CN101409838A - Coding method, decoding method, coding apparatus, decoding apparatus, and image processing system - Google Patents
Coding method, decoding method, coding apparatus, decoding apparatus, and image processing system Download PDFInfo
- Publication number
- CN101409838A CN101409838A CN200810145318.3A CN200810145318A CN101409838A CN 101409838 A CN101409838 A CN 101409838A CN 200810145318 A CN200810145318 A CN 200810145318A CN 101409838 A CN101409838 A CN 101409838A
- Authority
- CN
- China
- Prior art keywords
- information
- motion vector
- reference frame
- motion
- mentioned
- Prior art date
- Legal status (The legal status 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 status listed.)
- Granted
Links
Images
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
A coding method for performing compression/coding of a moving picture by using a motion compensation prediction so as to code or decode motion information in video with a small overhead. The method includes a motion compensation prediction step (processing performed by a motion detector (2), a motion compensation unit (5), and a space prediction unit (9)) for assigning one or more motion vectors (for example, two motion vectors at the most) on basis of sub-blocks into which macro blocks of a moving picture frame are divided, thereby performing a motion compensation prediction and an assigning information coding step (processing performed by a variable length coding unit (23)) for multiplexing information on the motion vector assignment for the sub-blocks with a bit stream and outputting it.
Description
The application is to be on September 12nd, 2002 applying date, and application number is 02803222.5, and denomination of invention is divided an application for " coding method, interpretation method, code device, code translator, image processing system, coded program and translator " application.
Technical field
The present invention relates to use motion compensated prediction carry out coding method, code device and the coded program of the compressed encoding of moving image, decompress(ion) that the use motion compensated prediction carries out the compression movement view data and decoding interpretation method, code translator and translator, comprise above-mentioned code device and code translator and the image processing system that constitutes.
Background technology
As the example of the image encoding mode that has earlier, called optical imaging have based on International standardization work document ITU-T SG16VCEG-M81, " H.26L Test Model LongTerm Number7 (TML-7) " draft0 (
Http:// standard.pictel.comThe image encoding decoded mode of/H.26L light mode ftp/video-site/0104_Aus/VCEG-M81d0.doc) put down in writing.
H.26L in the image encoding,, prepared a lot of motion segmentation figures (pattern) shown in Fig. 4 A~Fig. 4 G at this, can carry out exquisite motion compensation as inter-frame forecast mode.By means of such technology, can catch this activity effectively when the situation of trickle activity taking place, thereby can improve forecasting efficiency in the inside of macro block.
But, to prepare as inter-frame forecast mode in the mode of a lot of motion segmentation figures, the number of motion segmentation is many more, just must transmit many more motion vector informations, thereby expense will heighten.Fig. 5 represents the needed transmission amount of information of each pattern of Fig. 4 A~Fig. 4 G.It is long that " pattern position " shown in Figure 5 expression is used to show the code of variable length code of predictive mode, " minimum movement vector position " expression is as the amount of information about the minimum that should can get by the amount of information of the motion vector of macro block unit transmission, the value of the summation of " minimum needs figure place " expression pattern position and minimum movement vector position.The motion vector position is the total bit of 2 values of level, vertical direction, minimum situation be level with vertical all situations of the vector identical with motion vector predictor (motion vector in the rule of appointment with near motion vector as predicted value, its difference value is encoded).But so usually situation is very rare, always needs many information of this figure place of transfer ratio.As shown in Figure 5, more than pattern 4 or 5, overhead is very many.
On the other hand, in general image, be not to carry out motion detection more than 4 parts with the macro block region segmentation exquisitely, catch it and movable just can carry out abundant exercise and compensate and generally be divided into 2 parts.But, in the such fixing motion segmentation of 16 * 8 pixel units, 8 * 16 pixel units that coded system is H.26L prepared, can not correctly show the motion segmentation of interior macroblocks, the reason of 7 intra prediction modes of Here it is set-up dirgram 4A~Fig. 4 G.
Summary of the invention
The present invention is motion in order to address the above problem, and purpose aims to provide coding method, interpretation method, code device, code translator, image processing system, coded program and the translator that can the moving information in the image be encoded or be deciphered with few expense.
In order to achieve the above object, coding method of the present invention is to use motion compensated prediction to carry out the coding method of the compressed encoding of moving image, it is characterized in that: comprise and distribute one or more motion vectors by the 1st of the frame gained of each being cut apart moving image and then the 2nd unit cutting apart gained and carry out the motion compensated prediction step of motion compensated prediction and will turn to bit stream about the information multichannel to the allocation situation of above-mentioned the 2nd motion vector that carries out and the assignment information coding step exported.
As here the 1st can adopt for example piece of 16 * 16 pixels, as the 2nd, can adopt for example piece of 4 * 4 pixels.In this coding method, distribute one or more motion vectors in motion compensated prediction step with the 1st of the frame gained of each being cut apart moving image and then the 2nd unit cutting apart gained and carry out the motion compensated prediction of motion compensated prediction, at the assignment information coding step, to turn to bit stream about information multichannel and export the allocation situation of above-mentioned the 2nd motion vector that carries out, so, expense that can be by seldom and catch various activities in the 1st reliably, thus can carry out image transmission, record, regeneration effectively.
At this moment, in motion compensated prediction step, can with the 2nd position and number assigned motion vector irrespectively.Promptly, in motion compensated prediction step, with the 1st in the 2nd position and number irrespectively by the 2nd unit assigned motion vector, carry out motion compensated prediction, in the assignment information coding step, can with the 2nd position and about by the information of the motion vector of the 2nd distribution as information use about the allocation situation of motion vector.
In addition, in motion compensated prediction step, also can be according to the unit assigned motion vector of the distribution figure that is predetermined by the 2nd.Promptly, in motion compensated prediction step, by with one or more the 2nd marshalling, according to the distribution figure that is predetermined, unit assigned motion vector by the 2nd, with motion compensated prediction, in the assignment information coding step, the identifying information of above-mentioned distribution figure can be used as the information about the allocation situation of motion vector.
As form according to the distribution figure that is predetermined like this, in motion compensated prediction step, selection will be to a certain group in a plurality of graphical set of being organized into groups in advance by the 2nd motion vector distribution figure of the 1st block unit decision, according to the distribution figure in the graphical set that is included in this selection to the 2nd assigned motion vector, carry out motion compensated prediction, in the assignment information coding step, can adopt the form that to export about distributing the identifying information of figure as motion vector specific from the graphical set of above-mentioned selection by the information of the motion vector allocation situation of the 1st block unit coding.
Owing to distributing the identifying information of figure to export as this few motion vector of data volume about the information of motion vector allocation situation, so, the various activities in the 1st can be caught with seldom expense reliably, thereby image transmission, record, regeneration can be carried out effectively.
In addition, in the assignment information coding step, can adopt the form of distributing the predicted value of the shape decision motion vector of figure also the predicted difference score value of this predicted value to be encoded as motion vector information according to the 1st the motion vector that becomes coded object.
Like this, the predicted value of just distributing the shape decision motion vector of figure according to motion vector, and the predicted difference score value of this predicted value that data volume is few is encoded as motion vector information, so, various activities in the 1st can be caught reliably by few expense, thereby image transmission, record, regeneration can be carried out effectively.
Here, in the assignment information coding step, the prediction candidate position of distributing the shape decision motion vector of figure according to the 1st the motion vector that becomes coded object, to predict the some of candidate position according to the shape that is arranged near the 1st the distribution figure that becomes the 1st of above-mentioned coded object, and the predicted difference score value of this predicted value will be encoded as motion vector information as predicted value.
In addition, as the form that meets the distribution figure that is predetermined, in the motion-vector prediction step, the predicted value of distributing the shape decision motion vector of figure according to the 1st the motion vector that becomes coded object, setting is the motion vector exploration window at center with this predicted value, detect motion vector, in the assignment information coding step, can adopt the form that the predicted difference score value of the motion vector of above-mentioned detection and above-mentioned motion vector predictor is encoded as motion vector information.
Like this, the motion vector of the detection that data volume is few and the predicted difference score value of motion vector predictor are encoded as motion vector information, so, can catch various activities in the 1st reliably by few expense, thereby can carry out image transmission, record, regeneration effectively.
In the various forms that meets the above-mentioned distribution figure that is predetermined, in motion compensated prediction step, get and to be 2 to the maximum by the motion vector number that the 2nd block unit distributes, simultaneously according to the some figures in a plurality of distribution figures that are predetermined of the 2nd the distribution state of expression in the 1st that is included in by the decision of the 1st block unit to the 2nd assigned motion vector, carry out motion compensated prediction, in the assignment information coding step, just adopt the form of exporting as the identifying information of above-mentioned distribution figure about by the information of the allocation situation of the motion vector of the 1st block unit decision.
At this moment, distribute figure comprise with above-mentioned the 1st in the horizontal direction or the inhomogeneous image of cutting apart of vertical direction and with above-mentioned the 1st figure of cutting apart at incline direction.
In addition, in the assignment information coding step, distribute the shape of figure to distribute figure to distribute the code of variable length, the identifying information of above-mentioned distribution figure is encoded each according to each.
In addition, in the assignment information coding step, according to being positioned at the situation that becomes near the 1st distribution figure the 1st of coded object, for becoming the 1st of above-mentioned coded object, distribute figure to distribute the code of variable length to each, the identifying information of above-mentioned distribution figure is encoded.
In addition, in the assignment information coding step, distribute figure to distribute the code of variable length to each, the identifying information of above-mentioned distribution figure is encoded according to picture material.
In the form of distributing, can and then comprise that the predicted residual signal that will obtain is by the residual signals coding step of encoding with above-mentioned the 2nd measure-alike block unit in motion compensated prediction step at the various forms that meets the above-mentioned distribution figure that is predetermined with the 2nd location independent ground.
That is, can adopt comprise the motion compensated prediction step of carrying out motion compensated prediction so that the 2nd the unit of the frame of moving image being cut apart back each the 1st of gained and then being cut apart gained is distributed one or more motion vectors, will turn to bit stream about information multichannel to the allocation situation of above-mentioned the 2nd motion vector and the assignment information coding step exported and the predicted residual signal that will obtain in motion compensated prediction step by the form of the residual signals coding step of encoding with above-mentioned the 2nd measure-alike block unit.At this moment, because the predicted residual signal that will obtain in motion compensated prediction step is by encoding with the 2nd measure-alike block unit, so, the coded data amount can be suppressed for seldom, various activities in the 1st can be caught reliably by few expense, thereby image transmission, record, regeneration can be carried out effectively.
In addition, in the forms of irrespectively distributing at the various forms that meets the distribution figure that is predetermined with the 2nd position and number, in the assignment information coding step, can adopt allocation situation will be identified in the form of encoding by the information whether predicted residual signal that should encode is arranged in one or more the 2nd motion vector range of distribution that constitutes according to above-mentioned motion vector.At this moment, allocation situation according to motion vector will be identified in by the information the coding whether predicted residual signal that should encode is arranged in one or more the 2nd motion vector range of distribution that constitutes, so, the coded data amount can be suppressed for seldom, various activities in the 1st can be caught reliably by few expense, thereby image transmission, record, regeneration can be carried out effectively.
In addition, in the forms of irrespectively distributing with the 2nd position and number, in the assignment information coding step, can adopt the information conduct that represent change point that the motion vector under this allocation situation distributes according to the allocation situation that becomes the 1st of coded object and adjacent the 1st motion vector about multichannel turns to the form that bit stream is exported with the information of the allocation situation of above-mentioned motion vector.
In addition, in the forms of irrespectively distributing at the various forms that meets the distribution figure that is predetermined with the 2nd position and number, in motion compensated prediction step, except setting by the 1st block unit the situation of inter-frame forecast mode or intra prediction mode, comprise the situation of setting inter-frame forecast mode or intra prediction mode by by one or more units that constitute of the 2nd, can adopt the form of carrying out motion compensated prediction.
In order to achieve the above object, coding method of the present invention is to use motion compensated prediction to carry out the coding method of the compressed encoding of moving image, it is characterized in that: comprise by by selecting 1 reference frame to carry out the compensation prediction step of motion compensated prediction in a plurality of reference frame that one or more units that constitute of the 2nd of gained keep in order to carry out motion compensated prediction and will turn to bit stream about the information multichannel to the choice situation of above-mentioned one or more reference frame of the 2nd and the reference frame information coding step exported the frame of moving image being cut apart back each the 1st of gained and then being cut apart.
According to this coding method, about information to the choice situation of one or more reference frame of the 2nd, data volume is few, so, the coded data amount can be suppressed for seldom, various activities in the 1st can be caught reliably by few expense, thereby image transmission, record, regeneration can be carried out effectively.
In reference frame information coding step, can adopt the form of encoding as to the combined information of the choice situation of above-mentioned one or more reference frame of the 2nd of in above-mentioned the 1st unit, comprising.
In addition, in reference frame information coding step, ask predicted value respectively according near the choice situation of above-mentioned the 1st or one or more above-mentioned the 2nd reference frame being positioned at, can adopt the difference information of the reference frame of this predicted value and selection as form about the information use of the choice situation of reference frame to one or more reference frame of the 2nd.
In order to achieve the above object, coding method of the present invention is to use motion compensated prediction to carry out the coding method of the compressed encoding of moving image, it is characterized in that: comprise with by the motion compensated prediction step of the prediction direction of selecting bidirectional interframe predictives to use by one or more the 2nd unit that constitute of the frame of moving image being cut apart back each the 1st of gained and then being cut apart gained being carried out motion compensated prediction with will turn to bit stream about information multichannel and the information of forecasting coding step exported the choice situation of above-mentioned one or more prediction direction of the 2nd.
According to this coding method, can catch various activities in the 1st reliably by few expense, can transmit, write down, regenerate thereby can carry out image effectively respectively to catching one or more the 2nd two-way infra-frame prediction that application is suitable of other activity simultaneously.
In order to achieve the above object, coding method of the present invention is to use motion compensated prediction to carry out the coding method of the compressed encoding of moving image, it is characterized in that: comprise with by to the frame of moving image being cut apart back each the 1st of gained and then being cut apart the selecting inter-frame forecast modes or intra prediction mode to carry out the prediction steps of inter prediction or infra-frame prediction and will turn to bit stream about information multichannel and the prediction mode information coding step exported of gained the choice situation of above-mentioned one or more the 2nd inter-frame forecast modes or intra prediction mode by one or more the 2nd unit that constitute.
According to this coding method, can catch various activities in the 1st reliably by few expense, can transmit, write down, regenerate thereby can carry out image effectively respectively to catching one or more the 2nd inter-frame forecast mode or the intra prediction mode that application is suitable of other activity simultaneously.
In above-mentioned information of forecasting coding step or prediction mode information coding step, can adopt with above-mentioned information about choice situation with about to the information of the allocation situation that is included in above-mentioned the 1st motion vector in above-mentioned the 1st unit as combination to encoding and multichannel turns to bit stream and the form exported.
In addition, in above-mentioned information of forecasting coding step or prediction mode information coding step, can adopt will be about the information of above-mentioned choice situation with about encoding as combined information to the information of the choice situation of the reference frame that is included in above-mentioned one or more the 2nd a plurality of reference frame in above-mentioned one or more the 2nd block units and multichannel turns to bit stream and the form exported.
In order to achieve the above object, interpretation method of the present invention is to use motion compensated prediction to carry out the decompress(ion) of compression movement view data, the interpretation method of decoding is characterized in that: thus comprise expression by to each the 1st of gained behind the frame of cutting apart moving image and then to cut apart motion vector that the 2nd unit of gained distributes be assignment information decoding step of deciphering by which the motion vector assignment information in one or more motion vectors of above-mentioned the 1st unit decoding and the motion compensation step of using to carry out motion compensation generation forecast image by the specific motion vector of above-mentioned the 2nd unit according to this motion vector assignment information.
In this interpretation method, to represent by each the 1st of gained after to the frame of cutting apart moving image and then to cut apart motion vector that the 2nd unit of gained distributes be to decipher by which the motion vector assignment information in one or more motion vectors of above-mentioned the 1st unit decoding in assignment information decoding step, thereby carry out motion compensation generation forecast image according to this motion vector assignment information use by the specific motion vector of the 2nd unit in motion compensation step.Therefore, will decipher, and use according to this motion vector assignment information and carry out motion compensation, the generation forecast image by the specific motion vector of the 2nd unit about the motion vector assignment information of the motion vector that distributes by the 2nd unit.Like this, can decipher reliably, thereby can carry out image transmission, record, regeneration effectively by few expense.
At this moment, in assignment information decoding step, can adopt with the 2nd position distributing object and number irrespectively as the 2nd position and distribute to the 2nd motion vector information and with the form of above-mentioned motion vector assignment information decoding.
In addition, in assignment information decoding step, by with one or more pieces marshallings, can adopt as the identifying information of the distribution figure that is predetermined and the form of above-mentioned motion vector assignment information decoding.
Adopt like this as the identifying information of the distribution figure that is predetermined and with the form of motion vector assignment information decoding, in assignment information decoding step, with above-mentioned motion vector assignment information as distributing the information of figure to decipher by what the 1st block unit was predetermined to the 2nd motion vector as identification, when distributing the specific distribution figure of figure identifying information according to this, information decoding with the graphical set under this distribution figure of identification, can adopt according to definition, according to the form of above-mentioned distribution figure identifying information decision to the 2nd motion vector distribution by the specific graphical set of this graphical set identifying information.At this moment, the motion vector assignment information as the information of distributing figure to the 2nd motion vector that identification is predetermined, can be deciphered reliably by few expense, thereby can be carried out image transmission, record, regeneration effectively.
