coding by 3D wavelets with the data stream described at output 7 of the conventional coder.
• the coding configuration information ensures an optimal response to a cost criterion of the coding scheme.
• Two modes can be implemented: - unidirectional mode, or "forward / backward" (direct / reverse).
• the direction of motion compensation must be constant within a GOP.
• the bidirectional mode for which all the images of the GOP are candidates for the choice of the reference image.
• the two directions of compensation in motion can be used on either side of the reference image.
• Figures 6 and 7 illustrate the unidirectional mode, or "forward / backward" (direct / reverse); Figures 8 and 9 illustrate the bidirectional mode.
• FIG. 6 the decomposition of a flow of images in the unidirectional mode is shown, where only one direction of compensation is allowed per sequence of GOP images.
• Each image is represented in the time domain by a vertical line.
• the image stream is separated in the time sequence of the images from left to right into three sequences of GOP images which are designated successively in time by GOP1, GOP2 and GOP3.
• the number of images in each sequence GOP1, GOP2, GOP3, ... is linked to an increase constraint in the number of pixels not connected by GOP.
• the choice of the reference image R, marked by a strong line is carried out according to a given criterion. Several examples of criteria are given below.
• a first criterion is the minimization of the number of pixels not connected in the GOP. In this case, we calculate the number pixels not connected in the GOP for each of the two directions of compensation in motion. We then choose the direction and the reference image which minimize the number of unconnected pixels for the entire GOP.
• a second criterion is the minimization of the amplitude of the movement. In this case, the amplitude of the movement in the whole of the GOP is measured for each of the two directions, for example by calculating the sum of the squares of the components of each movement vector of each field. Then we choose the direction and the reference image which presents the minimum amplitude of movement.
• a third criterion is the wealth of information in the reference image.
• the texture information of the first and of the last image is measured by the amplitude of the high frequencies of their respective wavelet decomposition, and the richest is chosen as the reference image.
• the image sequences in the image stream are defined as follows in the following table: Sequence Size Reference image Direction
• GOP1 4 images Last image "backward”
• GOP2 8 images Last image "backward"
• GOP3 4 images First “forward” image
• FIG. 7 an embodiment of the method of the invention has been represented, in the unidirectional mode, using the minimization of the number of unconnected pixels.
• the identified steps can each be executed on an identifiable processor or part of an identifiable processor in the form of a process execution circuit and an implementation device is directly derived therefrom.
• a stream of images is acquired and transmitted to a step 21 for estimating the movement in the positive direction between images and concurrently to a step 22 for estimating the movement in the negative direction between images. Then, the construction of the trajectories of the movement points is carried out during steps 24 for the estimation of direct movement 21 and 25 for the estimation of reverse movement 22.
• step 26 the calculation of the number of unconnected pixels is carried out during a step 26 for the construction of trajectories of step 24 and 27 for the construction of trajectories of step 25.
• step 28 the best direction of motion compensation is chosen in step 28 to produce in step 29 the bit stream representative of the spatial transform of the GOP image sequence thus adapted.
• step 28 The choice of the best direction of motion compensation is carried out in step 28 by the smallest of the numbers of unconnected pixels noted in step 26 or in step 27. It is noted that if the positive direction is detected as minimizing the number of unconnected pixels, the reference image chosen is the last of the GOP image sequence being adapted. Conversely, if the negative direction is detected as minimizing the number of unconnected pixels, the reference image chosen is the first of the GOP image sequence being adapted.
• the invention can be extended by allowing all the images of the GOP to be candidates for the choice of the reference image. It is the bidirectional mode. The two directions of temporal decomposition can then be retained for the same sequence of GOP images, on either side of the reference image.
• each GOP sequence has one or two directions of decomposition S and S '.
• the reference image is determined within the GOP sequence.
• the choice of the reference image R can be made according to the same criteria as in the unidirectional mode, that is to say the minimization of the number of pixels not connected on the whole of the GOP, the minimization range of motion across the GOP, and the choice of image
• an embodiment of the method of the invention is represented in the bidirectional mode, that is to say when all the images of the GOP are candidates for the choice of the reference image, with the criterion of minimization. the number of unconnected pixels.
• the identified steps can each be executed on an identifiable processor or part of the identifiable processor in the form of a process execution circuit and an implementation device is directly derived therefrom.
• a stream of images is acquired and transmitted to a step 31 of estimating the movement in the positive direction between images and concurrently to a step 32 of estimating the movement in the negative direction between images.
• a loop 33 is performed for each image of the GOP sequence being adapted during which successively a step 34 of construction of the trajectories is executed, then a step 35 of calculation of the number of unconnected pixels.
• a step of selecting the best reference image R in the sequence of GOP images being adapted is carried out.
• the criterion for selecting the best reference image is that of the lowest number of unconnected points.
• the method of the invention is integrated into a video compressor-decompressor diagram using 3D wavelets.
• the method of the invention is implemented in three separate locations which can be: - a location in which the coding step is implemented; - a place in which the state in which the binary data stream comprising the configuration information unit and the 3D wavelet coding data exists in the form of a transmission signal or else of recorded data medium; and
• the method is integrated into an MPEG TM coding-decoding scheme. This provision is particularly relevant in the context of the implementation of the MPEG-7 standard.
• FIG 10 there is shown a block diagram detailing an embodiment of a 3D wavelet coding device implementing the method of the invention.
• Such a device comprises an input terminal 100 of a video data stream 106, which is connected to the input of an image group analyzer circuit 101 of which a first output terminal 107 is connected to a terminal d 'entry 109' of a circuit
• the circuit 101 finally includes a second output terminal for continuing the 3D wavelet processing on the 3D wavelet encoder 102 proper.
• Output terminals 110 of the encoder 102 and 1 1 1 of the generator circuit 103 are connected to corresponding inputs of a circuit 104 realizing a binary data stream FDB presented at its output 105 and which comprises according to a protocol determined a sequence:
• the data stream thus formed can be transmitted to a new processing circuit, to a recording device or even to a transmission line.
• FIG. 1 there is shown a block diagram detailing an embodiment of a wavelet decoding device
• Such a device comprises an input terminal 120 connected to a discrimination circuit 121 whose output terminals 126 and 127 are respectively connected respectively to a circuit generating information for 3D decoding configuration 122 and to a wavelet decoding circuit 3D 123.
• the discriminator 121 When the flow of binary data from the circuit of FIG. 10 is transmitted to the input terminal 120, the data is analyzed by the discriminator 121 so that, on a first output terminal 126 of the discriminator 121, are presented information characteristic of the 3D wavelet coding applied to the binary data stream.
• the latter is then presented at the output 127 of the discriminator and loaded into the decoder 123.
• the latter includes a configuration input terminal which receives the configuration information from a terminal 128 of the configuration information generator circuit 122 on the base of the information contained in each unit of configuration information such as the datum of the reference image, the direction of decomposition and / or the number of images of the sequence of images represented in the data stream 120.
• a the output terminal 124 presents the sequence of decoded GOP images.
• the present invention finds application in the field of video compression based on decomposition by 3D wavelets. It can therefore be integrated into a 3D wavelet video compressor-decompressor scheme.
• the present invention finds application in a conventional hybrid coding scheme, and particularly in an MPEG type scheme in which the type I images and the P type images serve as reference images. In this case, the method of the invention is integrated into an MPEG coding-decoding scheme.

