CN101622877A - Systems and methods for efficient spatial intra predictabilty determination (or assessment) - Google Patents

Abstract

Techniques for efficient determination of a macroblock's spatial predictability quality with respect to the H.264 specification are provided. A device comprises a processor operable to estimate a first subset of spatial prediction modes based on a first pixel subset of a current subblock of an intra-frame. The processor is also operable to estimate a second subset of spatial prediction modes based on a second pixel subset of the current subblock. The first subset of prediction modes is different from the second subset of prediction modes.

Description

The system and method that is used for predictability definite (or assessment) in the useful space frame

The application's case is advocated the rights and interests of the 60/871st, No. 648 U.S. Provisional Application case of application on December 22nd, 2006, and the full content of described application case is incorporated herein by reference.

Technical field

The present invention relates generally to video coding and decoding, and more particularly, relates to the technology of determining the space predictability quality of macro block with respect to standard H.264 effectively.

Background technology

The intra-encoded frame that can independently decode is for realizing the frame to the random-access common form of vision signal.MPEG-x and H.26x standard use the thing be called image sets (GOP), its comprise intra-encoded frame (also being called the I frame) with through the P-frame of interim prediction or with reference to the I frame in the GOP and/or other P and/or B frame through bi-directional predicted B frame.Long GOP is desirable for the compression ratio that increases, but short GOP allows to obtain faster and arbitrary access.The number that increases the I frame will permit obtaining faster and arbitrary access, but with lower boil down to cost.

Before, for predictability in the air-frame of determining macro block, used based on the algorithm of exhausting formula search or rim detection.Yet, complicated on these algorithm computation, and specifically, produce estimation based on the algorithm of rim detection with higher relatively error probability.

Known space predicting method does not generally provide and makes computational complexity minimize the method stage by stage of yet the error probability that provides relatively low in the estimation procedure simultaneously.By calculating effective means and being the important component part of the mode decision algorithm carried out by encoder/transcoder with predictability in the frame of high accuracy assessment macro block.

Therefore, exist effectively determine the technology requirement of the space predictability quality of macro block with respect to standard H.264.

Summary of the invention

The technology of effectively determining the space predictability quality of macro block with respect to standard H.264 is provided.A kind of device is provided, and it comprises processor, and described processor can be operated with first subset of pixels based on the current sub-block of frame in the frame and estimate the first spatial prediction mode subclass.Described processor also can be operated to estimate the second spatial prediction mode subclass based on second subset of pixels of described current sub-block.The described first predictive mode subclass is different with the described second predictive mode subclass, and on the one hand, described first subset of pixels and described second subset of pixels do not comprise mutually.Described device also comprises memory, and it is coupled to described processor.In another aspect, described first subset of pixels and described second subset of pixels are overlapped.

In aspect another, provide a kind of multimedia system, it comprises encoder, and described encoder can be operated with first subset of pixels based on the current sub-block of frame in the frame and estimate the first spatial prediction mode subclass.Described encoder can be operated to estimate the second spatial prediction mode subclass based on second subset of pixels of described current sub-block.The described first predictive mode subclass is different with the described second predictive mode subclass.Described system further comprises decoder, and described decoder can be operated with frame in the described frame of decoding based on the described first spatial prediction mode subclass and the described second spatial prediction mode subclass.

According to detailed description, especially when in conjunction with the accompanying drawings, will understand additional aspect easily.

Description of drawings

Fig. 1 explanation is according to the block diagram of the exemplary multimedia communications system of some embodiment.

Fig. 2 A illustrates the block diagram of the exemplary encoder device in the system that can be used for Fig. 1.

Fig. 2 B illustrates the block diagram of the example decoder device in the system that can be used for Fig. 1.

In the frame in Fig. 3 A explanation H.264 _ the 4x4 predictive mode.

Fig. 3 B explanation is according to the directivity classification of the grouping of some embodiment.

The partial view of (I) frame in the frame of Fig. 4 explanation according to a certain embodiment.

Fig. 5 A explanation is according to first subset of pixels of the sub-piece in border of some embodiment.

Fig. 5 B explanation is according to second subset of pixels of the sub-piece in border of a certain embodiment.

Fig. 6 A explanation is according to first subset of pixels of the sub-piece in inside of some embodiment.

Fig. 6 B explanation is according to second subset of pixels of the sub-piece in inside of some embodiment.

Fig. 6 C explanation is according to an alternative arrangements of second subset of pixels of the sub-piece in inside of some embodiment.

Fig. 7 explanation is according to the block diagram of the intra prediction mode taxon of some embodiment.

Fig. 8 illustrates that the first intra prediction direction sort module handles the flow chart of the calculating of sub-piece.

Fig. 9 illustrates that the second intra prediction direction sort module handles the flow chart of the calculating of sub-piece.

Figure 10 illustrates that power calculates and summation module.

Figure 11 is to the flow chart of Figure 13 explanation according to the process of the accurate estimation intra prediction direction sexual norm of some embodiment.

For illustrative purposes, the image in graphic is simplified, and and not drawn on scale.For promoting to understand, under possible situation, used same reference numbers to represent common similar elements for described figure, just adding suffix in due course distinguishes these elements.

Description of drawings exemplary configuration of the present invention, and therefore, it should be considered as limiting other equal effectively scope of configuration of admitting of the present invention.Can be expected under the situation of not having further narration, feature or piece with a kind of configuration can advantageously be incorporated in other configuration.

Embodiment

" exemplary " this speech is in this article in order to mean " serving as example, example or explanation ".Be described as any configuration of " exemplary " or design herein and may not be interpreted as than other configuration or design preferably or favourable, and use term " core ", " engine ", " machine ", " processor " and " processing unit " interchangeably.

