[go: up one dir, main page]

WO2002013537A2 - Procede et dispositif de codage et de decodage de donnees video et d'images transformees par ondelettes - Google Patents

Procede et dispositif de codage et de decodage de donnees video et d'images transformees par ondelettes Download PDF

Info

Publication number
WO2002013537A2
WO2002013537A2 PCT/EP2001/009128 EP0109128W WO0213537A2 WO 2002013537 A2 WO2002013537 A2 WO 2002013537A2 EP 0109128 W EP0109128 W EP 0109128W WO 0213537 A2 WO0213537 A2 WO 0213537A2
Authority
WO
WIPO (PCT)
Prior art keywords
wavelet
group
wavelet coefficients
coefficients
coefficient
Prior art date
Application number
PCT/EP2001/009128
Other languages
German (de)
English (en)
Other versions
WO2002013537A3 (fr
Inventor
Axel Becker
Original Assignee
Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e. V.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e. V. filed Critical Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e. V.
Priority to AU2001293733A priority Critical patent/AU2001293733A1/en
Publication of WO2002013537A2 publication Critical patent/WO2002013537A2/fr
Publication of WO2002013537A3 publication Critical patent/WO2002013537A3/fr

Links

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
    • H04N19/60—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
    • H04N19/64—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding using sub-band based transform, e.g. wavelets characterised by ordering of coefficients or of bits for transmission
    • H04N19/647—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding using sub-band based transform, e.g. wavelets characterised by ordering of coefficients or of bits for transmission using significance based coding, e.g. Embedded Zerotrees of Wavelets [EZW] or Set Partitioning in Hierarchical Trees [SPIHT]