In addition, the form that adopts as the identifying information of the distribution figure that is predetermined and the motion vector assignment information is deciphered, in assignment information decoding step, the form that the motion vector information that can adopt the predicted difference score value of predicted value of the motion vector of shape decision that will be by distributing figure according to the 1st the motion vector that becomes the decoding object to encode to obtain is deciphered.At this moment, because just the predicted difference score value of the predicted value of the motion vector of the shape decision by distributing figure according to the motion vector that becomes decoding the 1st of the object motion vector information that obtains of encoding is deciphered, so, can decipher by the motion vector information that the predicted difference score value coding that data volume is few obtains, thereby can carry out image transmission, record, regeneration effectively.
At this moment, in assignment information decoding step, to distribute motion-vector prediction candidate position some of the shape decision of figure according to the 1st the motion vector that becomes the decoding object according near the shape of the 1st distribution figure being arranged in, to deciphering by the motion vector information that obtains that the predicted difference score value of this predicted value is encoded as predicted value.
And in the form of deciphering as the identifying information of the above-mentioned distribution figure that is predetermined, to be taken as 2 by the 1st their motion vector maximum of unit, the motion vector assignment information as identification expression by deciphering that the 1st of the 1st block unit decision comprises to the information of a plurality of distribution figures motion vector distribution state, that be predetermined of the 2nd.
At this moment, distribute figure comprise with above-mentioned the 1st in the horizontal direction or the inhomogeneous figure of cutting apart of vertical direction and with above-mentioned the 1st figure of cutting apart at incline direction.
In addition, in assignment information decoding step, will distribute the identifying information of figure to decipher as each according to each code that distributes the shape of figure to distribute to the variable length that respectively distributes figure.
In addition, in assignment information decoding step, will distribute to the code of the variable length that respectively distributes figure and decipher according to being positioned at the situation that becomes near the 1st distribution figure the 1st of decoding object as the 1st the identifying information that respectively distributes figure that becomes the decoding object.
In addition, in assignment information decoding step, will distribute the identifying information of figure to decipher as each according to the code that picture material is distributed to the variable length that respectively distributes figure.
The form of deciphering as the identifying information of the above-mentioned distribution figure that is predetermined and with the 2nd position and number irrespectively in the forms with the decoding of motion vector assignment information, and then comprise the residual signals decoding step of deciphering by the predicted residual signal of the motion compensated prediction that obtains according to the encoding process of carrying out with the 2nd measure-alike block unit.
That is, thus can adopt and comprise expression by to each the 1st of gained behind the frame of cutting apart moving image and then to cut apart motion vector that the 2nd unit of gained distributes be the assignment information decoding procedure deciphered by which the motion vector assignment information in one or more motion vectors of above-mentioned the 1st unit decoding, use by above-mentioned the 2nd specific motion vector of unit according to this motion vector assignment information and carry out the motion compensation step of motion compensation generation forecast image and the structure of the residual signals decoding procedure that will decipher by the predicted residual signal of processing the motion compensated prediction that obtains according to the coding that carries out with the 2nd measure-alike block unit. At this moment, owing to just decipher by the predicted residual signal of the motion compensated prediction that obtains according to the encoding process of carrying out with the 2nd measure-alike block unit, so, can will answer the few predicted residual signal decoding of coded data amount, thereby can carry out image transmission, record, regeneration effectively.
In addition, the form of deciphering as the identifying information of the above-mentioned distribution figure that is predetermined and with the 2nd position and number irrespectively in the forms with the decoding of motion vector assignment information, in assignment information decoding step, will distribute the shape recognition of figure in by one or more the 2nd motion vector range of distribution that constitute, whether to have the information of the predicted residual signal that should decipher to decipher according to the 1st the motion vector that becomes the decoding object.At this moment, can will answer the few predicted residual signal decoding of coded data amount, thereby can carry out image transmission, record, regeneration effectively.
In addition, with above-mentioned the 2nd position and number irrespectively in the forms with the decoding of motion vector assignment information, in assignment information decoding step, the information of the change point that will distribute as the expression motion vector of exporting about the information of the allocation situation of motion vector is deciphered.
In addition, the form of deciphering as the identifying information of the above-mentioned distribution figure that is predetermined and with the 2nd position and number irrespectively in the form with the decoding of motion vector assignment information, motion vector be except the situation of setting inter-frame forecast mode or intra prediction mode by the 1st block unit by also comprising the motion vector that obtains with motion compensated prediction by the situation of the 2nd one or more units that constitute setting inter-frame forecast modes or intra prediction mode.
In order to achieve the above object, interpretation method of the present invention is to use motion compensated prediction to carry out the decompress(ion) of compression movement view data and the interpretation method of decoding, it is characterized in that: thus comprise with about with by the frame of moving image is cut apart back each the 1st of gained so that cut apart reference frame information decoding step that the information of the choice situation of the reference frame that motion compensated prediction that one or more units that constitute of the 2nd of gained distribute uses deciphers and according to this reference frame information from selecting 1 reference frame to carry out the compensation process of motion compensation generation forecast image a plurality of reference frame that motion compensation was kept in order to carry out.
According to this form, about the information to the choice situation of above-mentioned one or more reference frame of the 2nd, data volume is few, so, information decoding that can data volume is few about the choice situation of this reference frame, thus can carry out image transmission, record, regeneration effectively.
At this moment, in reference frame information decoding step, will decipher above-mentioned one or more reference frame information of the 2nd according to combined information to the choice situation that is included in above-mentioned one or more the 2nd the reference frame in above-mentioned the 1st unit.
In addition, in reference frame information decoding step, ask predicted value respectively according near the choice situation of above-mentioned the 1st or one or more above-mentioned the 2nd reference frame being positioned at, according to deciphering as the difference information and the above-mentioned predicted value of obtaining of the reference frame of exporting about the information of the choice situation of reference frame to one or more reference frame of the 2nd.
In order to achieve the above object; Interpretation method of the present invention is to use motion compensated prediction to carry out the decompress(ion) of compression movement view data and the interpretation method of decoding, it is characterized in that: thus comprise with about with by the frame of moving image is cut apart each the 1st of rear gained so that cut apart information of forecasting decoding procedure that the information of the choice situation of the prediction direction that two-way infra-frame prediction that one or more units that consist of of the 2nd of gained distribute uses deciphers and according to this information of forecasting from the prediction direction of selecting two-way inter motion compensation to use the reference frame that forward keeps with reverse two aspects being carried out the motion compensation step of motion compensation generation forecast image.
According to this form, can catch various activities in the 1st reliably by few expense, can transmit, write down, regenerate thereby can carry out image effectively catching one or more the 2nd bidirectional interframe predictive that application is suitable of other activity respectively simultaneously.
In order to achieve the above object, interpretation method of the present invention is to use motion compensated prediction to carry out the decompress(ion) of compression movement view data and the interpretation method of decoding, it is characterized in that: thus comprise about with by the frame of moving image being cut apart back each the 1st of gained and then being cut apart inter-frame forecast mode that one or more units that constitute of the 2nd of gained distribute or the prediction mode information decoding step that the information of the choice situation of intra prediction mode is deciphered and the predicted picture of selecting inter-frame forecast mode or intra prediction mode to carry out inter prediction or infra-frame prediction generation forecast image according to this information of forecasting generate step.
According to this form, can catch various activities in the 1st reliably by few expense, can transmit, write down, regenerate thereby can carry out image effectively respectively to catching one or more the 2nd infra-frame prediction or the infra-frame prediction that application is suitable of other activity simultaneously.
In information of forecasting decoding step or prediction mode information decoding step, can adopt according to will about be included in above-mentioned the 1st unit will be to the information of the allocation situation of above-mentioned the 1st motion vector and information about the form that the information of above-mentioned one or more above-mentioned choice situations of the 2nd is deciphered about the information combination of above-mentioned choice situation.
In addition, in information of forecasting decoding step or prediction mode information decoding step, can adopt according to will be about will be about the form that the information of above-mentioned one or more above-mentioned choice situations of the 2nd is deciphered the information of the choice situation that is included in the reference frame in above-mentioned one or more the 2nd a plurality of reference frame in above-mentioned one or more units of the 2nd and information about the information combination of above-mentioned choice situation.
The present invention beyond invention as coding method, interpretation method, also can catch as the code device that uses coding method, use interpretation method code translator, comprise these code devices and code translator and each invention such as the image processing system that constitutes, the coded program of carrying out by code device and the translator carried out by code translator and obtain same effect and effect.That is, can carry out description shown below.
In order to achieve the above object, code device of the present invention is to use motion compensated prediction to carry out the code device of the compressed encoding of moving image, it is characterized in that: have by the 1st of the frame gained of each componental movement image and then the 2nd unit cutting apart gained are distributed one or more motion vectors and carry out the motion compensated prediction unit of motion compensated prediction and will turn to bit stream about the information multichannel to the allocation situation of above-mentioned the 2nd motion vector and the assignment information coding unit exported.
At this moment, can adopt the 2nd position in motion compensated prediction unit and the 1st and number irrespectively to carry out motion compensated prediction by the 2nd unit assigned motion vector and the assignment information coding unit with the 2nd position with distribute to the 2nd the information of motion vector as form about the information use of the allocation situation of motion vector.
In addition, can adopt the motion compensated prediction unit according to by with 1 give or distribution figure that a plurality of the 2nd marshalling is predetermined carry out motion compensated prediction by the 2nd unit assigned motion vector and the assignment information coding unit with the identifying information of above-mentioned distribution figure as form about the information use of the allocation situation of motion vector.
In order to achieve the above object; Code translator of the present invention is to use motion compensated prediction to carry out the decompress(ion) of compression movement view data and the code translator of decoding, it is characterized in that: thus have expression by to each the 1st of gained behind the frame of cutting apart moving image and then to cut apart motion vector that the 2nd unit of gained distributes be the assignment information decoding unit of deciphering by which the motion vector assignment information in one or more motion vectors of above-mentioned the 1st unit decoding and the motion compensation units of using to carry out motion compensation generation forecast image by above-mentioned the 2nd specific motion vector of unit according to this motion vector assignment information.
At this moment, can adopt the assignment information decoding unit as the 2nd position distributing object with distribute to the 2nd the information of motion vector and form that above-mentioned motion vector assignment information is deciphered.
In addition, also can adopt the assignment information decoding unit as form by deciphering with the identifying information of one or more the 2nd distribution figure that be predetermined of marshalling and with above-mentioned motion vector assignment information.
In order to achieve the above object; Image processing system of the present invention be comprise use motion compensated prediction carry out moving image compressed encoding code device and use motion compensated prediction to carry out the code translator of the decompress (ion) of compression movement view data and decoding and the image processing system that consists of; It is characterized in that: code device has the motion compensated prediction unit that carries out motion compensated prediction by the 1st of the frame gained of each componental movement image and then the 2nd unit cutting apart gained are distributed one or more motion vectors and will turn to bit stream about the information multichannel to the allocation situation of above-mentioned the 2nd motion vector and the assignment information coding unit exported, is the assignment information decoding unit of deciphering by which the motion vector assignment information in one or more motion vectors of above-mentioned the 1st unit decoding and the motion compensation units of using to carry out motion compensation generation forecast image by above-mentioned the 2nd specific motion vector of unit according to this motion vector assignment information thereby code translator has expression by the motion vector that each the 1st of gained behind the frame of cutting apart moving image and then the 2nd unit cutting apart are distributed.
At this moment, can adopt the 2nd position in the motion compensated prediction unit of code device and the 1st and number irrespectively to carry out motion compensated prediction and the assignment information coding unit uses the 2nd position and the information of distributing to the 2nd motion vector as the information about the allocation situation of motion vector by the 2nd unit assigned motion vector, the assignment information decoding unit of code translator as the 2nd position distributing object with distributing to the 2nd the information of motion vector form that above-mentioned motion vector assignment information is deciphered.
In addition, the motion compensated prediction unit that also can adopt code device according to carrying out motion compensated prediction by the 2nd unit assigned motion vector by the distribution figure that one or more the 2nd marshalling is predetermined the assignment information coding unit with the identifying information of above-mentioned distribution figure as the information about the allocation situation of motion vector use, the assignment information decoding unit of code translator is as the form of above-mentioned motion vector assignment information being deciphered by the identifying information of organizing into groups the distribution figure that is predetermined with one or more the 2nd.
In order to achieve the above object, coded program of the present invention is characterised in that: ensconce in making and use motion compensated prediction to carry out each step that computer in the code device of compressed encoding of moving image is carried out each invention of above-mentioned coding method.
In order to achieve the above object, translator of the present invention is characterised in that: ensconce in making and use motion compensated prediction to carry out each step that computer in the code translator of the decompress(ion) of compression movement view data and decoding is carried out each invention of above-mentioned interpretation method.
Above-mentioned purpose of the present invention and other features and advantage are read following detailed explanation by the reference accompanying drawing and can be understood.
The simple declaration of accompanying drawing
Fig. 1 is the structure chart of code device.
Fig. 2 is the structure chart of code translator.
Fig. 3 A is the figure of content of the employed former frame of explanation of expression motion compensation infra-frame prediction.
Fig. 3 B is the figure of content of the employed present frame of explanation of expression motion compensation infra-frame prediction.
Fig. 4 A is the figure that represents the pattern 1 of inter-frame forecast mode H.26L.
Fig. 4 B is the figure that represents the pattern 2 of inter-frame forecast mode H.26L.
Fig. 4 C is the figure that represents the mode 3 of inter-frame forecast mode H.26L.
Fig. 4 D is the figure that represents the pattern 4 of inter-frame forecast mode H.26L.
Fig. 4 E is the figure that represents the pattern 5 of inter-frame forecast mode H.26L.
Fig. 4 F is the figure that represents the pattern 6 of inter-frame forecast mode H.26L.
Fig. 4 G is the figure that represents the mode 7 of inter-frame forecast mode H.26L.
Fig. 5 is the table of expression to the transmission amount of information of the needed minimum of each pattern of intra prediction mode.
Fig. 6 A is the 1st group figure of the fixedly partitioning scheme of the expression embodiment of the invention.
Fig. 6 B is the 2nd group figure of the fixedly partitioning scheme of the expression embodiment of the invention.
Fig. 6 C is the 3rd group figure of the fixedly partitioning scheme of the expression embodiment of the invention.
Fig. 6 D is the 4th group figure of the fixedly partitioning scheme of the expression embodiment of the invention.
Fig. 7 A be the present macro block of expression be pattern 1 and its left side, go up, figure that upper right macro block is respectively the situation of pattern 0,1,8.
Fig. 7 B is the figure that the macro block in the expression uses the situation of the reference frame different with present macro block.
Fig. 8 A is the figure that is used to illustrate the flow process that the motion vector predictor decision under the situation of pattern 0 is handled.
Fig. 8 B is the figure that is used to illustrate the flow process that the motion vector predictor decision under the situation of pattern 1 is handled.
Fig. 8 C is the figure that is used to illustrate the flow process that the motion vector predictor decision under the situation of mode 7 is handled.
Fig. 9 is that expression is considered and near the continuity of the activity of the macro block table to an example of the top relative importance value of each predictive mode definition and left priority degree.
The figure of the predicted value of Figure 10 A motion vector that to be expression provide regularly to fixing each predictive mode of the 1st group of partitioning scheme.
The figure of the predicted value of Figure 10 B motion vector that to be expression provide regularly to fixing each predictive mode of the 2nd group of partitioning scheme.
The figure of the predicted value of Figure 10 C motion vector that to be expression provide regularly to fixing each predictive mode of the 3rd group of partitioning scheme.
The figure of the predicted value of Figure 10 D motion vector that to be expression provide regularly to fixing each predictive mode of the 4th group of partitioning scheme.
Figure 11 A is the figure of grammer of the macro block data of the compressive flow in representing H.26L.
Figure 11 B is the figure of grammer of macro block data of the compressive flow of expression embodiment 1.
Figure 12 is the flow chart of contents processing of the code translator of expression embodiment 1.
Figure 13 A is the figure of the 1st definition example of the predictive mode that shows as 4 * 4 binary map information in order to specify allocation situation arbitrarily of expression.
Figure 13 B is the figure of the 2nd definition example of the predictive mode that shows as 4 * 4 binary map information in order to specify allocation situation arbitrarily of expression.
Figure 13 C is the figure of the 3rd definition example of the predictive mode that shows as 4 * 4 binary map information in order to specify allocation situation arbitrarily of expression.
Figure 14 is the figure of order of the motion vector detection of expression embodiment 2.
Figure 15 is the figure of other order of expression motion vector detection.
Figure 16 A is the figure that is used for illustrating the example that the continuity of activity of coding method macro block A of prediction mode information of embodiment 2 is high.
Figure 16 B is the figure that is used for illustrating the example that the continuity of activity of coding method macro block B of prediction mode information of embodiment 2 is high.
Figure 17 is the flow chart of content of the encoding process of expression embodiment 2.
Figure 18 A is the table of the example of the expression long CB of code (4).
Figure 18 B is the table of the example of the expression long CB of code (3).
Figure 18 C is the table of the example of the expression long CB of code (2).
Figure 19 A is the figure of grammer of the macro block data of the compressive flow in representing H.26L.
Figure 19 B is the figure of the simplest grammer example of macro block data of the compressive flow of expression embodiment 2.
Figure 19 C is the figure of other grammer example of macro block data of the compressive flow of expression embodiment 2.
Figure 20 is the flow chart of contents processing of the code translator of expression embodiment 2.
Figure 21 is the flow chart of contents processing of the code device of expression embodiment 3.
Figure 22 is the key diagram of contents processing of the code device of embodiment 3.
Figure 23 A is the figure of example of the saw-tooth sweep of expression scan table.