Below describe in detail at some sample embodiment.Yet the present invention can embody as the many different modes that defined by claims and contain.Ginseng is wherein represented same parts with same numbers in the whole text with reference to graphic in this describes.

Can characterize vision signal according to a series of pictures, frame and/or field (wherein any one can further comprise one or more fragments).As used herein, term " frame " is a broad terms, and it can contain one or more in frame, field, picture and/or the fragment.

Embodiment comprises the system and method for the channel switching that promotes in the multi-media transmission system.Multi-medium data can comprise one or more in the audio-visual data of sport video, audio frequency, rest image, text or any other adequate types.

For example multimedia processing system such as video encoder can use the coding method based on international standard to come the encoded multimedia data, described international standard for example is mobile photographic experts group (MPEG)-1 ,-2 and-4 standards, International Telecommunication Union-T is standard H.263, with ITU-T H.264 standard and its homologue, ISO/IEC MPEG-4, the 10th part (that is, advanced video coding (AVC)), each in the described standard all is incorporated herein by reference to be used for all purposes.These codings and (extending) coding/decoding method are directed to the compressing multimedia data usually to be used for transmission and/or storage.Can broadly compression be considered as removing redundant process from multi-medium data.

Can be according to the sequence description vision signal of picture, described picture comprises frame (whole picture) or field (for example, interlaced video stream comprises the odd number that replaces of picture or the field of even rows).In addition, each frame or field can further comprise the subdivision of one or more fragments or frame or field.As used herein, individually or with other word combination, term " frame " can refer to a picture, a frame, a field or one fragment.Method for video coding is by using harmless or the lossy compression method algorithm compresses each frame with compressed video signal.Intraframe coding (being called intraframe coding in this article) refers to only use a frame to that frame coding.Interframe encode (being called interframe encode in this article) refers to based on other " reference " frame one frame be encoded.For instance, vision signal often shows time redundancy, wherein in the time series of frame frame close to each other have each other coupling fully or to the small part coupling to small part.

For example the multimedia processor of video encoder can come frame is encoded by the subclass that a frame is divided into pixel.These subclass of pixel can be called piece or macro block, and can comprise (for example) 16 * 16 pixels.Encoder can further become sub-piece with each 16 * 16 macroblock partitions.Each sub-piece can further comprise extra sub-piece.For instance, the sub-piece of 16 * 16 macro blocks can comprise 16 * 8 and 8 * 16 sub-pieces.For instance, each in 16 * 8 and 8 * 16 sub-pieces can comprise 8 * 8 sub-pieces, and 8 * 8 sub-pieces self can comprise (for example) 4 * 4,4 * 8 and 8 * 4 sub-pieces or the like.As used herein, term " piece " can refer to the sub-piece of macro block or any size.

Encoder uses the interframe encode algorithm based on motion compensation to utilize time redundancy between the successive frame.Movement compensating algorithm is discerned the part to one of small part coupling in one or more reference frames.Can make described compatible portion displacement in frame with respect to reference frame.This displacement is characterized by one or more motion vectors.Any difference between the partial matching partial of piece and reference frame can characterize according to one or more residual errors.Encoder can be encoded to a frame and comprise at the motion vector of a particular division of described frame and the one or more data in the residual error.Can minimize the particular division that is used for coded frame of selecting piece by causing cost function (for example, the distortion of the content of its balance code frame big or small and that produce by encoding or the distortion that perceives) greatly.

Interframe encode realizes the compression efficiency higher than intraframe coding.Yet, when reference data (for example, reference frame or reference field) owing to channel error etc. and when losing, interframe encode can have problems.Except losing the reference data owing to wrong, reference data also may since vision signal through initially the obtaining or obtaining again of frame place of interframe encode and unavailable.In these cases, may maybe can cause improper mistake and error propagation for impossible to decoding through the data of interframe encode.These situations can cause the unacceptable mass loss in the video flowing.

Fig. 1 explanation is according to the block diagram of the exemplary multimedia communications system 100 of some embodiment.System 100 comprises an encoder apparatus 110, and it is communicated by letter with decoder device 150 via network 140.In an example, encoder apparatus 110 is from external source 102 multimedia signal receivings, and that signal of encoding is to be used for transmission on network 140.

In this example, encoder apparatus 110 comprises processor 112, and it is coupled to memory 114 and transceiver 116.Processor 112 is encoded from the data in multi-medium data source, and it is provided to transceiver 116 to be used for communication on network 140.

In this example, decoder device 150 comprises processor 152, and it is coupled to memory 154 and transceiver 156.Processor 152 can comprise one or more in general processor and/or the digital signal processor.Memory 154 can comprise solid-state or one or more based in the storage device of disk.Transceiver 156 is configured on network 140 receiving multimedia data and it is provided to processor 152 for decoding.In an example, transceiver 156 comprises wireless transceiver.Network 140 can comprise one or more in wired or the wireless communication system, comprise Ethernet, phone (for example, POTS), cable, one or more in power line and the fibre system, and/or comprise code division multiple access (CDMA or CDMA2000) communication system, frequency division multiple access (FDMA) system, OFDM (OFDM) system, one or more wireless system in time division multiple access (TDMA) system, for example GSM/GPRS (General Packet Radio Service)/EDGE (enhanced data gsm environment), terrestrial trunked radio (TETRA) mobile telephone system, Wideband Code Division Multiple Access (WCDMA) (WCDMA) system, high data rate (1xEV-DO or the multicast of 1xEV-DO gold) system, IEEE 802.11 systems, the MediaFLO system, DMB (DMB) system, digital video broadcast-handheld formula (DVB-H) system etc.