Definitions

  • the present invention relates to image and video compression, for example for archiving or transmitting video data.
  • FIG. 8 shows a schematic representation of the method.
  • An image 10 is an array of x values along the x axis and y values along the y axis, where. the image can be represented as a function f (x, y) can, and the x and y axes are indicated by reference numerals 20 and 30.
  • the image 10 is subjected to low-pass filtering along the x-axis, which results in a low-pass filtered image f (x, y) 50.
  • the image 10 is subjected to a high-pass filtering along the x-axis in a step 60, resulting in a high-pass filtered image f H (x, y) 70.
  • the filtered images are additionally undersampled by a factor of 2 along the x- in steps 40 and 60.
  • Ax is subjected, with no loss of information.
  • the subsampling is carried out by omitting every second filtered value.
  • the images f L (X / y) and f H (x, y) thus generated consist of fields with (x / 2) * y wavelet coefficients.
  • the image f L (x, y) 50 in a step 80 of low-pass filtering along the y-axis and a subsequent subsampling by a factor of two along the y-axis and in step 90 of high-pass filtering along the y- Axis and a subsequent subsampling by a factor of two along the y-axis.
  • the image f H (x, y) 70 becomes low-pass filtering along the y-axis at step 100 and then subsampled along the y-axis by a factor of two and high-pass filtering along the y-axis and one at step 110 Subsequent subsampling by a factor of two along the y-axis.
  • steps 80, 90, 100 and 110 thus results in four partial images 120, 130, 140 and 150, which with fu, (X / y) / fLH (x, y) / ffl_ (X / y) and f HH (x, y) are displayed, where L indicates low-pass filtering and H indicates high-pass filtering, and the respective first index indicates the corresponding filtering along the x-axis and the respective second index indicates the filtering along the y-axis.
  • Field 120 is low pass filtered along both axes and thus corresponds to an average signal, i.e. a smoothed image.
  • the partial images 130, 140 and 150 are high-pass filtered at least along one of the two axes and are consequently direction-sensitive detail signals.
  • the partial image 130 emphasizes horizontal image features
  • the partial image 150 diagonal image features Consequently, the wavelet transformation of the image 10 yields four fields 120, 130, 140 and 150, each of which, due to the subsampling along both axes, the number of wavelet coefficients is only a quarter of the number of image values of the input image 10.
  • the individual partial images are shown in sequencing a sequence of wavelet coefficients, whereupon the wavelet coefficients are subjected to quantization.
  • the quantization is carried out with the aid of a step function which represents the threshold values, whereby a wavelet coefficient is disregarded if the amount thereof does not exceed the relevant threshold value.
  • the sequence of quantized wavelet coefficients thus generated is subjected to entropy coding, in which code words of different lengths are assigned to the different quantization values depending on the frequency of their occurrence in order to achieve the highest possible compression rate.
  • a problem that arises with the aforementioned wavelet-based coding methods is that so-called “mosquito artifacts” occur during the decoding and playback of a sequence of images coded by these methods, which are caused by flickering in uniform image regions even at high levels and medium frame rates.
  • the reason for ⁇ as that conventional, wavelet-based method Mosquito Arte ⁇ fakte produce, is that the value of a wavelet coefficient ⁇ at a fixed position within a part of the image during successive frames of a video sequence oscillates strongly. In the case of a wavelet coefficient, this can lead to the fact that the same value for quantization is exceeded and coded, but already in the following frame does not exceed the threshold value and is neglected because the amount is too small.
  • mosquito artefacts arise at the image position corresponding to this wavelet coefficient.
  • the problem of mosquito artefacts is exacerbated by the fact that in many of the areas of application mentioned above, motion compensation must not take place, so that the mosquito artefacts cannot be filtered in the time domain, ie between successive frames. Spatial filtering, on the other hand, would degrade the image quality more than the mosquito artifacts do anyway.
  • IEEE Transactions on Consumer Electronics, Vol. 45, 1, 1999, pp. 13 to 20 describes a hierarchically progressive image coding which is intended to convey a recognizable visual impression of an image as quickly as possible during image transmission. This is achieved in that the image to be encoded is first subjected to a modified region growth algorithm which divides the image into closed regions within which the image is almost constant. A difference ⁇ image r 0 is generated from this image s o divided into regions by subtracting a bilinearly interpolated image s 0 from the image s 0 , as described on page 15, first column, third paragraph.
  • the difference image rO is then subjected to a wavelet transformation, the selection which of the wavelet coefficients is coded and which is not coded is divided into a preliminary selection and a zero tree or zero tree coding. In the preselection, those wavelet coefficients are excluded from further processing, firstly to the
  • each difference image itself is divided into tree structures in a suitable manner.
  • the object of the present invention is to provide a method and a device for coding wavelet-transformed video and image data, so that fewer mosquito artefacts are generated when an image sequence is played back.
  • a further object of the present device consists in ⁇ represents a method and an apparatus for decoding co-founded ⁇ wavelet-transformed video and image data fen to sheep ⁇ so that less ⁇ Mosquito artifacts are generated when playing back a sequence of images.
  • a method for coding wavelet-transformed video and image data, the video and image data comprising wavelet coefficients of a plurality of subbands comprises combining at least two wavelet coefficients into a group of wavelet coefficients, the group of wavelet coefficients comprising at least two different wavelet coefficients Subbanders has.
  • the method further comprises comparing the group's wavelet coefficients to a predetermined threshold to obtain an actual relationship of each wavelet coefficient to the predetermined threshold.
  • the group overriding is provided.
  • An inventive apparatus for encoding wavelet transformed video and image data said video and image data wavelet coefficients comprise a plurality of subbands, comprising means for combining at least two wavelet coefficients to a set of wavelet ⁇ , wherein the group of wavelet coefficients of wavelet coefficients from at least two different sub-bands. Means are also provided for comparing the wavelet coefficients of the group with a predetermined threshold value in order to determine an actual relationship of each wavelet coefficient to the predetermined threshold value.
  • a device for decoding coded wavelet-transformed video and image data, the coded video and image data having at least one significance word to which a predetermined threshold value is assigned, the significance word the actual relationships of a group of at least two original wavelet coefficients to the predetermined one Threshold, and wherein the actual relationship of at least one of the at least two original wavelet coefficients is equal to a predetermined relationship to the threshold, and wherein at least two of the original wavelet coefficients belong to different ones of a plurality of subbands of the wavelet transformed video image includes means for determining the original wavelet coefficient of the at least two original wavelet coefficients of the group, the actual relationship of which is equal to the predetermined relationship, using the Threshold value in order to obtain a determined wavelet coefficient, the determined wavelet coefficient corresponding to the original wavelet coefficient within an accuracy determined by the predetermined threshold value.