Figure 23 B is the figure of example of the horizontal sweep of scan table.
Figure 23 C is the figure of example of the Hilbert scanning of expression scan table.
Figure 24 A is illustrated in embodiment 4 the angular direction is carried out the difference of inside/centre and use the figure of definition example of the predictive mode of 2 motion vectors in inside.
Figure 24 B is the figure of the definition example of the predictive mode predicted by the centre by the bottom, inside for two parts up and down, top of the macroblock partition of expression embodiment 4.
Figure 25 A is the figure on the plane of expression inside/middle classification.
Figure 25 B is the figure on the plane of expression intra-frame prediction method.
Figure 25 C is the figure on the plane of expression motion vector distribution.
Figure 26 A is the figure on the plane of expression inside/middle classification.
Figure 26 B be only represent will be with inner, middle corresponding sub-piece the figure on plane of intra-frame prediction method during as bit-plane data.
Figure 26 C be only represent will be with inner, middle corresponding sub-piece the figure on plane of motion vector distribution during as bit-plane data.
Figure 27 is the figure that the example of conversion is carried out the situation of Figure 24 A by the coding method 2 of embodiment 4 in expression.
Figure 28 is the figure of grammer of the macro block data of the expression compressive flow corresponding with the coding method 1 of embodiment 4.
Figure 29 is the figure of grammer of the macro block data of the expression compressive flow corresponding with the coding method 2 of embodiment 4.
Figure 30 is the figure of grammer of the macro block data of the expression compressive flow corresponding with the coding method 3 of embodiment 4.
Figure 31 is the figure of grammer of the macro block data of the expression compressive flow corresponding with the coding method 4 of embodiment 4.
Figure 32 A is the figure that the definition example that the predictive mode of minimal predictive mode sets is used in expression.
Figure 32 B is the figure that the definition example that the predictive mode of few motion vector of prompting in embodiment 1 sets is used in expression.
Figure 32 C is the figure of the definition example set of the most predictive modes of cutting apart most motion vectors of expression.
Figure 33 A is the table of the code structure of the reference frame information in representing H.26L.
Figure 33 B is the table of expression to the example of the combination assigned code of reference frame information.
Figure 33 C is the figure of the grammer of the expression macro block data corresponding with Figure 33 B.
Figure 34 A is to represent before changing or is the table of the code that switched according to predicted value in 0 o'clock in predicted value.
Figure 34 B is that the expression predicted value is 4 o'clock tables according to the code of predicted value switching.
Figure 35 A is the figure of cutting apart example of efficient aspect macro block of the coding of expression explanation orthogonal transformation coefficient data.
Figure 35 B is expression has or not the example of meaningful coefficient by the transmission of 8 * 8 block units figure.
Figure 35 C is that expression is to cutting apart the figure that the field transmission has or not the example of meaningful coefficient.
Figure 36 A is the figure that the motion vector detection of motion vector when exploring the adaptation of scope handled that carries out that is used to illustrate embodiment 5.
Figure 36 B is used to illustrate that embodiment's 5 does not carry out the figure that the motion vector detection of motion vector when exploring the adaptation of scope handled.
Figure 37 A be used to illustrate embodiment 5 about the fixing figure of the 1st group motion-vector prediction rule of partitioning scheme.
Figure 37 B be used to illustrate embodiment 5 about the fixing figure of the 2nd group motion-vector prediction rule of partitioning scheme.
Figure 37 C be used to illustrate embodiment 5 about the fixing figure of the 3rd group motion-vector prediction rule of partitioning scheme.
Figure 37 D be used to illustrate embodiment 5 about the fixing figure of the 4th group motion-vector prediction rule of partitioning scheme.
Figure 38 is the flow chart that the motion vector detection of expression embodiment 5 is handled.
Figure 39 is the figure of the bi-directional predicted example of expression embodiment 6.
Figure 40 A is the figure of the grammer in bi directional motion compensation when prediction in representing H.26L.
Figure 40 B is the figure of grammer of macro block data of the compressive flow of expression embodiment 6.
Figure 41 is the code table of the macro block type information of embodiment 6.
Figure 42 is the code table of the reference frame information of embodiment 6.
Figure 43 is the flow chart of the basic action of presentation code device.
Figure 44 is the flow chart of the basic action of expression code translator.
Embodiment
Below, order illustrates various embodiment of the present invention.
At first, in embodiments of the invention 1, illustrate and undertaken by macro block unit in the image encoded code device being that each sub-piece maximums of 4 * 4 pixel units can distribute 2 motion vectors to carry out the picture coding device and the code translator of motion compensated prediction with macroblock partition.
At first, illustrate based on the image encoding of using H.26L coded system of the present invention and the structure and the action of code translator.
The structure of Fig. 1 presentation code device, Fig. 2 represents the structure of code translator.In the code device of Fig. 1, cut down the redundancy that time orientation exists by motion compensation interframe prediction, by orthogonal transform and then the redundancy of cutting down the remnants that direction in space exists, carry out the Information Compression of picture signal.The key diagram of motion compensation interframe prediction is shown in Fig. 3 A and Fig. 3 B respectively, and the summary of piece matching treatment is shown in Fig. 4 A~Fig. 4 G respectively.In addition, Figure 43 is the flow chart of the basic action of presentation code device, and Figure 44 is the flow chart of the basic action of expression code translator.
Below, with reference to the coding of these figure key diagrams 1 and Fig. 2 and the action of code translator.
The received image signal 1 of Fig. 1 is the time series of two field picture, after, as the signal of expression two field picture unit.In addition, be configured to the present frame 30B of the two field picture of coded object for Fig. 3 B.Frame 30B is divided into 16 pixels * fixing square area (being called macro block) of 16 row now, carries out following encoding process by this unit.
Now each macro block of frame 30B transmits to motion detection portion 2 earlier, carries out the detection of the motion vector 4 of the decision of predictive mode 3 and the bar number corresponding with it at this.In coded system H.26L, as predictive mode 3, having prepared a plurality of uses in advance has finished intra prediction mode that near the pixel of the coding on the same space carries out spatial prediction and has used the different in time two field pictures of finishing coding (the former frame 30A of Fig. 3 A) to carry out the inter-frame forecast mode of motion compensated prediction, character according to the local place of picture signal is switched them by macro block unit, thereby can carry out effective Information Compression.
That is, in the flow chart of Figure 43, received image signal 1 input coding device (S01) carries out motion detection (S02) and motion compensation (S03), simultaneously, utilizes spatial prediction portion 9 to carry out the described spatial prediction in back (S04).
So-called motion compensated prediction, shown in Fig. 3 A, Fig. 3 B, in the exploration scope 32 of appointment, explore and the present similar signal graph of input picture (predicted value) of frame 30B according to the decoding image of former frame 30A, the displacement in space (is for example detected as motion vector, the displacement in the space of circle 31 is detected as motion vector 33), be the technology that predicted residual signal and motion vector are transmitted as coded message.In H.26L, shown in Fig. 4 A~Fig. 4 G, 7 kinds of intra prediction modes are arranged.Say that correctly the predictive mode (dancing mode) of the pixel of the useful same position of directly duplicating former frame still, has omitted for convenience of explanation when also having prepared image freeze etc. herein.The pattern 1 of Fig. 4 A detects motion vector by the mode 7 of 8 * 4 pixel units, Fig. 4 G by 4 * 4 pixel units by the pattern 6 of 4 * 8 pixel units, Fig. 4 F by the pattern 5 of 8 * 8 pixel units, Fig. 4 E by the pattern 4 of 16 * 8 pixel units, Fig. 4 D by the mode 3 of 8 * 16 pixel units, Fig. 4 C by the pattern 2 of 16 * 16 pixel units, Fig. 4 B.That is, these 7 kinds of predictive modes can segment the motion detection unit in the macro block, are provided with in order correctly to catch the comings and goings that may exist in macro block.
In motion detection portion 2, by each macro block all inter-frame forecast modes of Fig. 4 A~Fig. 4 G are estimated, select to judge the highest pattern of code efficiency.Intra prediction mode equally also carries out the evaluation of code efficiency, just selects intra prediction mode when efficient is better than inter-frame forecast mode.When having selected inter-frame forecast mode, just detection and transmission are by bar number (16 of maximum every macro blocks) the transmitting moving vector 4 of predictive mode 3 decisions.When having selected intra prediction mode, motion vector 4 does not transmit.
When having selected inter-frame forecast mode,, use the motion compensated prediction image 8 of finishing encoded image signal 7 each macro block of taking-up of the former frame in the motion vector 4 reference frame memories 6 to the predictive mode 3 and the motion vector 4 of dynamic compensating unit 5 transmission decisions.
When having selected intra prediction mode, in spatial prediction portion 9, with reference near the encoded image signal 7 generation forecast values 10 of finishing.The final predicted value of using 11 is selected some in motion compensated prediction images 8 and the predicted value 10 according to predictive mode 3 by switch 12.
Generate the predicted residual signal 13 (S05 of Figure 43) of the difference value of above-mentioned predicted value 11 of expression and received image signal 1, predicted residual signal 13 becomes orthogonal transform coefficient 15 (S06) to 14 transmission of orthogonal transform portion by orthogonal transform.16 transmission of orthogonal transform coefficient 15 vectorization portions quantize orthogonal transform coefficient 17 (S07) by quantizing to be treated as, and to variable-length encoding part 23 transmission.Carry out entropy coding at this, and multichannel turns to compressive flow 24 (S08).
On the other hand, though in Figure 43, omitted,, quantize orthogonal transform coefficient 17 and carry out partial decode through re-quantization portions 18 and inverse orthogonal transformation portion 19, and with predicted value 11 additions, become partial decode image 20.Partial decode image 20 is in order to remove in deteriorations such as the piece distortion of the boundary of orthogonal transform block and carry out Filtering Processing by 21 pairs of boundary pixels of loop filter, becomes partial decode image 22 after the filtering, and stores in the frame memory 6.The view data that is stored in the frame memory 6 is used as the comparable data 7 of the motion compensated prediction of later frame.
In addition, predictive mode 3 also carries out entropy coding by the unit of macro block by variable-length encoding part 23, and multichannel turns to compressive flow 24 (S08).Then, export the compressive flow 24 (S09) that obtains like this.
Below, the action of code translator is described according to Fig. 2 and Figure 44.In code translator, reception as the compressive flow 24 (S21 of Figure 44) of coded data afterwards, detect the synchronization character of the beginning of each frame of expression by variable length decoding part 25, then, press macro block unit and restore predictive mode 3, motion vector 4 and quantification orthogonal transform coefficient 17 (S22).
When predictive mode 3 expression intra prediction modes (when S23 is affirmative determination), motion vector 4 is to dynamic compensating unit 5 transmission, the action of dynamic compensating unit 5 and code device is the same, and the image section with motion vector 4 activities from frame memory 26 (with frame memory 6 the same uses) takes out (S24) as motion compensated prediction image 8.On the other hand, during predictive mode 3 expression intra prediction modes (when S23 is negative evaluation), in spatial prediction portion 9 with reference near finish encoded image signal 7 generation forecast values 10 (S25).
Utilizing re-quantization portion 16 to carry out re-quantization to the predicted residual signal order concurrently with S23~S25 handles (S26) and utilizes inverse orthogonal transformation portion 19 to carry out inverse orthogonal transformation processing (S27), the decoding value of generation forecast residual signals.
On the other hand, according to predictive mode 3,, and, decipher image 20 (S28) thereby become with its decoding value addition with the predicted residual signal of above-mentioned generation by the final predicted value 11 of switch 12 decision.In addition, obtain final decoding reproduced picture 22 (S29), export (S30) to display device constantly in the demonstration of appointment by loop filter 21, thus reproduced picture.Decoding reproduced picture 22 stores frame memory 26 into for the motion compensation that is used for later frame simultaneously.Here, view data 20,22 become with code device in the identical value of data of duplicate numbers.
Below, describe present embodiment in detail, in the present embodiment, in above-mentioned code device and code translator illustrated in figures 1 and 2, the action that is included in definition, motion detection portion 2 and the dynamic compensating unit 5 of predictive mode 3 (particularly inter-frame forecast mode) in the compressive flow 24 and motion vector 4 has feature.Therefore, below, being defined as the center and describing for action and predictive mode 3, the motion vector 4 of code device with motion detection portion 2, is that the center describes for code translator with the order of picture signal being deciphered according to predictive mode 3 and motion vector 4.
1. the definition that fixing motion vector distributes figure
2 pairs of all predictive modes of being prepared of motion detection portion are estimated, and detect the pattern corresponding motion vector best with code efficiency.Formerly have in the H.26L coded system of example, as inter-frame forecast mode, selected figure shown in Figure 4, but, in the present embodiment, for method by 2 motion vectors of 4 * 4 (sub-piece) unit maximum allocated, prepared the fixing partition graph shown in Fig. 6 A~Fig. 6 D as inter-frame forecast mode, adopt the structure of therefrom selecting the best figure of code efficiency.As fixing partitioning scheme, use the reason of the figure of Fig. 6 A~Fig. 6 D to be, when being divided into 2 parts in the macro block, these figures predict that as cutting apart rule its usage frequency is the highest.For example, the diagrammatic representation of carrying out vector assignment at middle body and peripheral part has the so very situation of local activity in inside of 16 pixels * 16 row, and still, such situation is rare in image.On the other hand, the figure that the typical activity that the partition graph shown in Fig. 6 A~Fig. 6 D is represented level, vertical, incline direction also interrelates outward with macro block estimates that usage frequency is very high.Wherein, in Fig. 6 B~Fig. 6 D, white region and gray area are represented the difference that motion vector distributes, and can carry out motion compensated prediction with dynamically being divided into these figures in the macro block.
In the pattern 0 of Fig. 6 A, with 1 motion-vector prediction macro block (16 * 16 pixel) zone.This is and the H.26L pattern 1 identical definition of the inter-frame forecast mode of coded system.In addition, the pattern 1,2 of Fig. 6 B is the zone of 16 * 8,8 * 16 pixels with macroblock partition, pattern 2,3 equivalences among this and Fig. 4 B, Fig. 4 C.On the other hand, the later pattern of mode 3 of Fig. 6 C, Fig. 6 D is the distinctive mode-definition of present embodiment, in mode 3~6 of Fig. 6 C, allows macro block is cut apart unevenly in level, vertical direction.Cutting apart of uneven level, vertical direction, special very high in the local value that engenders new pattern as the picture end.In addition, the mode 7 of Fig. 6 D allows later on macro block is cut apart at incline direction.Incline direction cut apart the discontinuity of seizure along the activity of the boundary member of object, value is very high aspect the new pattern that occurs according to the prediction of picture angle.More than, in all patterns except the pattern 0 of Fig. 6 A, motion vector only is 2, but, by preparing various partition graphs in advance, utilize few motion vector number can catch reliably with the pattern 1,2 of Fig. 6 B such evenly cut apart the activity in the macro block that can not catch.
The difference of the size of code during each predictive mode of 4 groups (the group G4 of the group G3 of the group G2 of the group G1 of Fig. 6 A, Fig. 6 B, Fig. 6 C, Fig. 6 D) expression transmission that the pattern 0~14 of Fig. 6 A~Fig. 6 D is divided.In the present embodiment, compare with the H.26L prediction mode information of Fig. 4 A~Fig. 4 G, has many 1 times pattern count, so, for its expense is encoded effectively, according to each pattern cut apart number, cut apart direction or the area in the zone cut apart than carrying out the long Variable Length Code that distributes of variable code.At first, from cutting apart several viewpoints, to the number of motion vector less, pattern 0 that expense is few distributes the shortest code in all mode.Usually, because the relation of picture signal and time correlation is very high, so, if compare, can think very low according to the necessity in the motion segmentation macro block with the situation of carrying out motion compensation (Motion Compensation:MC) by macro block unit.In addition, for the figure that is divided into 2 parts, in picture signal in nature, can think that the frequency of level, vertical segmentation compares the figure form height that tilts to cut apart.In addition, be that unit carries out the viewpoint consideration that MC reduces average prediction residual power easily from, two data segment area identical even with the piece size, set the code length of evenly cutting apart short.On the other hand, when particularly new pattern occurring in the picture end etc., can think the inhomogeneous purposes of cutting apart the pattern part that is applicable to that prediction has better existed in frame.Tilt to cut apart and catch the activity that level-vertical movable institute can not catch that only depends in the profile that is applicable to the movable body in image, still, can think that usually its generation frequency lacks than other patterns.
But variable-length coding method is not unique when the Variable Length Code of the predictive mode that carries out a certain macro block, can adopt the situation (context, context relation) of considering near the predictive mode it yet and carries out the long method of distributing of code.For example, when the predictive mode of a certain macro block is encoded, be pattern 0, when upper left macro block is pattern 9, can think that the motion segmentation alive of the figure of succession pattern 9 takes place present macro block easily at left macro block.That is, under these circumstances, considered that the code of shortening pattern 9 long waits a left side, goes up, near the predictive mode situation (context relation) of the macro block in upper right side etc. changes such practices such as code table as parameter.No matter the kind of image how,, can carry out assignment of code effectively by determining general context relation.Simultaneously, by finding such context relation, adopt arithmetic coding to replace fixing assignment of code, based on context relation is switched the wattmeter of arithmetic coding, can expect to cut down by coding the effect of redundancy.