Fig. 2 A explanation is according to the block diagram of the exemplary encoder device 110 in the system that can be used for Fig. 1 100 of some embodiment.In this embodiment, encoder apparatus 110 comprises interframe encode encoder 118, intraframe coding encoder 120, reference data generator 122 and reflector 124.Interframe encode encoder 118 encoded videos through interframe encode allusion quotation part, described part through interframe encode is the other parts that are arranged in frame At All Other Times of reference video data and predicted in time (for example, using motion compensated prediction).The intra-encoded allusion quotation part that can independently decode in intraframe coding encoder 120 encoded videos, and not with reference to the video data of going up the location At All Other Times.In certain embodiments, but intraframe coding encoder 120 usage spaces predict and utilize the redundancy that is arranged in same time frame in the video data.

In one aspect, reference data generator 122 produces indication and is positioned at data where by produce intra-encoded of encoder 120 and 118 with through the video data of interframe encode respectively.For instance, reference data can comprise by decoder and is used in frame the piece of one position, location and/or the identifier of macro block.Reference data also can comprise the number of frames that is used for location one frame in sequence of frames of video.

Reflector 124 in the transmission over networks of the network 140 of for example Fig. 1 through interframe encode allusion quotation data, intra-encoded data and (in certain embodiments) reference data.Can be at one or more communication links transmission of data.Use the term communication link in general sense, and can comprise any communication channel, include, but is not limited to wired or wireless network, pseudo channel, optical link etc.In certain embodiments, in the defeated intra-encoded data of basic unit's communication links, and in the defeated data through interframe encode of enhancement layer communication links.In certain embodiments, in the defeated intra-encoded data of same communication links with through the data of interframe encode.In certain embodiments, can be in one or more in the data of interframe encode, intra-encoded data and reference data of transmission on the communication with side information link.For instance, for example can use H.264 supplemental enhancement information (SEI) message or the communication with side information link of the user_data message of MPEG-2.In certain embodiments, in intra-encoded data of transmission on the pseudo channel, one or more in the data of interframe encode and reference data.Pseudo channel can comprise packet, and described packet contains packet is identified as the packet headers discerned that belongs to described pseudo channel.Other form of known identification pseudo channel in this technology, for example, frequency division, time cuts apart, the sign indicating number exhibition frequently or the like.

Fig. 2 B explanation is according to the block diagram of the example decoder device 150 in the system that can be used for Fig. 1 100 of some embodiment.In this embodiment, decoder device 150 comprises receiver 158, selectivity decoder 160, reference data determiner 162 and one or more reference data availability detector, for example, and channel switch detector 164 and error detector 166.

Receiver 158 receives encoded video data (for example, by Fig. 1 and Fig. 2 A encoder apparatus 110 coded datas).Receiver 158 can be gone up at wired or wireless network (for example, the network 140 of Fig. 1) and receive encoded data.Can on one or more communication links, receive described data.In certain embodiments, on basic unit's communication link, receive intra-encoded data, and on the enhancement layer communication link, receive data through interframe encode.In certain embodiments, on same communication link, receive intra-encoded data and through the data of interframe encode.In certain embodiments, one or more in data through interframe encode, intra-encoded data and the reference data can received on the communication with side information link.For instance, for example can use H.264 supplemental enhancement information (SEI) message or the communication with side information link of the user_data message of MPEG-2.In certain embodiments, receiving intra-encoded data, one or more in the data of interframe encode and reference data on the pseudo channel.Pseudo channel can comprise packet, and described packet contains packet is identified as the bag the discerned header that belongs to described pseudo channel.Other form of known identification pseudo channel in this technology.

Interframe encode and intra-coded video data that 160 decodings of selectivity decoder are received.In certain embodiments, the data that received comprise the intraframe coding version of the described part of the interframe encode version of a part of video data and video data.Can decode through the data of interframe encode in decoding reference data (predict based on described reference data through interframe encode data) back.For instance, the data of use motion compensated predictive coding comprise the frame identifier of the position of motion vector and identification reference data.If reference frame is available (for example, decoded), the then described interframe encode version of selectivity decoder 160 decodable codes by the motion vector of interframe encode version and the part of frame identifier identification.Yet, if reference data is unavailable, selectivity decoder 160 decodable code intraframe coding versions.

In one aspect, the reference data that reference data determiner 162 identification is received, intra-encoded in the encoded video data that the reference data that is received indication is received and where be positioned at through the video data of interframe encode.For instance, reference data can comprise by selectivity decoder 160 and is used in frame the piece of one position, location and/or the identifier of macro block.Reference data also can comprise the number of frames that is used for location one frame in sequence of frames of video.Use this reference data that receives to make decoder can determine whether reference data (data through interframe encode depend on described reference data) is available.

The channel that the user is switched multichannel communication system can influence the reference data availability.For instance, by using one or more communication links, receiver 158 can utilize a plurality of video broadcastings.If user command receiver 158 is changed into different broadcast channels, then on new channel be used for may be not available immediately through the reference data of the data of interframe encode.Channel switch detector 164 detects that the channel switching commands have sent and with signalisation selectivity decoder 160.Whether selectivity decoder 160 can then use the information that obtains from the reference data determiner to discern the reference data of interframe encode version unavailable, and then discern the position of nearest intraframe coding version and the intraframe coding version discerned of decoding optionally.