  • the directionally sensitive wavelet partial images are scanned along their respective sensitivity directions and stored in one-dimensional arrays.
  • each of the first wavelet coefficients of the directional ⁇ sensitive subbands are combined into a first group, the respective second wavelet coefficients of these subbands to a second group combined, etc.
  • the thus generated groups of wavelet coefficients a significance test are subjected, in which the amounts of the wavelet coefficients with a smoldering 10
  • a wavelet coefficient is significant
  • this wavelet coefficient is positive or negative, i.e. is positively significant or negatively significant.
  • a significance word is generated as a result of the significance test. If none of the group's wavelet coefficients is significant, the group is transitioned and a run length counter value incremented. Otherwise, the significance word is saved and encoded. After all groups have undergone the significance test, the significance test for the groups is repeated using a halved threshold. In this way the quantization and coding of the wavelet coefficients is carried out.
  • the video and image data coded in this way contain run length counter values and significance words, each of which is assigned a threshold value.
  • the run length counter values When decoding, using the run length counter values, the number of groups that have been passed over or skipped over during coding can be determined, so that the meaningful words can be assigned to the correct groups of original wavelet coefficients during decoding. If a significance word is received during decoding, a determined wavelet coefficient is brought up to the original wavelet coefficient using the threshold value assigned to the significance word, so that each determined wavelet coefficient deviates from the original wavelet coefficient by at most twice the smallest threshold value after decoding.
  • An advantage of the present invention is that under 11 -
  • the mosquito artifacts arise from the fact that the absolute amount of a wavelet coefficient during successive frames of a video sequence exceeds and does not exceed a quantization threshold, and is therefore encoded irregularly and not encoded.
  • at least two wavelet coefficients are combined to form a group, one of these wavelet coefficients being encoded only in the event that no wavelet coefficient in the group is significant, or the amount of none of the wavelet coefficients in the group exceeds the predetermined threshold value.
  • the likelihood that none of the wavelet coefficients in a group is significant is significantly less than the likelihood that a single one of the wavelet coefficients is not significant.
  • Another advantage of the present invention is that although by combining at least two wavelet coefficients into a group and passing over a group only in the event that none of the wavelet coefficients is significant, as previously described, the number of to coding wavelet coefficients is increased, the 12
  • Compression rate can still be kept low.
  • the reason for this is that the quantization of the wavelet coefficients by means of the significance words produces an encoder output signal with a limited set of output code words.
  • the effect of a subsequent entropy coding can hereby be increased, since the frequently occurring significance words can be coded with code words of shorter length.
  • FIG. 1 shows a flow chart of a method according to the invention for encoding wavelet-transformed video and image data
  • FIG. 2 shows a schematic diagram for illustrating the on from a wavelet-transformed video image
  • Fig. 3 is a schematic diagram illustrating the sequencing of subbands and Zusammmen--SETTING of wavelet coefficients into groups in accordance with egg ⁇ nem particular embodiment
  • Fig. 4 is a schematic diagram illustrating the structure of one word according to a significance exporting ⁇ approximately example of the present invention
  • FIG. 5 is a flowchart of a procedure for encoding video and image data according to a special embodiment. - 13 -
  • FIG. 6 shows a diagram for illustrating a method according to the invention for decoding coded wavelet-transformed video and image data
  • FIG. 7 is a flowchart of a procedure for decoding encoded wavelet-transformed video and image data in accordance with a particular embodiment of the present invention.
  • FIG. 8 shows a flow chart of a method for performing a wavelet transformation of a video image according to the prior art.
  • FIG. 1 a method according to the invention for encoding wavelet-transformed video and image data is first described, but it is pointed out that the flowchart of FIG. 1 also serves as a block diagram of a device for encoding wavelet-transformed video and image data can by replacing each step of the flow chart of Figure 1 with a corresponding means for performing the step.
  • a device could, for example, be implemented in an ASIC (application-specific integrated circuit), programmable logic, software or firmware or other hardware.
  • ASIC application-specific integrated circuit
  • the video image to be encoded is first made available.
  • the video image is in digital form and consists of an array of image values.
  • the video image can - 14 -
  • the video image can be generated by a variety of devices, such as a digital camera, digital library, digital camcorder, or the like.
  • a step 210 the video image is subjected to a wavelet transformation, as is shown, for example, in FIG. 8.
  • the wavelet transformation can be carried out once or recursively several times, a plurality of wavelet partial images being generated, as has been described with reference to FIG. 8.
  • the wavelet coefficients contained in the wavelet fields and generated in step 210 are combined in a step 220 such that at least two wavelet coefficients form a group of wavelet coefficients. In this way, successive groups of wavelet coefficients are formed, each group of wavelet coefficients having wavelet coefficients from at least two different subbands. It is preferred that each wavelet coefficient be included in at most one group, but it is also possible that one or more wavelet coefficients belong to multiple groups.
  • a next step 230 the successive groups of wavelet coefficients are subjected to a comparison.
  • the wavelet coefficients of each group are compared to a predetermined threshold.
  • the comparison of a wavelet coefficient with a predetermined threshold value can comprise, for example, comparing the amount of the wavelet coefficient with the predetermined threshold value.
  • the determination can also be provided - 15 -
  • Step 230 obtains an actual relationship of each wavelet coefficient of a group to the predetermined threshold. If the actual relationship of at least one wavelet coefficient of a group is equal to a predetermined relationship to the predetermined threshold, the relationship obtained by comparison of each wavelet coefficient of that group to the threshold is stored in step 240 as a significance word, thereby quantizing the wavelet coefficients of that group , Otherwise, i.e. if the actual relationship of each wavelet coefficient of a group is not equal to the predetermined relationship, the group is skipped in step 250.
  • the predetermined relationship may be defined such that the actual relationship of a wavelet coefficient is equal to the predetermined relationship if its amount exceeds the predetermined threshold.