The Variable Length Code table of predictive mode coding usefulness also can switch according to the content of activity of imagination.For example, in the frame of image that comprises movable very fierceness or scene change, can think that the such inclination in mode 3~14 of Fig. 6 C, Fig. 6 D is cut apart or the inhomogeneous irregular figure of cutting apart etc. will take place in a large number, on the other hand, in rest image, can think that these predictive modes not too take place.That is, can think in the unit of frame or sequence that sizable variation takes place the ratio of effective predictive mode is very natural thing.In order to carry out such difference, the flag information multichannel that can consider the Variable Length Code table of predictive mode coding usefulness that will use in the frame or be used for the context relation definition turns to the part of heading message of this frame for example or frame group's beginning.In addition, consider that the distribution of predictive mode value is relevant with compression rates, can switch the Variable Length Code table of the predictive mode value of a plurality of preparations according to the quantification parameter initial value of frame unit.
Like this, by carry out the setting of the context relation of predictive mode according to the local property of picture signal, the predictive mode that can occur can be limited to a certain degree, so, even prepared, by the prediction mode information coding can be suppressed few with expense than the more predictive mode shown in Fig. 6 C, Fig. 6 D.For example, except the variation of Fig. 4 A~Fig. 4 G, also can use the predictive mode of Fig. 6 C, Fig. 6 D to change.But, be in the Variable Length Code of prerequisite with the context relation, because packed data error of transmission of near zone etc. and when correctly not deciphering, prediction mode information that just can not independently and correctly will be present be deciphered.For fear of such problem, can in for example MPEG-2 image encoding mode etc., use the unit that in two field picture, is divided into like that thinner absolute coding image-region (section), and the context relation of not using leap to cut into slices on the border of section.In the method, on the border of section, though the reduction of some code efficiencies is arranged,, in it is coordinated, can improve mistake.
In addition, also can switch a plurality of prediction mode information according to encoding condition.In the past, according to have earlier seen in the described H.26L former standard code mode (MPEG-1 or MPEG-2 etc.) of example to the predictive coding that utilizes motion compensation in, 1 motion vector is distributed to 1 macro block, and the adjustment of size of code can be carried out with the precision as the result's of residual coding DCT coefficient.Yet, according in recent years coding (H.263 or MPEG-4), having the predictive mode that in 1 macro block, is divided into a plurality of, the adjustment of size of code is not only carried out in the DCT coefficient but also in the balance about the size of code of prediction.Thereafter H.26L so that prepared a plurality of predictive modes, the adjustment of size of code is very strong with the tendency of carrying out with the precision of the corresponding motion compensation of employed motion vector number.In H.26L, when two-forty, can utilize the pattern of a lot of motion vector numbers in the pattern of use Fig. 4 A~Fig. 4 G, adjust size of code by the sophistication of motion compensation.But, under low rate, in order to carry out the adjustment of size of code with the motion vector number, it is 1 or 2 pattern (pattern 1~3) that the pattern that can select in the pattern of Fig. 4 A~Fig. 4 G almost limits per 1 macro block, does not wish that the problem of utilizing the motion compensation sophistication to improve code efficiency can be used as 1 problem and considers.
The optimized viewpoint of error (Rate-Distortion) is considered that best motion compensation model should be different with given coding rate (size of code) from size of code.That is the motion compensation model that, uses under low rate should be different with the motion compensation model that uses under two-forty.Under high code rate, utilize multi-mode, most motion vector on the meaning that waveform is preserved with the motion compensation sophistication, on the other hand, under low code rate, the approximation method that need utilize limited motion vector number to carry out more high efficiency motion compensation.Approximation method according to such speed motion switch compensation model can not be undertaken by coding standard up to now.
Here, under low rate, shown in Fig. 6 A~Fig. 6 D, though consider in 1 macro block, to exist single movable or do not have a single motion compensation model that also has the border (the motion vector number of per 1 macro block is defined as 2) of the different object of 1 activity when movable.Motion vector as per 1 macro block is defined as 2 situation, not to be 2 parts with macroblock partition by simple level, vertical segmentation, but by expansion for having increased cutting apart of level, vertical, inclination and phase place, the situation on the border of existence activity in 1 macro block different object is disposed.On the other hand, under two-forty, the also motion compensation model of allowing a lot of motion vector bar numbers shown in admissible chart 4A~Fig. 4 G.That is, the structure of switching the change of such motion compensation model according to the encoding condition of image is useful.For example, in frame or frame group (for example the unit with the video object layer of the sequence of for example MPEG-2 image encoding mode or GOP, MPEG-4 image encoding mode, video object plane etc. is suitable) unit, the identifying information multichannel that will be used to carry out the dynamic change of employed motion compensation model (being the sets definition of predictive mode) turns to the part of the heading message of its beginning.
In addition, the change of such motion compensation model as the calibration structure, can improve and install the convenience that utilizes with coded data.For example, shown in Figure 32 A~Figure 32 C, defined the set of 3 predictive modes.In Figure 32 A, can use minimal predictive mode, in Figure 32 B, in addition, can use present embodiment prompting use the predictive mode of few motion vector.In Figure 32 C, so adopted increase utilize cut apart more, the set of the predictive mode of a plurality of motion vectors.
By defining such set, if when reducing the amount of calculation burden, use the predictive mode set of Figure 32 A for prepreerence situation, use when in high code rate, utilizing a plurality of motion vectors of multi-mode to pursue exquisiter motion compensation Figure 32 C the predictive mode set, when not requiring especially, use the predictive mode set of Figure 32 B, then can be according to encoding condition motion switch compensation model.In addition, in the structure of Figure 32 A~Figure 32 C, each predictive mode becomes the subclass of the more set of predictive mode number.For example, the subclass of the set of the predictive mode of Figure 32 B set becoming Figure 32 C.Therefore, the coded data of encoding according to each predictive mode set can be deciphered in based on the installation of more high-order set.
The change of such motion compensation model, the profile that also can be used as the image encoding mode defines, thereby by the utilizable motion compensation model of the different regulations of purposes.So-called profile, in the coded system of a plurality of encoding functions of regulation, be by the subclass by the needed encoding function of purposes different definition obtain to install, the structure of amount of calculation burden and the balance of code efficiency.The set of the predictive mode shown in Figure 32 A~Figure 32 C can be handled them as the predictive mode set that utilizes in each profile, in addition, also the subclass of each profile as more high-order profile can be adopted calibration profile structure.
At this moment, when the unit of the change of motion compensation and residual coding is handled independently, then encoder and decoder installing just very desirable.When situation H.26L, residual coding is undertaken by 4 * 4 pixel units, so the parted pattern of macro block also adopts 4 * 4 pixel unit pieces.Like this, the change of motion compensation model is handled as the change to the ownership of each motion vector of residual coding piece, so from the viewpoint of installation effectiveness, effect is very big.
2. the predictive coding of motion vector information
Usually, under low bitrate,, the motion vector information that detects must be encoded expeditiously in order to improve the expense ratio of motion vector information.Usually, in the unit of macro block, movable size and Orientation all is a balance, and in this sense, motion vector wherein similarly is worth with near piece usually.From the viewpoint, motion vector determines predicted value according near motion vector, and its prediction residual data are encoded.Below, the motion vector encoder method of present embodiment is described.
In coded system H.26L, have a plurality of two field pictures of having encoded with the past and store in the frame memory 6 as the reference image and specify wherein some structures of carrying out motion compensated prediction.This exists moving zone and because the activity in moving zone and in the image that background is sheltered, can expect effective prediction before static background particularly.In the present embodiment, in motion detection portion 2, this structure can be utilized, and when in macro block, having 2 motion vectors, reference frame can be changed respectively.
In the inter-frame forecast mode of image encoding mode H.26L, no matter in macro block, how to carry out dynamic partition, because the bar number of motion vector is many, must ask each motion vector according to same reference frame, but, in the present embodiment, do not use most motion vectors to cut down expense, by allowing to give each motion vector, can further improve the efficient of motion compensated prediction with reference to the frame handover information.At this moment, reference frame handover information majority can adopt in the same macro block or contiguous macro block or the value approximate with the handover information of sub-piece.By utilizing this point prediction handover information or unified the coding, can cut down the needed size of code of reference frame handover information.
Certainly, itself must transmit the reference frame handover information as additional information, so, be when needing the simple image of such switching, can adopt the structure that in macro block, can select to make the identical rule of reference frame.At this moment, for example, in frame or frame group's unit, can consider that demultiplexing is used to discern the flag information of the reference frame change that whether allows in the macro block.Like this, code translator can not deciphered with reference to the frame handover information one by one to macro block when macro block internal reference frame does not change.
On the other hand, in the time of can making reference frame by the unit change of motion detection, between adjacent mutually motion detection unit reference frame not simultaneously, the dependency relation between these motion vectors reduces.That is, at this moment the precision of prediction of motion vector reduces.
For example, Fig. 7 A represent present macro block be pattern 1 and its left side, go up, situation that upper right macro block is respectively pattern 0,1,8.At this moment, shown in the arrow among the figure, be the continuity of dynamic partition according near the context relation of predictive mode, the motion vector of left fragment all with the dependency relation height of a left side, top, the motion vector of right fragment all with last, top-right dependency relation height.That is, in Fig. 7 A, now, a left side, go up, upper right during all with reference to identical reference frame, no matter which predicted value can be predicted accurately.But for example, when the macro block of 2 of Fang macro block motion vectors reference frame different with the motion vector of present macro block, left used same reference frame in the use, then the precision of prediction of the motion vector of Zuo Ce macro block was higher.
On the other hand, in Fig. 7 B, when the macro block of top uses the reference frame different with present macro block, just do not had with the continuity of the activity of the macro block of top, and, not only context relation can not determine to predict should use, which motion vector in the macro block of left, and because continuity that can not the expectation activity, so, can not expect high precision of prediction.
In order to solve such problem, in the present embodiment, add the relative importance value of prediction candidate according near the context relation of predictive mode, determine motion vector predictor adaptively.The information that near prediction mode information and the use of reference frame handover information own have been deciphered, so, the special additional information that does not need the motion vector predictor deciding step to use.Below, the flow process that concrete motion vector predictor decision is handled is described.
1) decision becomes near the motion vector of predicted value candidate to each fragment.Fig. 8 A~Fig. 8 C represents the example of pattern 0,1,7.In the pattern 0 of Fig. 8 A, motion vector has only 1, so, will draw the motion vector of piece position of oblique line as candidate.In the pattern 1 of Fig. 8 B, there are motion vector respectively in white region, gray area (having drawn the zone of oblique line in the macro block of Fig. 8 B), so, motion vector for white region, to draw the piece of oblique line as the predicted value candidate, for the motion vector of gray area, will draw the motion vector of piece position of horizontal line as the predicted value candidate.In Fig. 8 of pattern 1 B, the central portion above macro block, the piece of having drawn oblique line superposes with the piece of having drawn horizontal line.In the mode 7 of Fig. 8 C, the continuity of white region and top shortage activity, so, only drawing of left the motion vector of piece position of oblique line as the predicted value candidate.At gray area, will draw the motion vector of piece position of horizontal line as the predicted value candidate.Each pattern is predetermined these prediction candidate position, and common between code device and code translator.
2) the specific macro block that comprises the predicted value of using same reference frame in the macro block that comprises the predicted value candidate.As long as kept near the information of the macro block deciphered, whether used same reference frame near just can judging.
3) 2) use high macro block and the motion vector of relative importance value according to the prediction mode information particular prediction in the specific macro block, with this motion vector as predicted value.Prediction uses relative importance value to consider as shown in Figure 9 and near the continuity of activity, and each predictive mode predetermined fixed ground is defined, and is common in code device and code translator.In Fig. 9, be that example is when describing for example with pattern 5, when the predictive mode of macro block that becomes the object of present motion-vector prediction is pattern 5, in the macro block above it, defining mode value movable successional relative importance value height in 0,6,2,8,9,12,13 order.In addition, the mode value of the macro block of left is defined in relative importance value height in 5,0,2,8,12 the order.For top-right macro block, equally also define.For example, when identical with the left priority up, any a certain side can be decided to be predicted value.Promptly, in this example, when predictive mode itself is the prediction of pattern 5 motion vector that carries out gray area, the motion vector of the prediction candidate position of top, left uses same reference frame, the reference frame that top-right use is different, and, pattern 13 appears up, when pattern 5 appears in left, just with the motion vector of left as predicted value, be in top priority.
4) as the white region of the mode 7 of Fig. 8 C, when the prediction candidate position only is 1, if same reference frame, just directly it is used as predicted value, if different reference frame just uses a plurality of predicted value candidates of assigned address to determine predicted values (for example, using inter prediction).
5) although a plurality of predicted value candidate position are arranged, no matter during near the continuity that macro block all can not the expectation activity which (being when not being included in the predictive mode of priority of Fig. 9), just use all predicted value candidates to carry out the decision of predicted value (for example, inter prediction).
6) all motion vectors of predicted value candidate use all predicted value candidates to carry out the decision of predicted value (for example, inter prediction) when using the reference frame different with the motion vector that will predict.The predicted value that the coding of motion vector uses the result according to above processing to obtain is carried out Variable Length Code with the prediction residual data.
In addition, the predicted value of motion vector also can be supplied with as Figure 10 A~Figure 10 D regularly.In these figure, white arrow represents that black arrow is represented the direction to the prediction of the motion vector of black region to the direction of the prediction of the motion vector of white region.When being example with the pattern 4 of Figure 10 C, in the prediction of the motion vector of the white region of pattern 4, the top from white region begins to be positioned at motion vector with the sub-piece of adjacent position, the 3rd sub-piece left side as predicted value.In this rule, do not take into full account the continuity with near activity, still, owing to distributed predicted position regularly, so, can obtain the device of handling about motion-vector prediction easily.In Figure 10 A~Figure 10 D, when using different reference frame as the employed predetermined piece of predicted value, its continuity will not exist, and preferably switch to methods such as inter prediction.
3. compressive flow grammer
To be shown in Figure 11 B from the existing grammer comparison H.26L of grammer with Figure 11 A of the macro block data of the compressive flow 24 of the code device of present embodiment output.
Travel information shown in these figure (RUN), be the data of number encoder that will be judged to be the macro block of above-mentioned dancing mode, from n macro block to n+5 macro block be jump, when a n+6 macro block does not jump, the travel information=coded data of 5 central authorities is just given the beginning of n+6 macro block data.The predictive mode 3 of macro block type information (MB_Type) presentation graphs 1.
As inter-frame forecast mode, discern 7 kinds shown in Fig. 4 A~Fig. 4 G herein.The intra prediction mode information (Intra_pred_mode) of Figure 11 A is the information of using in order to discern a plurality of space predicting methods in intra prediction mode, can save (arrow P 1 of Figure 11 A) when having selected intra prediction mode.Reference frame information (Ref-frame) is the data of the number of the reference frame of designated movement vector indication when having selected intra prediction mode, can save (arrow P 2) when only general's 1 frame before this uses as reference frame.Motion vector difference value (MVD) is the difference value with the predicted value of motion vector, by the order demultiplexing respectively of level, vertical composition.Here, shown in Fig. 4 G, in H.26L, motion vector is 16 to the maximum, so, 16 of the maximum demultiplexings of the data of motion vector difference value (that is, arrow Q1 represents maximum 16 times circulation).The number of motion vector difference value is determined by macro block type information.When having selected intra prediction mode, can save (arrow P 3).After above data, the orthogonal transformation coefficient data of demultiplexing predicted residual signal (Texture Coding Syntax).
On the other hand, in the grammer of the present embodiment shown in Figure 11 B, the 14 kinds of inter-frame forecast modes (intra prediction mode is considered as with H.26L identical) shown in identification Fig. 6 A~Fig. 6 D in macro block type information.In addition,, make it corresponding, so reference frame information and motion vector difference value are always as 1 group and demultiplexing with 1 reference frame information for 1 motion vector.The number of the group of reference frame information and motion vector difference value is determined by macro block type information.Maximum number is 2 (that is, the arrow Q2 of Figure 11 B represents maximum 2 times circulation).The such intra prediction mode information of omitting shown in arrow P 4 just when having selected inter-frame forecast mode, in addition, such reference frame information and the motion vector difference value of omitting shown in arrow P 5 just when having selected intra prediction mode.
At this moment, reference frame information adopts the approximate value of reference frame information of the macro block or the sub-piece of or vicinity interior with same macro block mostly, so, can utilize it to encode.For example, distribute 1 code, constitute code, can reduce the size of code of most combinations that take place according to the probability of happening of composite figure for the combination of the value of 2 reference frame information in the same macro block.
Figure 33 A represents the code structure of reference frame information H.26L, and Figure 33 B represents the example to the combination assigned code of reference frame information.For example, no matter which piece of reference frame information in the same macro block all is 0 o'clock, then in the code of Figure 33 A, 2 of 2 reference frame informational needs in contrast, in the code of Figure 33 B, just can be encoded with 1, thereby can reduce size of code.At this moment, different about the code of reference frame information with the grammer of Figure 11 B, shown in Figure 33 C, with the motion vector difference value not in groups, only with 1 of each macro block demultiplexing.The arrow P 26 of Figure 33 C, P27 represent to omit this processing and jump, and arrow Q11 represents maximum 2 times circulation.
In addition, also can predict switch code according to the contiguous macro block or the reference frame information of sub-piece.For example, the same with the prediction of above-mentioned motion vector, the position of distribution forecast like that regularly that can be shown in Figure 10 A~Figure 10 D, reference frame information that will be corresponding with predicted position is as predicted value.Owing to also exist to resemble that pattern 0 is not distributed the situation of fixing predicted position or owing to be intra prediction mode etc. and the predicted position situation not corresponding with reference frame information, so, for these situations with regard to the fixed predicted value of use.The probability of happening of the value of the reference frame information when obtaining this predicted value constitutes code, and according to the predicted value switch code.