The influence of the mistake in the video data that the reference data availability also can be subjected to being received.Error detector 166 can utilize error detection techniques (for example, forward error correction) to discern the uncorrectable error in the bit stream.If have uncorrectable error in reference data (the interframe encode version depends on described reference data), then error detector 166 can signalisation selectivity decoder 160, is subjected to erroneous effects thereby discern which video data.Selectivity decoder 160 can then determine it is still version of code (for example, if reference data is unavailable) in the decoded frame of version of code between decoded frame (for example, if reference data can with).

In certain embodiments, can rearrange and/or the element of the encoder apparatus 110 of constitutional diagram 2A in one or more.Described element can be implemented by hardware, software, firmware, middleware, microcode or its any combination.In certain embodiments, can rearrange and/or the element of the decoder device 150 of constitutional diagram 2B in one or more.Described element can be implemented by hardware, software, firmware, middleware, microcode or its any combination.

For instance, can use and be used to use FLO air interface specification (" only forward link [FLO] air interface specification (Forward Link Only[FLO] Air Interface Specification forTerrestrial Mobile Multimedia Multicast) that is used for land mobile multimedia multicast ", publish and be technical standard (Technical Standard) TIA-1099, in August, 2006, it is incorporated herein by reference fully to be used for all purposes) transmit the MediaFLO that real-time video is served in the TM3 system ^TMVideo coding is implemented some embodiment of the present invention.

Encoder apparatus 110 is carried out effective process, be used for by intraframe coding encoder 120 according in the frame H.264 _ 4x4 prediction standard determines whether that accuracy/quality is come the spatial prediction macro block fully.The process of being used by the intraframe coding encoder 120 of encoder apparatus 110 is used for the one or more codings that are provided to the bit stream image information that comprises a plurality of image subblocks of a plurality of spatial prediction mode that intra mode block predicts, wherein determines the spatial prediction mode of current sub-block based on estimated predictive mode.As determining, can significantly reduce misjudgment and estimate with the predictive mode that pin-point accuracy is provided from following description.

Basically, the intraframe coding encoder 120 of encoder apparatus 110 utilizes addition, subtraction and signed magnitude arithmetic(al).Only need very small amount of division of integer (under request, can provide accurate computing counting).In different embodiment, accurate computing counting will change, yet it is convenient to calculate.In addition, need not classify/subregion.As finding out from following description, implementation 900 is with the predictive mode of a set of estimator piece by stages.Phase I uses first subset of pixels with first grid arrangement that the estimation very accurately of first subclass of predictive mode is provided.Second stage use have from first grid be furnished with angle ground forward second subset of pixels of second grid arrangement of skew provide predictive mode second subclass estimate that very accurately second subclass of predictive mode is different with first subclass of predictive mode.Phase III adopts first subset of pixels with first grid arrangement and second subset of pixels with second grid arrangement to come the three subsetss of interpolation prediction pattern.

In foundation is used to be applied to the frame that counts on one's card of macro block _ estimator/tolerance of 4x4 prediction, by process 900 provider tropism's clues that the intraframe coding encoder 120 of code device 110 is carried out, it can be used for advantageously quickening in the final frame after a while _ the determining of 4x4 predictive mode (if the decision of macro block mode decision logic is used in the frame _ the 4x4 coding at the luminance channel of macro block).

The method (efficiency enabler) that realizes of the purpose of described embodiment and true efficient is can be with the consistent direction structure that exists with the scale greater than 4 * 4 sub-pieces in the mode identification signal simply and fast in this article, but not drops into and 4 * 4 handle (take to get rid of the probability of homogeneity/in the blindness mode of the fairly large structure that exists) at once.Therefore, if 8 * 8 sub-pieces are used for 4 * 4 intra prediction modes, then in the overlapping region identical, will be used for 44 * 4 sub-pieces through the same pattern that identification is used for described 8 * 8 sub-pieces with described 8 * 8 sub-pieces.

In one aspect, at this point and opinion,, 8 * 8 sub-pieces can be divided size and be chosen as and reduce computational complexity and still keep rationally trading off between the high frequency of occurrences for sub-piece with the consistent direction structure that exists or smoothness.Note, increase the probability (according to statistics) that sub-block size (exceeding 8 * 8) will reduce to run into the subimage of direction structure with homogeneous or smoothness, because the more details of signal will be captured by these big zones.Therefore, if identification and when consistent direction structure that exists when being identified in 8 * 8 scales under or smoothness can suppose that then it is applicable to existingly in that 8 * 8 district own (four) 4 * 4 sub-pieces.

Secondary analysis by subsequently and utilize the result of hereinafter described process 900, encoder apparatus 110 also can be evaluated at and use on the luminance channel of same macro block in the frame _ the rate-distortion advantage of 16x16 coding.To be similar in the frame at luminance channel _ mode of 16x16 spatial prediction mode assessment, can estimate in the frame of chrominance channe _ the calling the shot of 8x8 pattern.

Rate-distortion (R-D) is analyzed the framework that is provided for selecting in the best way macro-block coding pattern for encoder apparatus 110.Encoder apparatus 110 will be attempted to reduce to represent the bit rate that visual information is required by the distortion of lossy compression method algorithm introducing with its compressed format.The two is the target that conflict is arranged, and rate-distortion (R-D) framework defines compound cost metric, and described compound cost metric is captured the distortion dimension and the bit rate dimension of a described difficult problem, and it will be minimized by encoder apparatus 110.For most real-time coding apparatuses, the rough power method of R-D cost metric of macro-block coding pattern of assessing all permissions is too expensive on calculating.Therefore, make it possible to effectively to determine that the algorithm that the complexity of forced coding pattern to be utilized reduces is priceless.