  • the significance words generated by step 240 are entropy-encoded in step 260.
  • a different encoding of the significance of words could be used to ⁇ , such as a coding with fixed length code words or a tree coding.
  • the code words generated in step 260 are formatted into a bit stream in step 270, thereby generating an encoded video signal.
  • the bit stream in step 270 for example, the consideration of a byte budget (a good - 16 -
  • step 230 may also include determining the sign of those wavelet coefficients of the group whose magnitude has exceeded the predetermined threshold. The determined sign is stored in the significance word at step 240 so that when decoding the encoded video signal it can be determined whether the original wavelet coefficient was greater than the positive of the predetermined threshold or less than the negative of the threshold, i.e. whether the original wavelet coefficient was positively significant or negatively significant with respect to the predetermined threshold. In some industrial applications, however, it may be sufficient to merely determine in step 230 whether the magnitude of a group's wavelet coefficient exceeds the predetermined threshold.
  • Such applications include, for example, video surveillance, in which it is only determined whether changes occur in successive video images, e.g. for burglar protection.
  • other comparisons could also be used in step 230, such as determining whether a group's wavelet coefficients are greater than a predetermined threshold, disregarding negative wavelet coefficients.
  • step 250 the following is pointed out - 17 -
  • step 250 it can be provided, for example, in step 250 to also add a run length counter value Increment, whose instantaneous value shows how many groups of wavelet coefficients have been passed since the last time a significance word was saved. In this case, the run length counter value would be saved the next time step 240 is performed. In this way, the encoded video signal contains information about how many passed groups are between successive significance words.
  • a significance word identification is stored, so that the significance word can be identified on the basis of the significance word ID during decoding.
  • steps 230, 240 and 250 in several ⁇ ren runs carried out may be, wherein the predetermined threshold is changed in each run.
  • the predetermined threshold value may be halved before each pass, so that the wavelet coefficients are sampled or quantized more precisely with each pass, as will be described in more detail below. It should be noted that although the quantization of the wavelet coefficients described above by forming significance words in step 240 in only one step - 18 -
  • FIG. 2 illustrates the structure of a wavelet-transformed video image
  • FIG. 3 illustrates the combination of wavelet coefficients into a group of wavelet coefficients
  • FIG. 4 illustrates the structure of the significance word in accordance with a specific exemplary embodiment of the present invention
  • FIG. 5 shows a flow diagram of a method according to a special exemplary embodiment of the present invention.
  • FIG. 2 shows the structure of a wavelet-transformed video image which has been generated by recursively applying a wavelet transform three times to a video image.
  • the partial images created by the recursive wavelet transformation are represented by ten squares, which are numbered with Roman numerals from I - IX or with 0.
  • the partial images VII-IX correspond to the directionally sensitive partial images as they arise after the first-time application of a wavelet transformation, as has been described, for example, with reference to FIG. 8.
  • Each field VII - IX is half as long and half as wide as the original, non-wavelet-transformed video image.
  • sub-image VII reinforces horizontal image features, sub-image VIII vertical image features and sub-image IX diagonal image features.
  • Sectional image that was created when the wavelet transformation was used for the first time has undergone a further wavelet transformation in which the sub-images IV, V and VI have been created, the size of which in turn corresponds to a quarter of the size of the sub-images VII - IX.
  • the average partial image created in the second wavelet transformation has in turn been subjected to a wavelet transformation, the partial images 0, I, II, III having arisen, the sizes of which are also a quarter of the size of the partial images IV-VI. Due to the fact that partial images with the size of a quarter of the input image are generated during a wavelet transformation, the partial images can be displayed arranged in such a way that they combine to give the size of the original video image, as shown in FIG.
  • the partial images 0 - IX can be divided into levels, as indicated by curly brackets in FIG. 2.
  • drawing files 0, I, II and III belong to the first level
  • drawing files IV, V and VI belong to the second level
  • drawing files VII, VIII and IX belong to the third level.
  • the lowest or first level in each case consequently contains an average partial image, namely partial image 0, and consequently has four partial images as the only level.
  • the respective average partial image has been subjected to a further wavelet transformation.
  • Each field 0 - IX consists of an array of wavelet coefficients ⁇ , wherein in the partial images VII - IV respectively shown an array of 4x4 wavelet coefficients of the array and 20 -
  • the partial images IV-VI contain array partial fields 330, 340 and 350 which have arisen from the wavelet transformation of the average partial image of the third level and which are also correlated with the 8x8 field of the video image.
  • the fields 330-350 are 2x2 fields, the position coordinates of the upper left field point in each case corresponding, for example, to half the position coordinates of the upper left field point of the fields 300-320.
  • the first level contains a wavelet coefficient 360, 370, 380 and 390, which are correlated with the 8x8 field of the video image. From what has been described above it is clear that the wavelet coefficients 360-390 and the wavelet coefficients of the fields 300-350 are correlated with one another and relate to the same part of the original video image. The correlation is represented by arrows 400 and 410.
  • individual wavelet coefficients of a wavelet-transformed video image are represented by individual boxes.
  • the wavelet coefficients indicated with 500 are, for example, the wavelet coefficients of a subband 0 of the first level.
  • the wavelet coefficients indicated by 510, 520 and 530 are, for example, the wavelet coefficients of the subbands I, II and III of the first level.
  • 540, 550 and 560 indicate the wavelet coefficients of the sub-bands IV, V and VI of the second level.
  • an order is first defined among the wavelet coefficients of the partial images in order to obtain a subband for each partial image.
  • the wavelet coefficients of the subbands 0 - VI are each stored in a one-dimensional array.
  • Subfield 0 of FIG. 2 is scanned line by line, ie horizontally, and sequentially stored in a one-dimensional array, the first wavelet coefficient of subband 0 being the top left wavelet coefficient of subfield 0 of FIG. 2 and the last wavelet coefficient of the sub-band 0 to the lower right Waveletkoeffizi ⁇ ducks of the field 0 corresponds.
  • the wavelet coefficients of the sub-images I and IV are also scanned horizontally, ie line by line, from the top left corner and stored sequentially in one-dimensional arrays, so that the last wavelet coefficient of the subbands I and IV is respectively that 22 -
  • the wavelet coefficients of fields 2 and 5 are sampled diagonally and stored sequentially in arrays, the wavelet coefficients being sampled, for example, from the top right corner to the bottom left corner, and fields 3 and 6 are sampled in the vertical direction and stored sequentially in arrays