For example, the code of Figure 34 A adopted before changing or predicted value is 0 o'clock a code, as an example, predicted value is 4 o'clock a code, the code that uses Figure 34 B to constitute like that.Reference frame information adopts the value identical with predicted value mostly, predicted value is 4 o'clock, if reference frame information becomes 4 probability and is higher than reference frame information and becomes 2 probability, by using the code of Figure 34 B, reduce the size of code of reference frame information in the time of can be than the code that still uses Figure 34 A.At this moment, reference frame information is identical with Figure 11 B, with the demultiplexing in groups always of motion vector difference value.
Though it is not shown, but, in orthogonal transformation coefficient data, there is expression whether to exist the meaningful coefficient that should carry out coded data to have identifying information (abbreviation of CBP:Coded Block Pattern) by 8 * 8 unit in the macro block, by using this data, in having 8 * 8 of meaningful coefficient, all message transmission can be do not saved, thereby the code efficiency of orthogonal transformation coefficient data can be improved.Figure 35 A~Figure 35 C represents this example.Macro block cutting apart like that and encode shown in Figure 35 A, and in each sub-piece (4 * 4), with not having meaningful coefficient in the blank piece of representing.At this moment, shown in Figure 35 B, will have or not the information of meaningful coefficient to encode and transmit by 8 * 8 units, have or not each 4 * 4 the orthogonal transformation coefficient data that is included in 8 * 8 by indication in advance, each message transmission of 4 * 4 can be saved, thereby size of code can be cut down.
When using the predictive mode of present embodiment, under the situation of the figure of having selected to use 2 motion vectors, cut apart by this and to have improved forecasting efficiency, so, can envision the coefficient data that should encode in the predicted residual signal that is arranged in the partitioning boundary place seldom.Particularly in later inhomogeneous of mode 3 cut apart, tilts to cut apart, there is partitioning boundary, can thinks not exist the probability of the coefficient data of needs coding to increase for this part 8 * 8 inside.Therefore, with this situation accordingly, can change meaningful coefficient according to predictive mode and have the Variable Length Code table of identifying information or the probability tables of arithmetic coding.Like this, just, can improve the code efficiency that there is identifying information in meaningful coefficient according to predictive mode.
In addition, when using the predictive mode of present embodiment, part and the variation big part of macroblock partition for not having to change can think that having or not meaningful coefficient is certain in each cut zone.At this moment, as using H.26L, each 8 * 8 be not meant be shown with no meaningful coefficient, but, there is the needed size of code of identifying information thereby can reduce meaningful coefficient as indication has or not and constitutes the meaning coefficient the meaningful coefficient and have identifying information and encode to each cut zone.Figure 35 C represents this example.Transmitting when having or not the information of meaningful coefficient like that shown in Figure 35 B by 8 * 8 block units, its figure figurate number increases, but, shown in Figure 35 C like that transmission has or not the information of meaningful coefficient to cut zone the time, can reduce the figure figurate number, have the needed size of code of identifying information thereby can cut down meaningful coefficient.
Below, the decoding processing sequence of code translator is described.Figure 12 represents the handling process of code translator.
At first, in variable length decoding part 25, detect the synchronization character (S101) of the beginning of expression picture (being the coded data of each frame of image), with figure head decodings (S102).In the figure head, comprise this picture of identification and be the value of the picture/mb-type information of the picture (hereinafter referred to as the P picture) that the picture (by the picture of infra-frame prediction with all macroblock codings, hereinafter referred to as the I picture) of in-line coding also is to use inter prediction and the quantification parameter of orthogonal transformation coefficient data etc.Then, enter the decoding of each macro block data that the grammer by Figure 11 B constitutes.In the layer of macro block, earlier with travel information decoding (S103).Skipped macroblocks with several equivalent of the trip information takes place in the data of travel information described as defined above.For skipped macroblocks, on the reference frame of the appointment of storage, 16 * 16 pixel regions that will be positioned at same position directly obtain (S104, S105) as the decoding view data in frame memory 26.Obtain the decoding treatment of picture, undertaken by the reference frame number of motion vector 0 and appointment is supplied with dynamic compensating unit 5.When the final macro block of this picture of stroke information representation (S106), handle in this variable length decoding that finishes this picture constantly, in recursive filter 21, carry out the piece distortion and remove processing back (S118), transfer to the variable length decoding of next picture and handle.
About common coded macroblocks, enter step S107, carry out the decoding of macro block type information.In view of the above, determine predictive mode 3.(when S108 is sure judgement) carries out the decoding (S109) of intra prediction mode information during predictive mode 3 expression intra prediction modes.Correctly say, in H.26L, have by block (4 * 4) unit and carry out inside 4 * 4 patterns of infra-frame prediction and carry out 2 kinds of intra prediction modes of inside 16 * 16 patterns of infra-frame prediction by the unit of 16 * 16 pixels, and need inner 4 * 4 patterns of having only of intra prediction mode information, as the explanation of present embodiment because the purport difference, so omitted different detailed description to these processing.When having determined intra prediction mode information, in spatial prediction portion 9, generate intra prediction value (S110) according near view data.
On the other hand, when predictive mode 3 expression intra prediction modes (when S108 is the judgement negating), this pattern just becomes the value of expression a certain pattern shown in Figure 6.Therefore, determine constantly the reference frame information of which group and motion vector fully to be worth at this decipher (in pattern 0 is 1 group, is 2 groups) in pattern in addition.According to this information, decipher with reference to the group that frame information and motion vector fully are worth.But, the whether demultiplexing of reference frame information, with the information reconciliation of above-mentioned picture/mb-type, so, must judge in the value of picture/mb-type, whether there is reference frame information (S111).
In addition, though it is not shown, but, as previously mentioned, in utilizing same macro block or the value of the reference frame information of contiguous macro block or sub-piece when carrying out the coding of reference frame information, with its correspondingly according to the combined information of reference frame information or after utilizing predicted value to carry out the switching of code, carry out the decoding of reference frame information.
When reference frame information is arranged, after should being worth decoding, with motion vector difference value decoding (S112,113), when not having reference frame information, only with motion vector difference value decoding (S113).The predicted value that relative importance value decision motion vector is used in the predictive mode of representing according to the reference frame information that obtains like this, motion vector difference value, by macro block type information and near prediction mode information and the predicted value candidate position identical with Fig. 8 A~Fig. 8 C and code device shown in Figure 9 and prediction, each motion vector information that will be corresponding with all sub-pieces in the macro block restores.To dynamic compensating unit 5 transmission, dynamic compensating unit 5 generates motion compensation prediction value (S114) to each height piece according to the reference frame information of appointment and motion vector with this motion vector information 4.
When above processing finishes, carry out the decoding (S115) of orthogonal transformation coefficient data 17 of the predicted residual signal of this macro block by variable length decoding part 25, carry out the recovery (S116) of predicted residual signal then by re-quantization portion 18 and inverse orthogonal transformation portion 19.At last, by the predicted value addition that will obtain at S110 or S114, obtain the decoding image (S117) of this macro block.Then, the decoding of transferring to next macro block is handled.
According to above-described code device and code translator, can catch various activities in the macro block reliably by few expense, thereby can carry out image transmission, record, regeneration expeditiously.In addition, in the present embodiment, in with various shape dynamic partition macro blocks, improve in the forecasting efficiency, the zone that becomes the motion compensated prediction object constitutes by the combination of sub-piece, so, when as image encoding mode H.26L, carrying out the predicted residual signal coding of sub-block unit, owing to there is not the partitioning boundary of prediction in the inside of sub-piece, so, discontinuous waveform can not take place.Unnecessary high frequency waves coefficient can not take place in this when using DCT in the predicted residual signal coding, when using vector quantization, also can suppress the specific signals waveform, on with meaning, can carry out high efficiency coding by the performance of the general waveform of code book regulation.
Below, embodiment 2 is described.In the present embodiment, illustrate undertaken by macro block unit in the image encoded code device each macro block maximumly use 2 motion vectors, to being that each sub-pieces of 4 * 4 pixel units distributes above-mentioned 2 motion vectors to carry out the picture coding device and the code translator of motion compensated prediction with macroblock partition.The motion vector of embodiment 1 antithetical phrase piece distributes with fixedly figure supply, and in contrast, in the present embodiment, explanation can be specified code device and the code translator to the motion vector distribution of sub-piece arbitrarily.
In the present embodiment, except with code device illustrated in figures 1 and 2 that example is formerly arranged and code translator in be included in the action of definition, motion detection book 2 and motion compensation book 5 of predictive mode 3 (particularly inter-frame forecast mode) in the compressive flow 24, motion vector 4 different, can be with identical parts and action specification.Below, be defined as the center and describe for action and predictive mode 3, the motion vector 4 of code device with motion detection portion 2, for code translator being that the center describes with the order of picture signal decoding according to predictive mode 3 and motion vector 4.
At first, be the processing of center explanation code device and in proper order with the detection of the structure of the predictive mode 3 of present embodiment and its corresponding motion vector 4 to the structure of the compressive flow of code translator transmission.The predictive mode 3 of present embodiment is shown in Figure 13 A~Figure 13 C, and the fixing motion vector of not only above-mentioned Fig. 6 A~Fig. 6 D distributes but also the information expanded as specifying allocation situation arbitrarily defines.That is,, then can as Figure 13 A~Figure 13 C, the binary map information as 4 * 4 show if make the ID of 2 motion vectors be respectively 0,1.Here, the distribution of ID for example has the meaning as the transmission sequence in the compressive flow.
Below, the coding method of the predictive mode 3 that determining method (implementing in motion detection portion 2) as Figure 13 A~Figure 13 C motion vector 4 that works distributes and Figure 13 A~Figure 13 C show is described.As long as it is just passable to have estimated the determined a plurality of figures of distribution of fixing motion vector in embodiment 1, so, the detection of motion vector, implement just passable regularly in each zone, but, the detection of motion vector in the present embodiment must be considered the optimal cases that motion vector distributes and implements.
Figure 14 represents the order of the example that detects.At first, implementing with 16 * 16 pixels at S201 is the detection of the motion vector of unit.This can be the identical motion vector detection of situation with the pattern 0 of the pattern 1 of Fig. 4 A, Fig. 6 A.If the motion vector that obtains according to testing result is the 1st motion vector V1.Secondly, the prediction and evaluation value that the predicted value that use obtains according to this check result is asked each sub-piece of 4 * 4 (here, use abundant absolute value and SAD:Sum of AbsoluteDifference), the by way of exception sub-piece of the sub-piece that the prediction and evaluation value is bigger than designated value zone and extract (S202) out.Like this, can obtain oblique line portion represent the to make an exception motion vector in sub-piece zone and distribute Figure 40.This zone that sub-piece region representation can not be predicted fully with the 1st motion vector V1 of making an exception by particular case my husband piece zone, can be carried out motion vector according to forecasting efficiency and distribute.Then, the sub-piece that carries out only will making an exception detects the 2nd motion vector V2 (S203) as the detection of motion of objects vector.
In above processing procedure, prediction and evaluation value but also consider motion vector value or detect processing not only by the size of code of the motion vector assignment information (that is the predictive mode 3 of Fig. 1) of the sub-piece of exception zone decision.In addition, also can individually change reference frame respectively and detect as the detection of embodiment 1 the 1st and the 2nd motion vector.
Figure 15 represents the order of other test example.In this embodiment, be the detection that unit carries out motion vector at first with sub-piece, ask 16 vectors (S204).Then, all for macro block, 16 vectors are grouped into 2 vector V 1, V2, so that be thought of as optimal cases (S205) from the viewpoint of size of code and forecasting efficiency.Particularly carry out vector assignment, when macro block is encoded, can detect the motion vector that becomes best speed distortion balance by the size of code of considering predictive mode 3.According to above method, decision predictive mode 3 and motion vector 4 in motion detection portion 2.
Below, use the coding method of the information of Figure 16 A, B explanation predictive mode 3.The order of coding is shown in the flow chart of Figure 17.If Figure 16 is A, the macro block C of B is that the macro block of coded object, macro block A are that macro block, the macro block B that is positioned on the C is the macro block that is positioned at the left side of C.Each macroblock partition is 4 * 4 sub-block unit, with the different situations of representing motion vectors distribution (predictive mode 3 of Fig. 1) of white and grey.Coding is encoded the change point of the motion vector distribution of sub-piece according to the situation of the predictive mode of macro block A and B.At first, inherit the situation (predicting preferential context relation) (S206 of Figure 17) of a certain each predictive mode of A and B according to the predictive mode situation decision of macro block A and B.In addition, linage-counter L is reset to 0.In Figure 16 A, obviously can know the continuity height of the activity of macro block A, so as shown by arrows, the variation that the motion vector of vertical direction is distributed is as detected object.After, be referred to as the preferential context relation of vertical prediction.On the contrary, shown in Figure 16 B, when the continuity of the activity of judging macro block B was high, the variation that the motion vector of horizontal direction is distributed was as detected object.After, be referred to as the preferential context relation of horizontal forecast.In Figure 16 A, B, the change point when representing that with thick line macro block C predicts that by each preferential context relation is encoded.Below, describe with the example of Figure 16 A.
When handling row L, determine residue operation number r, operation starting point a earlier
0, end of run point a
1(S207).End of run point a
1Expression distributes the sub-piece position of the motion vector different with the motion vector that distributes by the most adjacent sub-piece by the macro block (situation of Figure 16 A is macro block A) of predicting preferential context relation appointment.
But, in fact in macro block A and C, or not to detect identical motion vector continuously usually, so, need to adopt will be for example to be considered as initial launch starting point a with the macro block of coded object by the motion vector allocation situation (0 or 1) of the adjacent sub-blocks of the macro block of the preferential context relation appointment of prediction
0Distribution identical and begin to handle etc. method.
As method for distinguishing, will predict that in addition preferential context relation is enclosed in useful method in the macro block C.That is, in 16 sub-pieces, the sub-piece in the upper left corner is keeping 0 or 1 value, and for 4 row of the top, the change point of the sub-piece that the left side is adjacent is taken as 1, and for leftmost 4 row, the change point of the sub-piece directly over always inciting somebody to action is taken as 1.By such decision, can determine the change point of remaining sub-piece according to the preferential context relation of prediction.
The be expert at beginning of L, r=4, a
0=0, in the example of Figure 16 A, a of L=0 as can be known
1Be 3.Like this, going the continuation of L distributes the sub-piece number of same movement vector by R=a
1-a
0Decision.Use is carried out Variable Length Code (S208) by the long CB of code (r) of residue operation number r decision with this value.The example of the long CB of code (r) is shown in Figure 18 A~Figure 18 C.When carrying out the coding of R, the situation occurred of remaining R is owing to the process in the way is restricted, so, by long, can improve code efficiency according to residue operation number r switch code.For example, when the L=0 of Figure 16 A, at first obtain r=4, R=3, so, use CB (4) to carry out the coding of R=3 (distributed code 1, code length is 3).Secondly, utilize r=r-R to upgrade residue operation number r, will move starting point a simultaneously
0Transfer to a
1(S209).Like this, then r=1, still, residue operation number be 1 just be illustrated in this motion vector of going L constantly distribute all definite, so, transmission code especially.Here it is does not define the reason of CB (1).Therefore, in r≤1 o'clock, the counting of L is finished, transfer to the processing (S210, S212) of next line.On the contrary, new change point (S211) is just detected in r>1 o'clock, carries out above-mentioned processing repeatedly.In the last capable L=3 of the macro block C of Figure 16 A, because the sub-piece that begins most becomes the end of run point suddenly, so, with the code coding of CB (4) " initial inversion 0 ", after, count r=3 as remainder stroke information and handle.Become for 4 the moment at line number L, finish encoding process.
More than, in Figure 16 A, when L=0, with 3 codings that carry out predictive mode 3, when L=1, with 4 codings that carry out predictive mode 3, L=2,3 o'clock, with 6 codings that carry out predictive mode 3, in addition, add the size of code (, the level of establishing all is 3 with vertical) of the preferential context relation of prediction here, just with 18 codings that carry out predictive mode 3.Under the situation of Figure 16 B, can encode with 4 L=0,1 o'clock, can encode with each 1 L=2,3 o'clock, add context relation, can encode with 13.Even binary message also can show with 16, so, becoming the situation of the performance more than 16, also can encode by methods such as escapes with 16 fixed length codes.In addition, shown in Figure 16 A like that, for as the high figure (situation shown in Fig. 6 B~Fig. 6 D) of partition graph anticipation occurrence frequency, when allocating in addition partition graph arbitrarily of fixing short code performance in advance, can use the code of above-mentioned coding.According to above-mentioned encoding process order, can be expeditiously the information of the predictive mode 3 of Fig. 1 of the allocation situation of motion vector arbitrarily of the sub-block unit of expression is encoded and transmit.
According to above-mentioned explanation, will compare from the grammer of the macro block data of the compressive flow 24 of the code device of present embodiment output and H.26L grammer (Figure 19 A), and be shown in Figure 19 B, C.Figure 19 B is the simplest grammer example of present embodiment, macro block type information only comprises the judgement information of infra-frame prediction or inter prediction, for infra-frame prediction, when 16 * 16 pixel unit motion compensation, distribute shortest code, for the information distribution time short code of the preferential context relation of prediction.Predict that it is the intra prediction mode that utilizes any partition graph that preferential context information is promptly represented.
The occurrence frequency of infra-frame prediction is lower than inter prediction usually, so distribution ratio is predicted the code that preferential context information is long.For example, if according to H.26L variable length code definition, for 16 * 16 pixel unit movement compensating modes with regard to allocation of codes " 1 " (1), for the preferential context information of prediction with regard to allocation of codes " 001 ", " 011 " (3).When selecting intra prediction mode,, intra prediction mode information is deciphered (shown in arrow P 9, having omitted intra prediction mode information during inter-frame forecast mode) with H.26L the same.Secondly, the motion vector assignment information demultiplexing of the sub-block unit that will encode with above-mentioned coding method.It also can be 16 a fixed length code.When intra prediction mode, this information has been omitted corresponding processing shown in arrow P 10.Later processing, described the same with embodiment 1.