H.264 standard provides the infra-frame prediction of two types for the luminance channel sub-block of pixels.The first kind is _ the 4x4 prediction that it supports the spatial prediction of 4 * 4 big boy's pieces under nine (9) individual different directions patterns in the frame.This type preferably is suitable for representing the image area of a large amount of details under on a small scale.Second type is _ the 16x16 prediction that it supports the spatial prediction of 16 * 16 big boy's pieces (that is whole macro block) under 4 different directions patterns in the frame.This type is more suitable for being used for representing the image area of smoothness or the even horizontal or vertical details that exists under 16 * 16 scales.Compare with luminance channel, both show quite low spatial detail (variation) usually chrominance channe U (Cb) and V (Cr).This is the common basic reason of using the 4:2:0 colourity subsample form in the video compression standard, whereby, the spatial resolution of original chrominance information before the compression level with vertical on all reduce by half.

Based on this relative smoothness that chrominance channe is compared with luminance channel, for the chrominance channe block of pixels, H.264 standard only provides one type infra-frame prediction.In defining the frame that is used for the chrominance channe block of pixels _ and 8x8 prediction supports the spatial prediction of the big boy's piece of 8x8 under 4 different directions patterns, described 4 different directions mode class are similar in the frame that is used for luminance channel _ and 4 direction modes of 16x16 type of prediction.It should be noted that, intra-macroblock does not exist only in the I type coding frame, but when the time, the motion compensated prediction process can not be discerned the reference of the satisfaction that is used for interframe encode, encoder apparatus 110 can determine to utilize the intra-macroblock of P type and category-B type coded frame.Therefore, for the decision-making of the macro block mode under all frame type of codings, the predictability assessment becomes important problem in the frame.When so that directly rough power mode is implemented, in the frame that is present in (16 * 16 pixel values) luminance macroblock in I, P or the category-B matrix section _ estimation that the assessment of 4x4 coding need produce 16 * 9=144 4 * 4 prediction signal, its residue signal that causes and the compound R-D cost of gained.This is generally the computation burden of the arduousness on the real-time coding apparatus.

The most important thing is that encoder apparatus 110 can detect, classifies and report the directional characteristic in sub-macro block district exactly, yet, can not keep and utilize the directional attributes of basis signal based on histogrammic method.Might be by synthetic two image section of identical or very similar histogram, one of them section represents direction structure (and therefore having appropriate space predictability) clearly, and another section is more as the texture that does not have clear directivity or smoothness, and the infra-frame prediction that therefore is not easy to carry out success.

Infra-frame prediction

Fig. 3 A explanation according to some embodiment H.264 in frame in _ 4x4 predictive mode 200.H.264 (4 * 4) prediction utilizes in 4 * 4 and along the direction predictability attribute by the basis signal of 8 directivity classifications of the vectorial 0-8 explanation among Fig. 3 A in the frame in.For the smooth signal that does not have clear directivity, supplying mode { 2} (DC).In Fig. 3 A, select height pattern based on vertical vector 0, wherein vector 0 is corresponding to pattern { 0}.Select the horizontal directive tendency (H.D.T.) based on horizontal vector 1, wherein vector 1 is corresponding to pattern { 1}.In described example, vector 7 and 5 has substantially from the directivity of 22.5 ° of vector 0 skews, and corresponding to pattern { 7 and 5}.Vector 3 (pattern 3}) have from the directivity of 45 ° of vector 0 skews.Equally, vector 4 (pattern { 4}) has from the directivity of 45 ° of vector 0 (pattern { 0}) or vector 1 (pattern { 1}) skews.Vector 8 (pattern 8}) have from the directivity of 22.5 ° of horizontal vector 1 skews.

Fig. 3 B explanation is according to the directivity classification 210 of the grouping of some embodiment.The directivity classification 210 of the grouping among Fig. 3 B is directly corresponding to above according to the { 0-8}, and will not further describe about the angular relationship between each pattern and each vector of the pattern among Fig. 3 A of standard H.264.Yet, in one aspect in, according to some configuration described herein, eight (8) individual directivity classifications 210 are oriented four (4) individual groups that are used for infra-frame prediction.These four groups comprise: (1) have vector 0, the rectangle group of 1}; (2) have vector 3, the diagonal group of 4}; (3) have 5,6,7, the inclination group of 8}; And (4) are at point { level and smooth group of 2} place.The vector of inclination group 5,6,7 and 8} have dotted line.Vector 5 is from 22.5 ° of vector 0 and 4 skews; From 45 ° of vector 6 and 7 skews; And from 90 ° of vector 8 skews, vector 6 is from 22.5 ° of vector 1 and 4 skews; Be offset 45 ° from vector 8 and vector 5; And from 90 ° of vector 7 skews.The vector of rectangle group 0,1} is offset 90 ° each other forward; And the vector 0,1} from the vector 4 the skew 45 °.Vector 0 is from 45 ° of vector 3 and 4 skews.

The partial view of (I) frame 300 in the frame of Fig. 4 explanation according to some embodiment.In one aspect, frame 300 is divided into the sub-pieces 310 in a plurality of borders (only displaying).The sub-piece 310 in border is for being positioned at the left hand edge or borderline 8 * 8 sub-pieces of I frame 300.Described 8 * 8 sub-pieces have 64 squares of its composition pixel in 8 * 8 frames of expression, select first subset of pixels from described 8 * 8 frames.In one aspect, select 16 pixels to be used for first subset of pixels.Sub-piece 310 in border and uneven neighbor with top and/or left side can permit described neighbor and be used for the prediction direction predictive mode, because the I frame end is in the periphery edge or the boundary of I frame 300.