  • the sequentialization begins, for example, with the lower left wavelet coefficient and ends with the upper right wavelet coefficient of field 3 or 6.
  • the wavelet coefficients sequenced in subbands are denoted by a (i, k), where i indicates the position of the wavelet coefficient within the one-dimensional array and k the subband, which also includes the wavelet coefficient.
  • the wavelet coefficients of the subband 0 are fed to the quantization one after the other, as in the conventional wavelet-based coding techniques.
  • the first wavelet coefficients of the subbands I-III are then each supplied to the quantization in succession as a group 580, as indicated by an arrow 581.
  • the respective second wavelet coefficients of the subbands I - III, the respectively third wavelet coefficients of the subbands I - III and then again the fourth fourth wavelet coefficients of the subbands I - III are fed to the quantization as described by - 23 -
  • the wavelet fields are first sequenced and then fed to the quantization in groups.
  • the wavelet coefficients of the wavelet-transformed video image are stored in a random access memory, however, it is also possible to directly carry out the supply of the wavelet coefficients for the quantization in the order described above by directly accessing the individual wavelet coefficients.
  • the wavelet coefficients subbandweise ribbon on ver ⁇ different lines with a sequential line per sub ⁇ , for grouping a multiple- could be xer used the first forwards the wavelet coefficients of the subbands 0 and then sequentially to each of the first wavelet coefficients of the Subbands III, II and then I, forwards the respective second wavelet coefficient of subbands I - III, etc.
  • this data can be temporarily stored in a di- - 24 -
  • right access memory are provided in order to then supply the wavelet coefficients to the quantization in a suitable sequence as described with reference to FIG. 3.
  • the groups since the groups only contain coefficients which are not correlated with one another by their position within a frame, in the event that the group is passed over, there remains only a wavelet coefficient of three wavelet coefficients which belong to a specific image area of the video image , disregarded and uncoded. As a result of this uncorrelation, the mosquito artifact suppression according to the present invention can be further increased. It is pointed out that the uncorrelation of wavelet coefficients within a group can also be achieved in a different way than in the manner described above.
  • the partial images could be scanned and sequenced in a different order.
  • groups of two could be formed from wavelet coefficients of subband I and III as well as subband II and III or corresponding combinations of subbands. It is also possible to include the 0th subband in the group formation. - 25 -
  • the wavelet coefficients 500-560 After the wavelet coefficients 500-560 have been placed in an order as indicated by arrows 570, 581-584 and 591-592, they can be fed to a device for comparing three successive wavelet coefficients each with a predetermined threshold as they do results from the corresponding step 230 of FIG. 1, after which they are quantized by forming significance words.
  • the quantization according to a special exemplary embodiment is described, in which the wavelet coefficients are grouped according to FIG. 3 and supplied, and can be used for steps 230, 240 and 250 from FIG. 1.
  • a significance word 600 comprises six bits 601, 602, 603, 604, 605 and 606, which are represented by squares.
  • a (i, k), a (i, k + l), a (i, k + 2) ⁇ it is checked whether it is significant or whether the amount of the ⁇ same a predetermined threshold value exceeds. If the wavelet coefficients a (i, k + 2) of the maximum sub-band in ⁇ nergur of Tripeis is significant, the most significant bit is set 601 of the significance of the word 600 to 1, otherwise shall be set to 0.
  • the Set bits 603 and 605 according to the significance of the wavelet coefficients a (i, k + l) and a (i, k) If the wavelet coefficient a (i, k + 2) is significant, it is checked whether it is positive or negative . In the case that the same is positive, the bit 602 is set to 1 ge ⁇ and otherwise to 0. In the case that the wavelet 26 -
  • bits 604 and 606 of the significance word 600 are set in accordance with the sign of the coefficients a (i, k + l) and a (i, k), respectively.
  • each wavelet coefficient of the trip ice is assigned two bits, a first bit for storing whether the amount of the corresponding wavelet coefficient exceeds the threshold value and a second bit for storing whether the corresponding wavelet coefficient is positive. is significant or negatively significant.
  • the significance word 600 can have more bits, for example, a significance word ID being stored in the further bits, for example. It can further be provided that there is no fixed assignment of two bits of the significance word to a wavelet coefficient of the group, and that instead each possible combination of the actual relationships of the wavelet coefficients to the threshold value is assigned a different significance word of the same length. In the present case, three wavelet coefficients per group game, five bits would suffice for ⁇ , encode the 27 possible combinations in different words significance.
  • a threshold value S and a run length counter value Z are initialized.
  • the threshold value S is set, for example, to half the upper limit of the value range of the wavelet coefficients.
  • the run length counter value Z is initialized to 0.
  • a group counter value is initialized to 0.
  • it is then checked whether at least one wave coefficient ai with i 1, 2 or 3 from the group ⁇ ai, a 2 , a3 ⁇ . gr exists so that the amount of the wavelet coefficient ai is greater than S. If this is not the case, Z is incremented in a step 640 and gr is incremented in a step 650. From step 650, the procedure returns to step 630.
  • step 660 For each wavelet coefficient whose magnitude is greater than S whether the wavelet coefficient ai is positive. If this is the case, the value of the wavelet coefficient ai is reduced by the value of the threshold value S in a step 670. If the wavelet coefficient ai, however, negative, the value of the threshold value S to the Wa ⁇ veletkostoryen ai is added in a step 680th In a step 690 represents ⁇ is checked alshin whether the run length counter value Z is zero. If not, it means that groups have already been skipped and in a step 700 the value of Z is stored. The value of Z is then set to 0 in step 710. From step 710 and then if - 28 -
  • step 720 in which a significance word containing the information obtained from polls 630 and 660 is stored.
  • step 730 it is checked whether the group counter value gr exceeds the number of groups in the sequence of groups within the wavelet-transformed video image. If this is not the case, the group counter value is incremented in a step 740 and the procedure returns to query 630. However, if the group counter value at query 730 exceeds the number of groups, the procedure jumps to query 750 where it is checked whether the procedure should be ended. The check in step 750 could, for example, provide a query as to whether the byte budget has already been used up.
  • the check could also provide a query as to whether the threshold has already reached a predetermined quantization level, i.e. whether the threshold value S has fallen below a predetermined value. If it is determined in step 750 that the procedure is to be continued, the threshold value is halved in step 760 and the procedure loops back to step 620. Otherwise the procedure ends. Variations on this procedure result from the description of the previous figures.