Figure 19 C represents other syntactic structure, utilize code " 001 ", " 011 " (3) conduct that the pattern information of the inter prediction of any partition graph is still used with the identifying information that fixed length (16) is transmitted with variable length, replace with the macro block type information performance and predict preferential context relation.When expression is variable length, in the preferential context relation of prediction (pred_ctx) afterwards, demultiplexing motion vector assignment information (pattern_info).On the other hand, when expression is fixed length, with regard to fixed length code demultiplexing with 16.In addition, in Figure 19 B, in macro block type information, also may be defined as the information and the expression that can show the high partition graph of the specific occurrence frequency of the expression shown in Fig. 6 B~Fig. 6 D is the escape code of figure in addition.At this moment, in the motion vector assignment information, comprise the value of predicting preferential context relation.Like this, the information of appending that the partition graph that occurrence frequency is high does not just need motion vector to divide adapted, thus can encode with higher efficient.In Figure 19 A~Figure 19 C, arrow P 6~P15 represents to have omitted corresponding information, and arrow Q3 represents maximum 16 times circulation, and arrow Q4, Q5 represent maximum 2 times circulation.
The motion flow of code translator is shown in Figure 20.The code translator of present embodiment is compared with the code translator of embodiment 1, except the decoding of prediction mode information 3 order difference, move identical, so, the part of the decoding processing of prediction mode information 3 only is described.
In Figure 20, S101~S118 is the identical treatment step of treatment step with the code translator of embodiment shown in Figure 12 1.When the S108 of Figure 20 represents to be inter-frame forecast mode, just the motion vector assignment information of Figure 19 B is deciphered at S215.At this moment, predict that preferential context information has been included in the macro block type information,,, use CB (r) to decipher processing here according to the preferential context information of the prediction of in the processing procedure of S214, deciphering.Because this can handle by general variable length decoding and implement, so, detailed explanation omitted.In the stage that the motion vector of having determined whole 16 sub-pieces distributes, finish the processing of S215, transfer to the decoding of reference frame information and motion vector difference value and handle.Later processing, described the same with embodiment 1.
In addition, in Figure 20, we can say that Benq handles in the decoding of the grammer of Figure 19 C.At this moment, at step S214, the motion vector assignment information replaces the preferential context information of prediction and represents variable length code or fixed length code.According to this information,, carry out the decoding of motion vector assignment information according to the grammer of Figure 19 C at step S215.When being the variable length code, predict that earlier the decoding of preferential context information is handled.And then, (for example comprise the high partition graph of several occurrence frequencies in macro block type information, during the pattern 1~14 of Fig. 6 B~Fig. 6 D) identifying information, step at S214, just determined partition graph, so, when detecting the escape code of any partition graph of expression, all skip the step of S215.
According to above-mentioned code device and code translator, can cut apart arbitrarily and catch various activities in the macro block by the unit of sub-piece, carry out image transmission, record, regeneration expeditiously by few expense beginning.
Below, embodiment 3 is described.Other coding method of predictive mode 3 of Fig. 1 of the partition graph arbitrarily of expression embodiment 2 is described in the present embodiment.Figure 21 represents its handling process, is the handling process of the concrete example that uses in Figure 16 A, and Figure 22 is the key diagram of this processing.At first, at the S301 of Figure 21, the preferential context relation of prediction that decision is identical with embodiment 2.Here, when having selected the preferential context relation of vertical prediction, will be from the sub-piece bottom of macro block A capablely see motion vector is distributed in where to change and detect (S302) as change point.For example, in macro block A, the C of Figure 22 A, the thick line part detects as change point.Like this, just generated to change a little be 1, binary map that the no change point is Figure 22 B of 0.Use the scan table (for example scan table of Figure 22 C) of appointment to scan this figure, generate zero stroke (S303).
For example, owing to adopted the preferential context relation of vertical prediction here, so, be used for scan table that the vertical direction to Figure 22 C scans binary map with Figure 22 B.As a result, just generated 3,1,3, the group of the such zero stroke of 5}.By this information being carried out Variable Length Code (S304~S307), carry out the coding of predictive mode 3.When change point disappears in the way of scanning (when S305 is sure judgement), later zero stroke information is just only replaced (S306) by 1 terminal code (EOB).
On the contrary, when before the end of scan, having change point, the summation of utilizing the number of zero stroke and change point is the code that 16 situation is determined last zero stroke, so, do not utilize terminal code encode (when S307 is sure judgement, not carrying out the processing of S306).
For scan table, can use the saw-tooth sweep of Figure 23 A for example, the horizontal sweep of Figure 23 B, the Hilbert scanning of Figure 23 C, in addition, also can be to show arbitrarily.Can also can use a certain specific general scanning rule regularly according to the preferential context relation change scanning of prediction (when having selected the preferential context relation of horizontal forecast, just adopting horizontal sweep etc.).In addition, also the special mark that is used for the switched scan table can be encoded as the part of the information of predictive mode 3.At this moment, in code device, can from a plurality of scan methods, use the highest scan table of code efficiency selectively.
Utilize the grammer of the predictive mode 3 that above-mentioned coding method obtains, can realize the motion vector assignment information shown in Figure 19 B, the C by the coding method that is replaced into embodiment 2.In addition, the structure of code translator is also deciphered the code of the coding method described in the present embodiment with the step of the S215 of Figure 20 and is determined that the processing of predictive mode 3 is corresponding.
In the present embodiment, the example that uses the preferential context information of prediction is described, but, with the preferential context relation of prediction irrespectively by a certain fixing rule detection change point, the structure that also can adopt the information that will not predict preferential context relation to encode.At this moment, scan table can the time fixing, also sign can be set, in order to the high table of switching efficiency.
According to above-mentioned code device and code translator, can cut apart arbitrarily and catch various activities in the macro block by the unit of sub-piece, thereby can carry out image transmission, record, regeneration expeditiously by few expense.
Below, embodiment 4 is described.In the present embodiment, illustrate that the change of antithetical phrase block unit not only motion vector distributes but also the code device and the code translator of the predictive mode 3 of pie graph 1 in order to distinguish inside/inter prediction.Like this, can select inside, centre (in a plurality of motion vectors some) that each sub-piece is encoded, so, can carry out the efficient coding that the character with actual image adapts to than the mode of carrying out mode switch by macro block unit.
At first, the definition of the predictive mode 3 of Fig. 1 of the example explanation present embodiment of use Figure 24 A, B.In coded system H.26L, as infra-frame prediction, prepared two kinds of predictive modes, promptly switched with a plurality of Forecasting Methodologies of intra prediction mode information Recognition and carry out inside 4 * 4 patterns (INTRA_4 * 4) of spatial prediction and switch inside 16 * 16 patterns (INTRA_16 * 16) that a plurality of Forecasting Methodologies are carried out spatial prediction by the unit of macro block by 4 * 4 sub-block unit.
To intra prediction mode information, defined 6 kinds of (1~6) space predicting methods altogether, can specify these methods by the unit of sub-piece.Therefore, can expand this method, for the predictive mode of 4 * 4 sub-pieces, coordinated movement of various economic factors vector assignment information.That is, in the time of can using 2 motion vectors, just the ID with its Forecasting Methodology is defined as 7 and 8.Example according to this definition antithetical phrase piece distribution forecast method 1~8 is shown in Figure 24 A, B.In Figure 24 A, the difference of inside/centre is being carried out in the angular direction, in inter-frame forecast mode, use 2 motion vectors.
In Figure 24 B, macroblock partition is two parts up and down, is that top is that infra-frame prediction, bottom are the example of inter prediction, and in this embodiment, inter prediction only uses 1 motion vector.Below, the prediction assignment information that this is many-valued is called many-valued prediction identification map information.
For the identification of inner 16 * 16 patterns and 16 * 16 pixel unit movement compensating modes (below, 16 * 16 patterns in the middle of being called), do not need the prediction assignment information of 4 * 4 sub-block units, so, only utilize macro block type information just can realize discerning.That is, macro block type information is divided into inner 16 * 16 patterns, middle 16 * 16 patterns and sub-block unit and predicts 3 kinds of distribution arbitrarily.Secondly, be sub-block unit when predicting the pattern of distribution arbitrarily, determining the Forecasting Methodology of all sub-pieces by will above-mentioned many-valued prediction identification map information encoding.
In this embodiment, in the motion detection portion 2 of code device, can predict that arbitrarily all possibilities of distribution estimate, select the highest pattern of code efficiency inner 16 * 16 patterns, middle 16 * 16 patterns and sub-block unit.Certainly, also can constitute and be fixed as a certain mode combinations and carry out apparatus for encoding.The detection of the best of sub-block unit prediction can be considered various methods, but, can extension example such as the method shown in Figure 14 of embodiment 2, the sub-piece that has the prediction and evaluation value bigger than the prediction and evaluation value of infra-frame prediction in the sub-piece zone of making an exception can utilize method such as infra-frame prediction to determine the identification map information.
Below, the coding method of the many-valued prediction identification map information shown in key diagram 24A, the B.This identification map information can be encoded with the coding method substantially the same with the motion vector assignment information shown in embodiment 2 and the embodiment 3.Different is, the information of the coded object of embodiment 2 and embodiment 3 is binary map, and the coded object of present embodiment is a grey scale mapping.
(1) coding method 1
For the method with embodiment 2 realizes, each pattern information is launched on bit plane.For example, the inner or middle plane of expression, expression be can be divided into and the plane (6 planes altogether) of using each intra-frame prediction method and the distribution plane of motion vector etc. had or not.Can encode with the method shown in the embodiment 2 in these planes.For example, when the many-valued prediction identification map information of Figure 24 A is decomposed on bit plane by above-mentioned rule, just become the situation shown in Figure 25 A~Figure 25 C.Here, the preferential context relation of prediction has all been used on the plane that the motion vector of the plane of inside/middle classification of Figure 25 A and Figure 25 C distributes, and still, for the plane of intra-frame prediction method, only the situation that has or not Forecasting Methodology is constituted bit plane.
The binary map that amounts to 8 planes of Figure 25 A~Figure 25 C is encoded by named order by method, can show the information of the predictive mode 3 of present embodiment with embodiment 2.In addition, beginning, identification by prioritised transmission inside/centre, the plane of distributing about the motion vector of 6 planes of Figure 25 B of intra-frame prediction method and Figure 25 C can be shown in Figure 26 B, C such difference only will with inner, middle corresponding sub-piece as bit-plane data (that is, the grey color part of Figure 26 B, C is not as coded object).Like this, just, can reduce the information encoded amount of answering effectively.In addition, also can use the coded system of embodiment 3 that above plane information is encoded.
(2) coding method 2
In addition, carry out Methods for Coding, can consider that no change point with Forecasting Methodology is taken as 0 and the method that stroke is represented is carried out in the no change interval that same Forecasting Methodology continues as the distortion of embodiment 3.When utilizing this method, the situation of Figure 24 A is carried out conversion (with the preferential context relation change detected point of horizontal forecast) as shown in Figure 27.Pressing the group of the Forecasting Methodology ID of zero stroke and change point encodes this information.In this example, { (5,7), (0,1), (2,7), (3,8), (1,7) } the such group of can using the horizontal sweep of Figure 23 B to be launched into.These set of dispense variable length codes are encoded.Scan table has been prepared sign, can switch to the high table of efficient and use.
In addition, at this moment, be considered as not being the variation of Forecasting Methodology as the variation of 7 and 8 value of inter prediction, thereby as the no change point, and encoded in the plane of motion vector assignment information in addition.
(3) coding method 3
Can expand to sign, with predicted portions between the form transmission frame on motion vector distribution plane to the intra prediction mode information adding inter prediction that routine H.26L coded system is arranged earlier.
(4) coding method 4
For example Fig. 6 A~Fig. 6 D distributes inside/middle classification figure and motion vector to distribute figure like that regularly in advance.Situation and motion vector allocation situation are cut apart in all inside/centres, can be as fixing pattern I D by encoder and decoder identification, so, according to the order of occurrence frequency the ID of these figures is carried out Variable Length Code.In addition, distribute figure (being equivalent to embodiment 1) for inside/middle partition graph and motion vector, consider to think with nigh macro block in the dependency relation height of the figure selected, so, partition graph for inside/centre, also the same with embodiment 1, based on context the situation defining context relation according near macro block concerns the probability tables that this Variable Length Code table of switching or arithmetic coding are used.
Grammer by the predictive mode 3 of above method coding is shown in Figure 28~Figure 31.Figure 28~Figure 31 is corresponding with above-mentioned coding method respectively.In coding method 1, as shown in figure 28, the motion vector assignment information is encoded on each bit plane.For the preferential context relation of prediction, the same with embodiment 2, can be enclosed in the same macro block and define, also can not use this definition.
In coding method 2, as shown in figure 29, after the preferential context relation of prediction, demultiplexing is used to discern the scan table flag information of scan table.During according to the preferential context relation of prediction decision scan table or when having determined scan table regularly, just do not need this sign.Then, transmission obtains the group number of the Forecasting Methodology ID of zero stroke and change point according to the result of scanning.
In coding method 3, as shown in figure 30, utilization is expanded to the extended edition information of intra prediction mode information adding inter prediction sign, can specify the 7 kinds of methods that amount to that 6 kinds of intra-frame prediction methods added inter-frame prediction method, thereby become the structure of the motion vector assignment information of the motion vector distribution when this extended edition message transmission is used to specify inter prediction.
In coding method 4, as shown in figure 31,, comprise the classification figure of inside/centre and comprise the fixing mode assignments figure that motion vector distributes figure as macro block type information, expand to the very data of many predictive mode values of identification uniquely.Secondly, only to carrying out identifying information by the other transmission frame inner estimation mode of 4 * 4 sub-pieces in the zone of the infra-frame prediction of macro block type information appointment.As grammer, can directly use intra prediction mode identifying information (intra prediction mode information) at the sub-block unit that for example H.26L uses in inside 4 * 4 patterns of coded system.
In Figure 28~Figure 31, arrow P 16~P25 represents to have omitted this information, and arrow Q7 represents maximum 16 times circulation, and arrow Q6, Q8~Q10 represent maximum 2 times circulation.
In code translator, receive the grammer of above various structures according to the decoding motion flow of Figure 12, Figure 20 etc., according to determining that with the order of the reversed in order of coding method regulation the predictive mode in the macro block distributes, and deciphers picture signal.
According to above-mentioned code device and code translator, can cut apart arbitrarily and catch various changing conditions in the macro block by the unit of sub-piece, thereby can carry out image transmission, record, regeneration expeditiously by few expense.
In the present embodiment, in predictive mode definition, the bar number of motion vector can be taken as 3~4, append Forecasting Methodology ID and distribute to them respectively.
Below, embodiment 5 is described.In the present embodiment, illustrate in the code device of embodiment 1 and code translator, to have installed and handle and the example of other parts of method of motion vector prediction about motion vector detection.
Utilization is when the inter-frame forecast mode of the A of the Fig. 6 described in the embodiment 1~Fig. 6 D, the detection method of the motion vector of each pattern can be with any method of the grammar explanation of appointment as adopting described in the action of the code translator of embodiment 1, but, in the present embodiment, the burden that not too increases operand is described and improves the motion vector detecting method of forecasting efficiency and the code device and the code translator of its corresponding motion-vector prediction rule of employing.In the following description, motion vector detection is handled and is carried out in the motion detection portion 2 of Fig. 1, motion-vector prediction is handled and is undertaken by the variable-length encoding part 23 (code device side) of Fig. 1 and the variable length decoding part 25 (code translator side) of Fig. 2, in the explanation afterwards, utilize the flow chart of these processing to carry out action specification.
Inter-frame forecast mode at the embodiment 1 shown in Fig. 6 A~Fig. 6 D is concentrated, and must detect 1 or 2 motion vectors to each macro block.In low rate coding, the only size of code minimizing that in the performance of orthogonal transform coefficient, can use of the part that increases of the bar number of motion vector, so, consider that preferably the cost of the size of code of motion vector carries out the detection of motion vector with the compromise selection of the precision of prediction that utilizes the motion vector that detects.In the H.26L image encoding mode as the basis of embodiment 1, the time that is located at is gone up the reference frame RF before the i frame
iThe motion vector that detects in (i 〉=1) is MV
i(n, j) (n: macro block position, j: the segment number sign indicating number), use R
pExpression motion vector MV
i(n is j) with reference frame identifying information RF
iThe needed position amount of coding, use the prediction residual power D that at this moment obtains
pDecision makes the cost C by following formula (1) decision with constant λ
pMotion vector MV for minimum
i(n is j) with reference frame RF
i(i 〉=1).The value of setting constant λ reduces and increases along with the target code bit rate, so, in low rate coding, position amount R
pRestriction be strict.
C
p=D
p+λR
p …(1)
As described in the embodiment 1, after the predicted value of decision motion vector, the predicted difference score value is encoded, so the motion vector of detection is as far as possible near the position amount R of predicted value
pReduce.Therefore, in image encoding mode H.26L, recommend the central point of scope that motion vector is explored to be set at motion vector predictor, the position that makes motion vector explore window changes with each section adaptively.Like this, can cut down the size of code of motion vector at least, make the cost C of following formula
pBecome hour, just relaxed a position amount R
pRestriction, utilize prediction residual power D
pMinimize, can under advantageous conditions, carry out motion vector detection.The continuity of the activity of consideration and near macro block also can expect to make adaptively the exploration window position to meet predicted value and prediction residual power D
pReduction interrelate.On the other hand, in the self adaptation of exploring window, there is the big problem of burden of the computing of prediction residual power.