Inner sub-piece 320 is 9 * 9 sub-pieces, and it comprises and is expressed as no any foursquare master's 8 * 8 sub-pieces 325 in main frame.External frame comprises described main frame and is listed as square (that is pixel) on one of the left periphery edge of leading 8 * 8 sub-pieces 325 or border.Described row square expression left side neighbor on left periphery or border.Equally, external frame is included in the periphery edge of main 8 * 8 sub-pieces 325 or delegation's top square (that is pixel) of top, border.Described capable top square expression top neighbor.

Fig. 5 A explanation is according to first subset of pixels of the sub-piece 410A in the border of some embodiment.In one aspect, the pixel of described first subset of pixels is not to be selected from first (ultra-Left side) row of the sub-piece 410A in border and first (topmost) OK.The pixel of described first subset of pixels is through selecting to form rectangular grid 412A.Rectangular grid 412A is included in the pixel r that arranges in the row and column alternately ₁₁, r ₁₂, r ₁₃, r ₁₄, r ₂₁, r ₂₂, r ₂₃, r ₂₄, r ₃₁, r ₃₂, r ₃₃, r ₃₄, r ₄₁, r ₄₂, r ₄₃And r ₄₄For instance, in second row of 8 * 8 sub-piece 410A, aim at r ₁₁, r ₁₂, r ₁₃And r ₁₄, between each selected pixel, have one and do not select pixel separation.Laying out pixel r in fourth line also ₂₁, r ₂₂, r ₂₃And r ₂₄, between each selected pixel, have one and do not select pixel.Equally, also the 6th the row in laying out pixel r ₃₁, r ₃₂, r ₃₃And r ₃₄, between each selected pixel, have one and do not select pixel.Also the 8th the row in laying out pixel r ₄₁, r ₄₂, r ₄₃And r ₄₄, between each selected pixel, have one and do not select pixel.

As seeing easily, the selected pixel of each row is shared row, makes that having one of selected pixel is listed as the row that the next-door neighbour only has the pixel do not selected.In addition, those row with selected pixel are separated by the delegation that does not only select pixel.In this configuration, selected pixel does not have side or angular contact or is close to another selected pixel in the described subclass.As describing in detail after a while, the rectangular grid 412A of selected pixel is used for predicting exactly corresponding to vector { 0} and the { directivity of 1} at first party tropism forecast period (stage 1 of Fig. 7).

Fig. 5 B explanation is according to second subset of pixels of the sub-piece 410B in the border of a certain embodiment.In one aspect, the pixel of described second subset of pixels is not to be selected from last (rightmost side) row of the sub-piece 410B in border and last (foot) OK.The pixel of described second subset of pixels is through selecting to form diagonal grid 412B.Diagonal grid 412B is included in the pixel d that arranges in the row of diagonal aligning ₁₁, d ₁₂, d ₁₃, d ₁₄, d ₂₁, d ₂₂, d ₂₃, d ₂₄, d ₃₁, d ₃₂, d ₃₃, d ₃₄, d ₄₁, d ₄₂, d ₄₃And d ₄₄For instance, make d ₁₁, d ₁₂, d ₁₃And d ₁₄In the 4th diagonal of 8 * 8 sub-piece 410A is capable, aim at.The 4th diagonal is capable to have four squares (that is pixel).Pixel d ₁₁Upper right corner contact pixel d ₁₂The lower left corner.Pixel d ₁₂Upper right corner contact pixel d ₁₃The lower left corner.Equally, pixel d ₁₃Upper right corner contact pixel d ₁₄The lower left corner.Therefore, in this configuration, pixel d only ₁₁Be to be selected from first (ultra-Left side) row, and d ₁₄It is only selected pixel in first (topmost) row.

Next-door neighbour's diagonal is capable only to have a pixel of choosing.Ensuing diagonal is capable to have pixel d ₂₁, d ₂₂, d ₂₃And d ₂₄, it is through layout and with above about d ₁₁, d ₁₂, d ₁₃, d ₁₄Described mode contact angle.In addition, pixel d ₂₁, d ₂₂, d ₂₃And d ₂₄The upper left corner contact pixel d respectively ₁₁, d ₁₂, d ₁₃And d ₁₄The lower right corner.Adjacent pixels d ₂₁, d ₂₂, d ₂₃And d ₂₄Diagonal capablely only have a pixel of choosing.Ensuing diagonal is capable to have to be similar to as about pixel d ₂₁, d ₂₂, d ₂₃And d ₂₄The pixel d that described mode is arranged ₃₁, d ₃₂, d ₃₃And d ₃₄Vicinity has pixel d ₃₁, d ₃₂, d ₃₃And d ₃₄The capable ensuing diagonal of diagonal capablely only have a pixel of choosing.The capable pixel d that comprises of ensuing diagonal ₄₁, d ₄₂, d ₄₃And d ₄₄, except right corner contact is not selected the pixel, it is to be similar to as about pixel d ₂₁, d ₂₂, d ₂₃And d ₂₄Described mode is arranged.As finding out easily, the selected pixel in the diagonal has 2,3 or 4 angles at the angle of the contiguous selected pixel of contact.As describing in detail after a while, the diagonal grid 412B of selected pixel is used for predicting exactly corresponding to vector { 3} and the { directivity of 4} at second party tropism forecast period (stage 2 of Fig. 7).

Fig. 6 A explanation is according to first subset of pixels of the sub-piece 520A in the inside of some embodiment, and its middle and upper part and left side neighbor are all available, and show the second rectangular grid 522A.In due course, the second rectangular grid 522A of the rectangular grid 412A of Fig. 5 A and Fig. 6 A each seed block of being used for I frame 300 by encoder apparatus 110 is to strengthen the accuracy of estimating directivity according to available left side and top neighbor.