  • Passes are generated only affect those groups in which at least one wavelet coefficient exceeds the threshold value S.
  • the threshold value is subtracted from the wavelet coefficient or added to the same, so that after the first pass, all wavelet coefficients are in a range of values which is smaller than the threshold value.
  • the threshold value S is halved, so that in the second pass significance words are only stored for those groups in which at least one significance word from the group exceeds the now halved threshold value. Consequently, in this procedure, several significance words are assigned to each wavelet coefficient, which are assigned to different threshold values. When decoding, the value of the wavelet coefficient can be determined from these significance words with the accuracy of the smallest threshold value.
  • run length coding is carried out in the procedure of FIG.
  • the run length counter value is pre-initialized to the value 0 and, in the event that groups are skipped, incremented until a significance word is written again.
  • decoding it can be concluded from this how many significance words have been passed over during coding, or to which group in the sequence of groups the relevant significance word relates.
  • the procedure is carried out until the byte budget is used up, or until the amount of data for coding the video image has been reached by the significance words and run length counter values written so far.
  • the bits required for storing the significance words and run length counter values are subtracted from the byte budget immediately upon storage, and the same is checked to see whether the byte budget still available has been used up.
  • the threshold may be changed in a different manner at each step in step 760.
  • a sequence of threshold values can be stored in a look-up table, which decrease exponentially with each pass.
  • the encoded video and image data contain a plurality of significance words, each of which is assigned a predetermined threshold value.
  • Each word contains the actual significance of relationships of a group of at least two original wavelet coefficients to the determined before ⁇ threshold value, whereby in particular the actual relationship at least one of the at least two original Wa ⁇ veletkostoryen equal to a predetermined relationship to the threshold value.
  • at least two of the Original Art ⁇ chen wavelet coefficients at different include a plurality of subbands of the wavelet transformed video image.
  • the original wavelet coefficients of at least two original wavelet coefficients are ermit- 31
  • the threshold value assigned to the significance word is used for this.
  • the original wavelet coefficients to which a certain significance word of the encoded video data relates can be determined, for example, by run length counter values also contained in the encoded video data or by significance word IDs. It is assumed here that during decoding the order in which the original wavelet coefficients were supplied for quantization is known, as has been described, for example, with reference to FIG. 3, or in which order the significance words are located. The number of significance words skipped in order can then be determined from the run length counter values. When using significance word IDs in decoding only the knowledge of the significance of association between the words and the significance word IDs is required, the Sieren during quantization of the significance of words and storing the signifi cance ⁇ words has been used.
  • Determining the original wavelet coefficients of the group using the threshold value in the step 800 is dependent lenwert of the encoding steps of the comparison of ur ⁇ sprün Mandarin wavelet coefficients with the predetermined smoldering ⁇ obtained and of storing by the comparison 32 -
  • each wavelet coefficient (e.g. steps 230 and 240 of Figure 1).
  • the determined wavelet coefficients are thereby obtained in step 800 that for each original wavelet coefficient, the amount of which has exceeded the threshold value, the determined wavelet coefficient is set to the threshold value assigned to the significance word.
  • the value of the determined wavelet coefficients is set to a value other than the threshold value, for example to compensate for a power loss that occurs as a result of the coding and decoding of the wavelet coefficients.
  • a lookup table could be used for this purpose, in which a wavelet coefficient is assigned to each threshold value.
  • FIG. 6 it is pointed out that this figure can also be regarded as the block diagram of a device according to the invention for decoding coded wavelet-transformed video and image data, by replacing the determination step in the flow diagram of FIG. 6 with a device for determining becomes.
  • a ent ⁇ speaking device could be implemented by an ASIC, a programmable logic, a software program, a farm product, or any other hardware that performs the function described.
  • the procedure begins at step 810 with initializing a threshold S and a pointer value Z, which is used to sample the order used in quantizing the original wavelet coefficients.
  • the threshold value S is initialized to the value that was used last when the video image was encoded, ie during the last run.
  • the threshold values used in the encoding can be contained in the encoded video data and can be read out in the decoding.
  • the pointer value is initialized to one. 'In a step 820 it is checked whether the pointer value Z has exceeded a certain size, corresponding to the number of groups to which the original wavelet coefficients were pooled.
  • step 830 the counter value is initialized to the value one and halves in egg ⁇ nem step 840, the threshold S (or alternatively read out from the encoded video data). From step 840 and in the event that the counter value Z does not exceed the determined size, the procedure proceeds to step 850. In step 850 the next element of the co-founded ⁇ video data is read. In a subsequent step - 34
  • a loop counter i is set to 1 in a step 870.
  • the first bit of the significance word associated with the wavelet coefficient ai of group z is checked to see if it is one, which means that the associated original wavelet coefficient was significant with respect to the threshold S. If this is the case, a query 890 checks whether the second bit, which is assigned to the wavelet coefficient ai of group z, of the significance word is one, which means that the assigned original wavelet coefficient is negatively significant.
  • the value of the corresponding wavelet coefficient ie the value of the wavelet coefficient to which the i-th two bits are assigned, is increased by the threshold value S in a step 900.
  • the value of the wavelet coefficient ⁇ a ⁇ is reduced by the threshold value, where ⁇ a ⁇ means the i-th wavelet coefficient of the z-th group.
  • the procedure jumps from steps 900 and 910 to a query 920, in which it is checked whether the loop counter i has reached the value 3. If this is not the case, the loop counter is incremented in step 930 and the procedure loops back to step 880. Was the first bit in the query 880 is zero, the procedure also proceeds to step 930.
  • step 940 the Zei ⁇ is Gerwert Z- incremented and the procedure returns to query 820.
  • the element read in step 850 is a run length counter value
  • step 950 the value of Z is increased by the run length counter value.
  • Figs. 5 and 7 can be implemented by suitable hardware, firmware or software, e.g. an ASIC or programmable logic.
  • wavelet coefficients it is pointed out that it is not necessary that all wavelet coefficients be combined into groups.
  • the wavelet coefficients of a group are combined in such a way that some wavelet coefficients belong to the same subband. It is also possible that wavelet coefficients of different levels are combined, or that some wavelet coefficients belong to several groups.