Landscapes

• Engineering & Computer Science (AREA)
• Multimedia (AREA)
• Signal Processing (AREA)
• Compression Or Coding Systems Of Tv Signals (AREA)
• Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)
• Compression, Expansion, Code Conversion, And Decoders (AREA)

Applications Claiming Priority (2)

Publications (1)

Family

ID=33396691

Family Applications (1)

Country Status (9)

Families Citing this family (6)

Family Cites Families (7)

• 2003
  • 2003-05-23 FR FR0306197A patent/FR2855356A1/fr active Pending
• 2004
  • 2004-05-01 US US10/558,140 patent/US7864863B2/en not_active Expired - Fee Related
  • 2004-05-12 WO PCT/FR2004/050190 patent/WO2004107761A1/fr active Application Filing
  • 2004-05-12 JP JP2006530435A patent/JP4990622B2/ja not_active Expired - Fee Related
  • 2004-05-12 CN CN200480012420A patent/CN100591127C/zh not_active Expired - Fee Related
  • 2004-05-12 EP EP04785573A patent/EP1636996A1/fr not_active Withdrawn
  • 2004-05-12 KR KR1020057022298A patent/KR101011894B1/ko not_active IP Right Cessation
  • 2004-05-21 MY MYPI20041955A patent/MY135286A/en unknown
  • 2004-05-21 TW TW093114357A patent/TWI339071B/zh not_active IP Right Cessation

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events