In one aspect, some in the pixel of first subset of pixels are first (ultra-Left side) row and first (topmost) row that are selected from inner sub-piece 520A.But when left side and top neighbor time spent, the pixel of first subset of pixels is through selecting to form the second rectangular grid 522A.The second rectangular grid 522A comprises and is arranged in the pixel r of first topmost in capable ₁₁, r ₁₂, r ₁₃And r ₁₄In one aspect, at pixel r ₁₁With r ₁₂Between exist one not select pixel.At pixel r ₁₂With r ₁₃Between and at pixel r ₁₃With r ₁₄Between exist two not select pixel.Among the sub-piece 520A first (ultra-Left side) row have selected pixel r ₁₁, r ₂₁, r ₃₁And r ₄₁In one aspect, at pixel r ₁₁With r ₂₁Between exist one not select pixel.At pixel r ₂₁With r ₃₁Between and at pixel r ₃₁With r ₄₁Between exist two not select pixel.

Has selected pixel r ₁₁, r ₁₂, r ₁₃And r ₁₄Row and selected pixel r ₂₁, r ₂₂, r ₂₃And r ₂₄Row between have the delegation only do not select pixel.Has selected pixel r ₂₁, r ₂₂, r ₂₃And r ₂₄Row and selected pixel r ₃₁, r ₃₂, r ₃₃And r ₃₄Row between have two row only do not select pixel.Equally, has selected pixel r ₃₁, r ₃₂, r ₃₃And r ₃₄Row and selected pixel r ₄₁, r ₄₂, r ₄₃And r ₄₄Row between have two row only do not select pixel.

Row pixel r ₂₁, r ₂₂, r ₂₃And r ₂₄With pixel r ₁₁, r ₁₂, r ₁₃And r ₁₄Aim at.Equally, pixel r ₃₁, r ₃₂, r ₃₃And r ₃₄With pixel r ₄₁, r ₄₂, r ₄₃And r ₄₄Also with pixel r ₁₁, r ₁₂, r ₁₃And r ₁₄Aim at.As seeing easily, the selected pixel of each row is shared row, makes the row next-door neighbour with selected pixel only not select one or more row of pixel.In addition, those row with selected pixel are separated by one or more row that only do not select pixel.In this configuration, sub-piece 520A does not have border or edge not to have at least one selected pixel.In this configuration, selected pixel does not have side or angular contact or is close to another selected pixel in the described subclass.As describing in detail after a while, the second rectangular grid 522B of selected pixel is used for predicting exactly corresponding to vector { 0} and the { directivity of 1} in the stage 1 (Fig. 7).

Fig. 6 B explanation is according to second subset of pixels of the sub-piece 520B in the inside of some embodiment, and its middle and upper part and left side neighbor are all available, and shows diagonal grid 522B.Inner sub-piece 520B is 9 * 9 sub-pieces.Diagonal grid 522B is included in the pixel d that arranges in the row of diagonal aligning ₁₁, d ₁₂, d ₁₃, d ₁₄, d ₂₁, d ₂₂, d ₂₃, d ₂₄, d ₃₁, d ₃₂, d ₃₃, d ₃₄, d ₄₁, d ₄₂, d ₄₃And d ₄₄

Diagonal grid 522B have the 5th diagonal at sub-piece 520B capable in the pixel d that aims at of diagonal ₁₁, d ₁₂, d ₁₃, d ₁₄Pixel d ₁₁Be in fifth line, first (ultra-Left side) row of sub-piece 520B.Diagonal grid 522B has the 5th row that are in sub-piece 520B, the pixel d in first (topmost) row ₁₄Pixel d ₁₂Contact pixel d ₁₁The upper right corner, and pixel d ₂₁Contact pixel d ₁₁The lower right corner.Pixel d ₁₃Contact pixel d ₁₄The lower left corner, and pixel d ₂₄Contact pixel d ₁₄The lower right corner.Pixel d ₂₂Contact pixel d ₁₂The lower right corner and d ₂₁The upper right corner.In one aspect, in the 5th diagonal is capable at d ₁₂With d ₁₃Between exist one not select pixel.

Existing between the 5th and the 7th diagonal is capable does not only select a diagonal of pixel capable.The 7th diagonal is capable to have pixel d ₂₁, d ₂₂, d ₂₃And d ₂₄, wherein at pixel d ₂₂With d ₂₃Between exist one not select pixel.At pixel d ₂₁, d ₂₂, d ₂₃And d ₂₄With pixel d ₃₁, d ₃₂, d ₃₃And d ₃₄Between exist one or more diagonal capable.

Diagonal grid 522B has the pixel d that is in the 5th row of sub-piece 520B, last (foot) row ₄₁ Diagonal grid 522B has the pixel d in the fifth line that is in sub-piece 520B, last (rightmost side) row ₄₄Pixel d ₃₂Contact pixel d ₃₁The upper right corner, and pixel d ₄₁Contact pixel d ₃₁The lower right corner.Pixel d ₃₃Contact pixel d ₃₄The lower left corner, and pixel d ₄₄Contact pixel d ₃₄The lower right corner.Pixel d ₄₂Contact pixel d ₃₂The lower right corner and d ₄₁The upper right corner.In one aspect, at pixel d ₃₂With d ₃₃Between and similarly at pixel d ₄₂With d ₄₃Between on diagonal, have one and do not select pixel.