Landscapes

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

Abstract

Procédé de codage de données vidéo et d'images transformées par ondelettes, selon lequel les données vidéo et d'images comprennent des coefficients d'ondelette d'une pluralité de sous-bandes. Ledit procédé comporte la réunion d'au moins deux coefficients d'ondelette pour former un groupe (580, 890) de coefficients d'ondelettes, ledit groupe (580, 590) de coefficients d'ondelettes contenant des coefficients d'ondelette provenant d'au moins deux sous-bandes différentes. Ledit procédé comporte en outre la comparaison des coefficients d'ondelette du groupe (580, 590) avec une valeur de seuil prédéterminée pour obtenir un rapport effectif de chacun des coefficients d'ondelette à la valeur de seuil prédéterminée. Si le rapport effectif d'au moins un coefficient d'ondelette est égal à un rapport prédéterminé à la valeur de seuil prédéterminée, le rapport effectif de chaque coefficient d'ondelette du groupe (580, 590) à la valeur de seuil est mis en mémoire pour obtenir un mot significatif (600) pour le groupe (580, 590) de coefficients d'ondelette. Si le rapport effectif de chaque coefficient d'ondelette n'est pas identique au rapport prédéterminé, le groupe (580, 590) est omis.
PCT/EP2001/009128 2000-08-07 2001-08-07 Procede et dispositif de codage et de decodage de donnees video et d'images transformees par ondelettes WO2002013537A2 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
AU2001293733A AU2001293733A1 (en) 2000-08-07 2001-08-07 Device and method for encoding and for decoding wavelet-transformed video and image data