For this problem is described, Figure 36 A will carry out motion vector and explore the situation of the self-adaptive processing of scope and carry out schematization and represent, Figure 36 B will not carry out motion vector and explore the situation of the self-adaptive processing of scope and carry out schematization and represent.Shown in Figure 36 B, the exploration scope is certain, and with section when irrelevant, that utilizes section is shaped as hierarchical structure, the calculation times that can cut down prediction residual power.
For example, in Fig. 6 D, upper segment (white segment) for pattern 13, by exploration window all motion vector candidates are carried out the prediction residual power calculation with unique decision, in pattern 9, needn't carry out the prediction residual power calculation once more to 3 sub-pieces in the upper segment of pattern 13, can only carry out the prediction residual power calculation remaining 3 sub-piece.In contrast, in Figure 36 A, during as prerequisite, the upper segment of pattern 13 is different with the exploration position in the same place of pattern 9 with the motion-vector prediction rule of using Figure 10 D, so, can not utilize this prediction residual power calculation result once more.
Therefore, in the present embodiment, illustrate that suppressing operand explores the self-adaptive processing of each section of window and handle and the motion-vector prediction rule in order to the motion vector detection that improves precision of prediction.The decision regularly as Figure 37 A~Figure 37 D of the motion-vector prediction rule of present embodiment.In this rule, to compare with Figure 10 A~Figure 10 D, the variation of the predicted position of use is little, roughly only uses 4 kinds of motion-vector predictions.
The 1st kind is the prediction (1. MV predicts) of the upper segment of pattern 0 and pattern 9,10,13,14, at this moment, adopts central authorities' prediction of the motion vector of sub-piece (4 * 4) position of using 3 places that surrounded by oblique line.
The 2nd kind and the 3rd kind is the prediction of pattern 1~8,11,12, and shown in Figure 37 B~Figure 37 D, (2., 3. MV predicts according to the prediction of the motion vector of the sub-piece of the different fixed positions of adopting use left or up side of section.Begin from the left side be predicted as 2., from being predicted as 3. that upside begins).
The 4th kind is the prediction of the lower section of pattern 9,10,13,14, and at this moment, the sub-piece of the dotted line position that use is adjacent with the sub-piece in the upper right corner of macro block is predicted (4. MV predicts).
In addition, when the motion vector of predicting employed place uses the reference frame different with the motion vector that will predict now or when carrying out not defining motion vector after the in-line coding, the employed inter prediction of upper segment of application model 0 and pattern 9,10,13,14 still.In Figure 10 A~Figure 10 D, each pattern and each section have been used different prediction rule, still,, can concentrate to be maximum 4 kinds of predictions by using the prediction rule of present embodiment.Like this, when carrying out the exploration of motion vector,, also can reduce the calculation times of prediction residual power significantly even set and the adaptive exploration window of predicted value.
The flow chart that the motion vector detection of utilizing above motion-vector prediction rule to carry out in the motion detection portion 2 of code device is handled is shown in Figure 38.Below, use Figure 38 explanation to utilize the motion vector detection of above-mentioned motion-vector prediction rule to handle.
At first, use the MV prediction 1. to ask motion vector predictor, and with it as exploring window center (SA1) 1..This explore window 1. in, pattern 0 is detected make cost C earlier
pFor the motion vector (and reference frame) of minimum (SA2).At this moment, for each sub-piece of macro block, will all 1. keep (SA3) to the prediction residual power ((SAD) calculates with the prediction and evaluation value) that the 1. interior motion vector candidate of exploration window obtains as sad value.1. sad value can utilize (SA4) once more when using identical MV prediction 1. to carry out the motion vector of pattern 9,10,13,14 and reference frame decision.That is, only, just can determine to make cost C about pattern 9,10,13,14 from 1. selecting the value that keeps as sad value and carrying out addition calculation
pBe the motion vector and the reference frame of minimum, thereby needn't carry out the calculating of prediction residual power (SAD).
Secondly, at SA5,2., 3. pattern 1~8,11,12 is judged whether to use the MV prediction.Here, for use the MV prediction 2., 3. pattern 1~8,11,12, use MV prediction 2., 3. in the different reference frame of motion vector of predicted position of use, when being in-line coding, just judge and to use the MV prediction 2., 3., thereby use sad value 1. to determine motion vector and reference frame (SA8).On the other hand, at SA5, if judge and can use the MV prediction 2., 3., just will use motion vector predictor that 2. MV predict, 3. obtain as the exploration window 2., 3. center (SA6), this explore window 2., 3. in, pattern 1~8,11,12 is detected make cost C respectively
pFor the motion vector (and reference frame) of minimum (SA7).Then, in SA9~12, the lower section of pattern 9,10,13,14 is carried out the processing identical with SA5~8.And, last, at SA13, from all mode, select to realize minimum cost C
pPattern.
Handle by above motion vector detection, in SA4,7,8,11, each step of 12, the size decision motion vector computation order by according to section can and then reduce the SAD operand.For example, in the relation of pattern 9,10,13,14,, pattern 13 is arranged for upper segment
Pattern 9
Pattern 10
The relation of pattern 14.That is, the SAD result calculated of the upper segment of pattern 13 still can be utilized in the SAD of the upper segment of pattern 9 calculates, thus with 1. directly carry out addition calculation according to sad value and compare, can reduce the addition calculation number of times.We can say it also is the same for pattern 1~8,11,12, so, handle by carry out motion detection according to their level, can suppress operand.
On the other hand, in code translator, in the decoding of the motion vector grammer of variable length decoding part 25 is handled, can import the motion-vector prediction rule that the motion-vector prediction rule shown in Figure 37 A~Figure 37 D replaces Figure 10 A~Figure 10 D and decide predicted value, motion vector is restored.
Below, embodiment 6 is described.In the present embodiment, coding and the interpretation method of using, can select bi-directional predicted prediction direction to each section of cutting apart in the macro block is described in to the motion vector partition graph of embodiment 1 bidirectional interframe predictive that not only forward inter prediction but also reverse inter prediction also use.
Present embodiment is the same with the situation of embodiment 1, formerly have in the code device illustrated in figures 1 and 2 and code translator of example, predictive mode 3 in being included in compressive flow 24, the grammer of motion vector 4, motion detection portion 2 are different with the action of dynamic compensating unit 5, can describe with identical parts and action.Below, for code device, be syntax of declaration when the center describes in action with motion detection portion 2, in addition,, be that the center describes with the order of generation forecast image in dynamic compensating unit 5 for code translator.
The definition that the promptly fixing motion vector of the motion compensation model of present embodiment distributes, use with in the identical definition shown in the embodiment 1.That is, shown in Fig. 6 A~Fig. 6 D, the level that is not simple of cutting apart that is divided into two of macro block is cut apart or vertical segmentation, expands to the motion vector of cutting apart that comprises level, vertical, inclination and phase place and distributes and be to use.
In the present embodiment, for the section of cutting apart by motion vector, each section and then can select the prediction direction of bidirectional interframe predictive.Like this, in 1 each macro block during the border of the different object of existence activity, for this section that can catch the various activities in the macro block by few expense reliably, and then can from bi-directional predicted a plurality of prediction direction, select and distribute to obtain the prediction direction of higher code efficiency, thereby can improve forecasting efficiency.
Figure 39 represents the example of the bi-directional predicted direction of present embodiment.In Figure 39, be illustrated in bi-directional predicted frames B
tMacro block M in, the cutting apart of pattern 9 of using Fig. 6, and, 2 each sections in the macro block are used in time in front reference frame P respectively
T-1forward predict F1 and be the reference frame P of back in time
T+1The situation of backward-predicted B1.
When carrying out the coding of macro block M, in the motion detection portion 2 of Fig. 1,, at first from each predictive mode of Fig. 6 A~Fig. 6 D, estimate and detect the predictive mode and the motion vector that can obtain higher code efficiency as inter-frame forecast mode.At this moment, in the present embodiment, in each predictive mode, and then 2 sections are estimated and detect the prediction direction that can obtain higher code efficiency from bi-directional predicted a plurality of prediction direction.The result that Figure 39 represents these evaluations and detection in macro block M as predictive mode preference pattern 9 and for the situation of the different prediction direction of each section selection.This selection result is transmitted as the information in the compressive flow shown below.
Will from the grammer of the macro block data of the compressive flow of the code device of present embodiment output with H.26L the bi directional motion compensation prediction time grammer (Figure 40 A) comparatively be shown in Figure 40 B.
In the grammer H.26L shown in Figure 40 A, the information of prediction direction only to 1 of macro block transmission, is discerned with predictive mode in the predictive mode macro block type information of expression motion vector partition graph.During grammer when using bi-directional predicted that the promptly fixing motion vector of motion compensation model of the present invention distributes, also adopt same structure, comprise information, can be used as the predictive mode macro block type information and constitute and transmit about the choice situation of the prediction direction of each section.But in the present embodiment, the structure of employing is different with structure H.26L, and employing prevents the structure that the Variable Length Code table of predictive mode is grown up.
Figure 40 B represents the grammer of present embodiment.Here, the information of prediction direction is to only to the macro block type information transmission of the identification of direct (Direct) pattern of becoming the predictive mode suitable with skipped macroblocks in bi-directional predicted, the information of other prediction direction is transmitted as reference frame information, discerns with the reference frame number.In addition, in H.26L, do not use the piece size that imports for the predictive mode of discerning bi-directional predicted middle all directions.Arrow P 28~P36 of Figure 40 A, B represents to omit corresponding processing and the situation of jumping, and arrow Q12~Q19 represents circulation.
Figure 41 represents the code table of the macro block type information of present embodiment.The structure of this code table can be identical with the structure of the macro block type information of using in embodiment 1 except Direct Model is positioned at the beginning of code table.In addition, omitted the definition to later inside 16 * 16 patterns of code number 17, still, these patterns all are in the code table H.26L inner 16 * 16 patterns are defined as benchmark.Follow-up grammer is different with the content of macro block type information, in a lot of predictive modes, be used for each section and have reference frame information, so, there are 2 reference frame information, in the pattern 0 of the predictive mode of cutting apart as the section of not having, reference frame information has only 1, in addition, and in inner 4 * 4 patterns, there is not reference frame information, has intra prediction mode information.
Figure 42 represents the code table of the reference frame information of present embodiment.In the structure of H.26L reference frame information, being used to discern forward, the number of the reference frame number of prediction becomes tactic structure, here, except the reference frame number, as the prediction direction in bi-directional predicted, comprise forward (Forward), reverse (Backward), use forward the information with two-way (Bidirectional) of the mean value of reverse predicted value, in addition, have the reference frame information of each section.For forward prediction, can be from this selection of a plurality of reference frame, so, for forward prediction and bi-directional predicted, change each code sequence arrangement of reference frame forward.Because with reference frame information Recognition prediction direction, so, follow-up grammer is difference with the content of reference frame information, becoming when forward predicting, have forward motion vector difference value FW, in backward-predicted, have reverse movement phasor difference score value BW, in bi-directional predicted, have motion vector difference value FW and motion vector difference value BW.In addition, the number of each motion vector difference is different with the predictive mode of discerning according to macro block type information.
When only existing 1 (1frame back of Forward), the reference frame number that the reference frame message code of Figure 42 is forward predicted when not using the function of a plurality of reference frame H.26L also exists in each section, at this moment, only bi-directional predicted prediction direction is discerned according to reference frame information.
On the other hand, in code translator, decipher processing according to these information that obtain from compressive flow, this is handled with identical in the flow process shown in the embodiment 1, so, omit its detailed description here.The code translator of the present embodiment place different with the code translator of embodiment 1 is, the information that obtains from compressive flow, comprise information about the choice situation of the bi-directional predicted prediction direction different with each section of predictive mode, according to this situation, the dynamic compensating unit of Fig. 25 uses with each section different bi-directional predicted prediction direction to carry out motion compensation and generation forecast image.Like this, just can generate and can obtain the prediction direction of higher forecasting efficiency and the predicted picture that obtains each section distribution, thus can be according to the be improved decoding image of code efficiency of compressive flow.
In the present embodiment, as the motion compensation model, use the model shown in Fig. 6 A~Fig. 6 D, still, also can and then use the motion compensation model of allowing a lot of motion vector bar numbers shown in Fig. 4 A~Fig. 4 G simultaneously or switch to the compensation model of Fig. 6 A~Fig. 6 D.At this moment, the change that can predict each piece of cutting apart, can be considered as cutting apart number greater than 2, and can use grammer shown in the present embodiment and coding, interpretation method equally, each piece in the macro block is selected and distributed the prediction direction that can obtain higher code efficiency, can improve forecasting efficiency.At this moment, cutting apart number greater than 2 o'clock, the expense of the information of the choice situation when each piece is carried out the selection of prediction direction increases, so, the selection of prediction direction is only cut apart in use and is allowed when number is 2 predictive mode, can switch the information of whether transmitting about the choice situation of each piece according to the predictive mode that uses in each macro block.
Information about the choice situation of the prediction direction of each section is transmitted as additional information, so, being when not needing to carry out the simple image of such switching, can be chosen in the identical rule of macro block intra prediction direction.At this moment, can consider that demultiplexing is used to discern the flag information of whether allowing the prediction direction change of carrying out in the macro block in for example frame or frame group's unit.Like this, only there is 1 grammer in the code translator grammer that can be used as the choice situation that comprises prediction direction when the macro block intra prediction direction does not change to macro block and moves.
In addition, in the present embodiment, coding, the interpretation method of the prediction direction of the two inter predictions uses of each section selection in the macro block being carried out motion compensated prediction have been described, but, also can constitute and utilize identical method each section selection inter prediction in the macro block or infra-frame prediction to be carried out coding, the interpretation method of inter prediction or infra-frame prediction.By adopting such structure, can only select and the distribution infra-frame prediction, thereby can and then improve forecasting efficiency the section that can not predict effectively by inter prediction.
Coding method and interpretation method according to above-mentioned present embodiment, can catch various activities in the macro block reliably by few expense, simultaneously, two inter predictions or infra-frame prediction that can be suitable to each section application of catching other activities respectively, thus can carry out image transmission, record, regeneration expeditiously.
In order to reach the purpose of the invention described above, image processing system of the present invention can adopt following various version.
Promptly in image processing system of the present invention, can adopt following version, promptly, in code device, a certain group of a plurality of graphical set of figure marshalling selected preferentially will to distribute by the motion vector to the 2nd of the 1st block unit decision in the motion compensated prediction unit, according to the distribution figure in the graphical set that is included in this selection to the 2nd assigned motion vector, carry out motion compensated prediction, the assignment information coding unit will be about distributing the identifying information of figure by the information of the motion vector allocation situation of the 1st block unit coding as motion vector specific from the graphical set of above-mentioned selection and exporting; In code translator, the assignment information decoding unit distributes the information of figure to decipher as identification by the motion vector to the 2nd that the 1st block unit is predetermined the motion vector assignment information, according to the specific distribution figure of this Figure recognition information the time, information decoding with the graphical set under this figure of identification, according to definition, distribute according to the motion vector of above-mentioned distribution figure identifying information decision to the 2nd by the specific graphical set of this graphical set identifying information.
In addition, in image processing system of the present invention, can adopt following version, promptly, in code device, the predicted value that the assignment information coding unit distributes the shape decision motion vector of figure according to the motion vector that becomes the 1st of coded object, and the predicted difference score value of this predicted value encoded as motion vector information; In code translator, the assignment information decoding unit will be deciphered the motion vector information that obtains of encoding according to the predicted difference score value of the predicted value that becomes decoding motion of objects vector.
In addition, in image processing system of the present invention, can adopt following version, promptly, in code device, the assignment information coding unit distributes the prediction candidate position of the shape decision motion vector of figure according to the 1st the motion vector that becomes coded object, to predict the some of candidate position according to being arranged in the shape that becomes near the 1st distribution figure the 1st of above-mentioned coded object, the predicted difference score value of this predicted value will be encoded as motion vector information as predicted value; The assignment information decoding unit will distribute prediction candidate position some as predicted value of the motion vector that the shape of figure determine according to the 1st the motion vector that becomes the decoding object according near the shape of the 1st distribution figure being arranged in code translator, will decipher by the motion vector information that obtains that the predicted difference score value of this predicted value is encoded.
In addition, in image processing system of the present invention, can adopt following version, promptly, in code device, the motion compensated prediction unit can be counted maximum by the motion vector that the 2nd block unit distributes and be taken as 2, simultaneously, according to 2nd piece the distribution state of expression in the 1st that is included in by the 1st block unit decision, some in a plurality of distribution figures that are predetermined to the 2nd assigned motion vector, carry out motion compensated prediction, the assignment information coding unit will about the information of allocation situation of motion vector of the 1st block unit decision export as the identifying information of above-mentioned distribution figure; In code translator, the motion vector maximum of deciphering by the 1st unit is taken as 2, and the motion vector assignment information is deciphered to the information of a plurality of distribution figures distribution state, that be predetermined that comprise the 2nd motion vector in the 1st that determines by the 1st block unit as the identification expression.
In addition, in image processing system of the present invention, can adopt following version, that is, code device has by by each the 1st of the frame of cutting apart moving image and then one or more units that constitute of the 2nd cutting apart being selected 1 reference frame carry out the compensation prediction unit of motion compensated prediction from a plurality of reference frame that keep in order to carry out motion compensated prediction and will being turned to bit stream about the information multichannel to the choice situation of one or more reference frame of the 2nd and the reference frame information coding unit exported; Thereby code translator has reference frame information decoding unit that the information of the choice situation of the reference frame of will distribute by the unit to each the 1st of the frame of cutting apart moving image and then one or more formations of the 2nd cutting apart, use about motion compensated prediction deciphers and according to this reference frame information from selecting 1 reference frame to carry out the compensating unit of motion compensation generation forecast image a plurality of reference frame that motion compensation keeps in order to carry out.
In addition, in image processing system of the present invention, can adopt following version, that is, code device has by the motion compensated prediction unit that is carried out motion compensated prediction by the prediction direction of selecting bidirectional interframe predictives to use to each the 1st of the frame of cutting apart moving image and then one or more the 2nd unit that constitute of cutting apart and will turn to bit stream about the information multichannel to the choice situation of above-mentioned one or more prediction direction of the 2nd and the information of forecasting coding unit exported; Thereby code translator have information of forecasting decoding unit that the information of the choice situation of the prediction direction of will distribute by the unit to each the 1st of the frame of cutting apart moving image and then one or more the 2nd formation of cutting apart, use about bidirectional interframe predictive deciphers and according to this information of forecasting to forward with the reference frame of reverse maintenance select two-way inter motion compensation to use prediction direction carry out the motion compensation units of motion compensation generation forecast image.
In addition, in image processing system of the present invention, can adopt following version, that is, code device has by by each the 1st of the frame of cutting apart moving image and then one or more the 2nd unit that constitute of cutting apart being selected inter-frame forecast modes or intra prediction mode carry out the predicting unit of inter prediction or infra-frame prediction and will being turned to bit stream about the information multichannel to the choice situation of above-mentioned one or more the 2nd inter-frame forecast modes or intra prediction mode and the prediction mode information coding unit exported; Thereby code translator has the prediction mode information decoding unit that will distribute by the unit to each the 1st of the frame of cutting apart moving image and then one or more the 2nd formation of cutting apart, decipher about the information of the choice situation of inter-frame forecast mode or intra prediction mode and select inter-frame forecast mode or intra prediction mode to carry out the predicted picture generation unit of inter prediction or infra-frame prediction generation forecast image according to this information of forecasting.
The possibility of utilizing on the industry
As mentioned above, according to the present invention, can catch reliably in the 1st by few expense Various activities, thereby can carry out expeditiously image transmitting, record, regeneration.
Claims (9)
1. a coding method of using motion compensated prediction to carry out the compressed encoding of moving image is characterized in that: comprise
With each the 1st the 2nd of further cutting apart gained to the frame gained of cutting apart moving image is unit, distribute one or more motion vectors, and with by one or more above-mentioned the 2nd units that constitutes, from a plurality of reference frame that keep in order to carry out motion compensated prediction, select 1 reference frame, carry out the motion compensated prediction step of motion compensated prediction;
With the assignment information coding step that turns to bit stream at the relevant information multichannel of above-mentioned the 2nd allocation situation and export with motion vector; With
With the reference frame information coding step that turns to bit stream at the relevant information multichannel of above-mentioned one or more choice situations of the 2nd and export with reference frame.
2. by the described coding method of claim 1, it is characterized in that: in above-mentioned reference frame information coding step, encode as the combined information that in above-mentioned the 1st unit, comprises at the choice situation of above-mentioned one or more reference frame of the 2nd.
3. by the described coding method of claim 1, it is characterized in that: in above-mentioned reference frame information coding step, ask predicted value respectively according to the choice situation of above-mentioned the 1st or one or more reference frame of the 2nd around being arranged at one or more reference frame of the 2nd, with the difference information between this predicted value and the selected reference frame as the information use relevant with the choice situation of reference frame.
4. one kind is used motion compensated prediction to carry out the decompress(ion) of compression movement view data, the interpretation method of decoding, it is characterized in that: comprise
Is the motion vector that unit distributes gained with expression with each the 1st the 2nd of further cutting apart gained to gained behind the frame of cutting apart moving image, be that corresponding which motion vector assignment information by in one or more motion vectors of above-mentioned the 1st block unit decoding is deciphered, and the assignment information that the unit with one or more above-mentioned the 2nd formations distributes information gained, relevant with the choice situation of employed reference frame in motion compensated prediction to decipher is deciphered step;
According to this motion vector assignment information, use with above-mentioned the 2nd to be the specific motion vector of unit, thereby carry out the motion compensation step of motion compensation generation forecast image; With
According to the information relevant, from selecting 1 reference frame a plurality of reference frame that motion compensation kept, thereby carry out the compensation process of motion compensation generation forecast image in order to carry out with the choice situation of above-mentioned reference frame.
5. by the described interpretation method of claim 4, it is characterized in that: in above-mentioned reference frame information coding step, according to the combined information that in above-mentioned the 1st unit, comprises, will decipher at above-mentioned one or more reference frame information of the 2nd at the choice situation of above-mentioned one or more reference frame of the 2nd.
6. by the described interpretation method of claim 4, it is characterized in that: in above-mentioned reference frame information decoding step, ask predicted value respectively according to the choice situation of above-mentioned the 1st or one or more above-mentioned the 2nd reference frame around being arranged in, decipher according to the difference information and the above-mentioned predicted value of obtaining of the reference frame of exporting as the information relevant with the choice situation of reference frame at one or more reference frame of the 2nd.
7. a code device that uses motion compensated prediction to carry out the compressed encoding of moving image is characterized in that: comprise
With each the 1st the 2nd of further cutting apart gained to the frame gained of cutting apart moving image is unit, distribute one or more motion vectors, and with by one or more above-mentioned the 2nd units that constitutes, from a plurality of reference frame that keep in order to carry out motion compensated prediction, select 1 reference frame, carry out the motion compensated prediction unit of motion compensated prediction;
With the assignment information coding unit that turns to bit stream at the relevant information multichannel of above-mentioned the 2nd allocation situation and export with motion vector; With
With the reference frame information coding unit that turns to bit stream at the relevant information multichannel of above-mentioned one or more choice situations of the 2nd and export with reference frame.
8. one kind is used motion compensated prediction to carry out the decompress(ion) of compression movement view data, the code translator of decoding, it is characterized in that: comprise
Is the motion vector that unit distributes gained with expression with each the 1st the 2nd of further cutting apart gained to gained behind the frame of cutting apart moving image, be that corresponding which motion vector assignment information by in one or more motion vectors of above-mentioned the 1st block unit decoding is deciphered, and the assignment information decoding unit that distributes information gained, relevant with the choice situation of employed reference frame in motion compensated prediction to decipher the unit with one or more above-mentioned the 2nd formations;
According to this motion vector assignment information, use with above-mentioned the 2nd to be the specific motion vector of unit, thereby carry out the motion compensation units of motion compensation generation forecast image; With
According to the information relevant, from selecting 1 reference frame a plurality of reference frame that motion compensation kept, thereby carry out the compensating unit of motion compensation generation forecast image in order to carry out with the choice situation of above-mentioned reference frame.
One kind comprise use motion compensated prediction carry out moving image compressed encoding code device and use motion compensated prediction to carry out the code translator of the decompress(ion) of compression movement view data and decoding and the image processing system that constitutes, it is characterized in that:
Above-mentioned code device has
With each the 1st the 2nd of further cutting apart gained to the frame gained of cutting apart moving image is unit, distribute one or more motion vectors, and with by one or more above-mentioned the 2nd units that constitutes, from a plurality of reference frame that keep in order to carry out motion compensated prediction, select 1 reference frame, carry out the motion compensated prediction unit of motion compensated prediction;
With the assignment information coding unit that turns to bit stream at the relevant information multichannel of above-mentioned the 2nd allocation situation and export with motion vector; With
With the reference frame information coding unit that turns to bit stream at the relevant information multichannel of above-mentioned one or more choice situations of the 2nd and export with reference frame,
Above-mentioned code translator has
Is the motion vector that unit distributes gained with expression with each the 1st the 2nd of further cutting apart gained to gained behind the frame of cutting apart moving image, be that corresponding which motion vector assignment information by in one or more motion vectors of above-mentioned the 1st block unit decoding is deciphered, and the assignment information decoding unit that distributes information gained, relevant with the choice situation of employed reference frame in motion compensated prediction to decipher the unit with one or more above-mentioned the 2nd formations;
According to this motion vector assignment information, use with above-mentioned the 2nd to be the specific motion vector of unit, thereby carry out the motion compensation units of motion compensation generation forecast image; With
According to the information relevant, from selecting 1 reference frame a plurality of reference frame that motion compensation kept, thereby carry out the compensating unit of motion compensation generation forecast image in order to carry out with the choice situation of above-mentioned reference frame.
Applications Claiming Priority (9)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2001279266 | 2001-09-14 | ||
JP2001-279266 | 2001-09-14 | ||
JP2001279266 | 2001-09-14 | ||
JP2001343540 | 2001-11-08 | ||
JP2001-343540 | 2001-11-08 | ||
JP2001343540 | 2001-11-08 | ||
JP2002-002944 | 2002-01-10 | ||
JP2002002944 | 2002-01-10 | ||
JP2002002944 | 2002-01-10 |
Related Parent Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN028032225A Division CN1513268B (en) | 2001-09-14 | 2002-09-12 | Coding method, decoding method, coding apparatus, decoding apparatus, image processing system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101409838A true CN101409838A (en) | 2009-04-15 |
CN101409838B CN101409838B (en) | 2011-01-12 |
Family
ID=40332561
Family Applications (4)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN200810145318.3A Expired - Lifetime CN101409838B (en) | 2001-09-14 | 2002-09-12 | Coding method, decoding method, coding apparatus, decoding apparatus, and image processing system |
CN200810145317.9A Expired - Lifetime CN101409837B (en) | 2001-09-14 | 2002-09-12 | Coding method, decoding method, coding apparatus, decoding apparatus, and image processing system |
CN200810145323.4A Expired - Lifetime CN101360240B (en) | 2001-09-14 | 2002-09-12 | Coding method, decoding method, coding apparatus, decoding apparatus, and image processing system |
CN200810145322.XA Expired - Lifetime CN101360239B (en) | 2001-09-14 | 2002-09-12 | Coding method, decoding method, coding apparatus, decoding apparatus, and image processing system |
Family Applications After (3)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN200810145317.9A Expired - Lifetime CN101409837B (en) | 2001-09-14 | 2002-09-12 | Coding method, decoding method, coding apparatus, decoding apparatus, and image processing system |
CN200810145323.4A Expired - Lifetime CN101360240B (en) | 2001-09-14 | 2002-09-12 | Coding method, decoding method, coding apparatus, decoding apparatus, and image processing system |
CN200810145322.XA Expired - Lifetime CN101360239B (en) | 2001-09-14 | 2002-09-12 | Coding method, decoding method, coding apparatus, decoding apparatus, and image processing system |
Country Status (1)
Country | Link |
---|---|
CN (4) | CN101409838B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102907100A (en) * | 2010-05-17 | 2013-01-30 | Sk电信有限公司 | Image coding/decoding device using coding block in which intra block and inter block are mixed, and method thereof |
CN103329538A (en) * | 2011-01-15 | 2013-09-25 | Sk电信有限公司 | Method and device for encoding/decoding image using bi-directional intra prediction |
CN103329538B (en) * | 2011-01-15 | 2016-11-30 | Sk电信有限公司 | Utilize the method and apparatus that image is encoded/decoded by bi-directional intra prediction |
CN107318015A (en) * | 2016-04-26 | 2017-11-03 | 谷歌公司 | Hybrid predicting pattern for Video coding |
Families Citing this family (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20100095992A (en) | 2009-02-23 | 2010-09-01 | 한국과학기술원 | Method for encoding partitioned block in video encoding, method for decoding partitioned block in video decoding and recording medium implementing the same |
EP3998776B1 (en) * | 2009-05-29 | 2023-08-30 | Mitsubishi Electric Corporation | Image encoding device, image decoding device, image encoding method, and image decoding method |
MX364986B (en) * | 2010-02-24 | 2019-05-17 | Velos Media Int Ltd | Image encoding device, image decoding device, and data structure. |
TW201515445A (en) * | 2010-04-01 | 2015-04-16 | Sony Corp | Image processing device and method |
JP2012023597A (en) * | 2010-07-15 | 2012-02-02 | Sony Corp | Image processing device and image processing method |
WO2014087861A1 (en) * | 2012-12-06 | 2014-06-12 | ソニー株式会社 | Image processing device, image processing method, and program |
US20150055697A1 (en) * | 2013-08-20 | 2015-02-26 | Media Tek Inc. | Method and Apparatus of Transform Process for Video Coding |
CN103561267B (en) * | 2013-09-10 | 2016-08-31 | 罗天明 | 3D Video coding transmission method based on movable information Yu depth information |
US9491473B2 (en) * | 2013-10-03 | 2016-11-08 | Amlogic Co., Limited | Motion compensated de-interlacing and noise reduction |
CN110015125B (en) * | 2017-08-17 | 2021-12-21 | 上海汽车集团股份有限公司 | Method and device for coding and transmitting cell voltage |
WO2019065915A1 (en) | 2017-09-29 | 2019-04-04 | 株式会社ニコン | Moving image compressing device, electronic equipment, and moving image compressing program |
US12022095B2 (en) * | 2019-03-15 | 2024-06-25 | Qualcomm Incorporated | Video coding with unfiltered reference samples using different chroma formats |
US11475601B2 (en) * | 2019-10-21 | 2022-10-18 | Google Llc | Image decoding during bitstream interruptions |
CN111797398B (en) * | 2020-06-28 | 2024-02-02 | 韩山师范学院 | Malicious code visualization and variant detection method, system, equipment and storage medium |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AU5632394A (en) * | 1993-03-05 | 1994-09-08 | Sony Corporation | Apparatus and method for reproducing a prediction-encoded video signal |
US5731840A (en) * | 1995-03-10 | 1998-03-24 | Kabushiki Kaisha Toshiba | Video coding/decoding apparatus which transmits different accuracy prediction levels |
JP2001238220A (en) * | 1995-10-27 | 2001-08-31 | Toshiba Corp | Moving picture coding apparatus and moving picture coding method |
JP3356004B2 (en) * | 1997-05-30 | 2002-12-09 | 日本ビクター株式会社 | Variable rate coding apparatus and method |
CA2265089C (en) * | 1998-03-10 | 2007-07-10 | Sony Corporation | Transcoding system using encoding history information |
JP3775630B2 (en) * | 1998-07-29 | 2006-05-17 | Kddi株式会社 | Video segmentation coding apparatus |
JP3694888B2 (en) * | 1999-12-03 | 2005-09-14 | ソニー株式会社 | Decoding device and method, encoding device and method, information processing device and method, and recording medium |
-
2002
- 2002-09-12 CN CN200810145318.3A patent/CN101409838B/en not_active Expired - Lifetime
- 2002-09-12 CN CN200810145317.9A patent/CN101409837B/en not_active Expired - Lifetime
- 2002-09-12 CN CN200810145323.4A patent/CN101360240B/en not_active Expired - Lifetime
- 2002-09-12 CN CN200810145322.XA patent/CN101360239B/en not_active Expired - Lifetime
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102907100A (en) * | 2010-05-17 | 2013-01-30 | Sk电信有限公司 | Image coding/decoding device using coding block in which intra block and inter block are mixed, and method thereof |
CN102907100B (en) * | 2010-05-17 | 2016-09-28 | Sk电信有限公司 | The video encoding/decoding apparatus of the encoding block that use intra block mixes mutually with interframe block and method thereof |
CN103329538A (en) * | 2011-01-15 | 2013-09-25 | Sk电信有限公司 | Method and device for encoding/decoding image using bi-directional intra prediction |
CN103329538B (en) * | 2011-01-15 | 2016-11-30 | Sk电信有限公司 | Utilize the method and apparatus that image is encoded/decoded by bi-directional intra prediction |
US9544584B2 (en) | 2011-01-15 | 2017-01-10 | Sk Telecom Co., Ltd. | Method and apparatus for encoding/decoding video using bidirectional intra prediction |
CN107318015A (en) * | 2016-04-26 | 2017-11-03 | 谷歌公司 | Hybrid predicting pattern for Video coding |
CN107318015B (en) * | 2016-04-26 | 2020-02-07 | 谷歌有限责任公司 | Video signal encoding method, decoding method and decoding device |
Also Published As
Publication number | Publication date |
---|---|
CN101360240A (en) | 2009-02-04 |
CN101409837B (en) | 2011-07-13 |
CN101409838B (en) | 2011-01-12 |
CN101360240B (en) | 2012-12-05 |
CN101409837A (en) | 2009-04-15 |
CN101360239A (en) | 2009-02-04 |
CN101360239B (en) | 2010-10-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN1513268B (en) | Coding method, decoding method, coding apparatus, decoding apparatus, image processing system | |
US10999597B2 (en) | Image encoding method and image decoding method | |
US20240223752A1 (en) | Method and apparatus for processing a video signal | |
CN110999304B (en) | Image processing method and image encoding/decoding method, and apparatus using the same | |
CN101409838B (en) | Coding method, decoding method, coding apparatus, decoding apparatus, and image processing system | |
CN1905679B (en) | Image encoding device, image encoding method, image decoding device, image decoding method, and communication device | |
US8948243B2 (en) | Image encoding device, image decoding device, image encoding method, and image decoding method | |
JP7311635B2 (en) | Codec method, device and equipment therefor | |
KR100987764B1 (en) | Method of and apparatus for determining reference data unit for predictive video data coding | |
KR102227680B1 (en) | Method and apparatus for fast image encoding for image compression | |
CN118414825A (en) | Prediction method, prediction device, prediction apparatus, prediction system, and storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CX01 | Expiry of patent term |
Granted publication date: 20110112 |
|
CX01 | Expiry of patent term |