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
DE10038400.5 2000-08-07
DE10038400A DE10038400C2 (de) 2000-08-07 2000-08-07 Vorrichtung und Verfahren zum Codieren von Wavelet-transformierten Video- und Bilddaten sowie Verfahren und Vorrichtung zum Decodieren von codierten Wavelet-transformierten Video- und Bilddaten

Publications (2)

Publication Number Publication Date
WO2002013537A2 true WO2002013537A2 (fr) 2002-02-14
WO2002013537A3 WO2002013537A3 (fr) 2002-07-18

Family

ID=7651529

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2001/009128 WO2002013537A2 (fr) 2000-08-07 2001-08-07 Procede et dispositif de codage et de decodage de donnees video et d'images transformees par ondelettes

Country Status (3)

Country Link
AU (1) AU2001293733A1 (fr)
DE (1) DE10038400C2 (fr)
WO (1) WO2002013537A2 (fr)

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5315670A (en) * 1991-11-12 1994-05-24 General Electric Company Digital data compression system including zerotree coefficient coding

Also Published As

Publication number Publication date
DE10038400A1 (de) 2002-02-21
AU2001293733A1 (en) 2002-02-18
WO2002013537A3 (fr) 2002-07-18
DE10038400C2 (de) 2002-11-21

Similar Documents

Publication Publication Date Title
DE69416364T2 (de) Bilddatenkompression
DE69321896T2 (de) Gerät und Verfahren zur Verarbeitung von digitalen Bildsignalen
DE69324958T2 (de) Kompression und Dekompression von Bildern
DE69525312T2 (de) Verfahren und Vorrichtung zum Dekodieren von Bilddaten
DE69324134T2 (de) Hdtv empfänger mit einer schaltung zur umsetzung von daten hoher auflösung in daten niedrigerer auflösung
DE69936723T2 (de) Verfahren zur Datenbearbeitung und zum Einfügen eines Wasserzeichens
DE69225211T2 (de) Verfahren und Vorrichtung zum Aufzeichnen von Videosignalen auf einem Aufzeichnungsträger
EP1599835B1 (fr) Procede et ensemble de videocodage, le videocodage comprenant l'analyse de texture et la synthese de texture, programme informatique correspondant et support d'enregistrement correspondant lisible par un ordinateur
DE69324090T2 (de) Intraframefilter für Bildkompressionssysteme
DE69228893T2 (de) Vorrichtung und Verfahren zur Datenmischung und -entmischung
DE69326366T2 (de) Verfahren und Vorrichtung zur Bildsignalkodierung mit merhstufigen Quantisierungnummernbestimmung
DE69022623T2 (de) Gegenständlich adaptiertes Bildkodiersystem.
DE69224227T2 (de) Verfahren und Anordnung zur Bestimmung eines Quantisierungsfaktors für Prozesse mit Datenkomprimierung und -dekomprimierung
DE69226160T2 (de) Vorrichtung zur Kompressionskodierung von Videosignalen
DE69523439T2 (de) Verfahren und Vorrichtung zur Bildsignalkodierung mit einer Klassifizieranlage
DE60203850T2 (de) System und Verfahren zum Verarbeiten von Bildern mit aufgelöster Mosaikstruk- tur, um durch Farb-Aliasing hervorgerufenen Artefakte zu verringern.
DE69321038T2 (de) Kodierung und Dekodierung von digitalen Videodaten
DE68921949T2 (de) System zur Kodierung eines Bildsignals.
DE69818751T2 (de) Dekompression von interpolierten Bildern
DE102009017436A1 (de) Erkennung einer Änderung zwischen Bildern oder in einer Sequenz von Bildern
DE69915843T2 (de) Teilbandkodierung/-dekodierung
DE69309732T2 (de) Verfahren und Vorrichtung ein Bild zu komprimieren und zu dekomprimieren
DE69329439T2 (de) Verfahren und Einrichtung zur Kodierung
DE60006651T2 (de) Verfahren, vorrichtung und verwendung zur bewertung von kodierten bildern
DE69126525T2 (de) Digitale bildverarbeitung mit filterung der blockkanten

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A2

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT TZ UA UG US UZ VN YU ZA ZW

AL Designated countries for regional patents

Kind code of ref document: A2

Designated state(s): GH GM KE LS MW MZ SD SL SZ TZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
AK Designated states

Kind code of ref document: A3

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT TZ UA UG US UZ VN YU ZA ZW

AL Designated countries for regional patents

Kind code of ref document: A3

Designated state(s): GH GM KE LS MW MZ SD SL SZ TZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
WA Withdrawal of international application
REG Reference to national code

Ref country code: DE

Ref legal event code: 8642