Fig. 6 C explanation is according to the alternative arrangements of second subset of pixels of the sub-piece 620 in the inside of some embodiment, and its middle and upper part and left side neighbor are all available, and show another diagonal grid configuration.In one aspect, last (rightmost side) row that can sub-internally piece 620 and last (foot) row are selected one or more in the pixel of second subset of pixels.The pixel of second subset of pixels is through selecting to form diagonal grid 622.Described diagonal grid 622 is included in the pixel d that arranges in the row of diagonal aligning ₁₁, d ₁₂, d ₁₃, d ₁₄, d ₂₁, d ₂₂, d ₂₃, d ₂₄, d ₃₁, d ₃₂, d ₃₃, d ₃₄, d ₄₁, d ₄₂, d ₄₃And d ₄₄For instance, in the 4th diagonal of 9 * 9 sub-pieces 620 is capable, aim at d ₁₁, d ₁₂, d ₁₃And d ₁₄The 4th diagonal is capable to have four squares (that is pixel).Pixel d ₁₁Upper right corner contact pixel d ₁₂The lower left corner.Pixel d ₁₂Upper right corner contact pixel d ₁₃The lower left corner.Equally, pixel d ₁₃Upper right corner contact pixel d ₁₄The lower left corner.Therefore, in this configuration, pixel d only ₁₁Be to be selected from first (ultra-Left side) row, it starts from fourth line, and d ₁₄It is only selected pixel in first (topmost) row.

Exist at least one next-door neighbour's who only has the pixel do not selected diagonal capable.In this regard, has pixel d ₂₁, d ₂₂, d ₂₃And d ₂₄With pixel d ₁₁, d ₁₂, d ₁₃And d ₁₄Two diagonal exist three diagonal capable between capable.In addition, pixel d ₂₁, d ₂₂, d ₂₃And d ₂₄The upper left corner respectively at pixel d ₁₁, d ₁₂, d ₁₃And d ₁₄The lower right corner between have one and do not select pixel.Adjacent pixels d ₂₁, d ₂₂, d ₂₃And d ₂₄Next diagonal capablely only have a pixel of choosing.Next diagonal is capable to have pixel d ₃₁, d ₃₂, d ₃₃, d ₃₄, it is to be similar to as about pixel d ₂₁, d ₂₂, d ₂₃And d ₂₄Described mode is arranged.Be adjacent to and have pixel d ₃₁, d ₃₂, d ₃₃, d ₃₄Capable one or more the ensuing diagonal of diagonal capable have only do not select pixel, then for having pixel d ₄₁, d ₄₂, d ₄₃, d ₄₄Diagonal capable.As seeing selected pixel d easily ₂₁, d ₂₂, d ₂₃And d ₂₄Respectively have and contact pixel d respectively ₃₁, d ₃₂, d ₃₃, d ₃₄The lower right corner.As describing in detail after a while, the diagonal grid 622 of selected pixel is used for predicting exactly corresponding to vector { 3} and the { directivity of 4} in the stage 2 (Fig. 7).As seeing easily, second subset of pixels that is arranged in first subset of pixels in rectangle (sampling) grid and is arranged in diagonal (sampling) grid of Fig. 5 A and Fig. 5 B or Fig. 6 A and Fig. 6 B (perhaps Fig. 6 C) forms inclination (sampling) grid arrangement together.

Fig. 7 explanation is according to the block diagram that is used for the intra prediction mode taxon 700 used by encoder apparatus 110 of some embodiment.At stage 1 place, intra prediction mode taxon 700 comprises the first subclass generator 705, and it produces the suitable rectangular grid 412A or the 522A (Fig. 5 A or Fig. 6 A) of the pixel of the corresponding current sub-block in the I frame 300.By using the rectangular grid 412A of border 8 * 8 sub-pieces, the first subclass generator 705 produces first subset of pixels, for example, is used for 16 pixels of those sub-pieces in border.In addition, by using the rectangular grid 522A of 9 * 9 sub-pieces in the frame, the first subclass generator 705 produces first subset of pixels, for example, is used for 16 pixels of those inner sub-pieces.

Intra prediction mode taxon 700 further comprises conversion module 710, and it has the conversion of Fourier (Fourier) shape, for example (but being not limited to) Hadamard (Hadamard) conversion, and it has 16 coefficients, as after a while with more detailed description.The output of the first subclass generator 705 (matrix) uses the lobate conversion of Fourier of conversion module 710 to stand two dimension (2-D) map function by transformation operator (being preceding and back matrix multiplication in one aspect) 715.The output of transformation operator 715 is sent to first intra prediction direction classification (IPDC) module 720.The one IPDC module 720 produces at least one output Out 1 and Out 2.If an IPDC module 720 assessment is from the result of transformation operator 715 and determine for { the 0} or { 1} then can finish operation or processing to that current sub-block of the sub-piece of current assessment predicted vector.Pattern { 0} or { 1} is sent to Out 1, and can be used for producing and the relevant bitstream information of coding of being undertaken by encoder apparatus 110.

In case pattern information is determined/is estimated through last, then in fact encoder will carry out corresponding predicted operation, that is, produce, deduct prediction signal to determine residue signal, (forward) conversion (to the integer transform approximation of 4 * 4DCT) and quantized residual signal and to determine that therefore gained is to quantification index from the pixel value of original 4 * 4 sub-pieces corresponding to the determined direction in space prediction signal of (comprising the unspecified directions that is used for the DC prediction).Therefore, encoder will write the quantification index of intra prediction mode information and gained (all in the entropy coding mode) to bit stream.

In one aspect, if { 2} then also can finish the operation or the assessment of current sub-block through determining to pattern.In the stage 1, if pattern { 0} or { 1} or { 2} is determined, then for the probability very low (if not being almost 0) of this mistake of determining.

Yet, if the result of an IPDC module 720 assessment transformation operators 715 and deterministic model 0,1 and 2} in none person be used for the sub-piece of current assessment by (clear/clearly) prediction, then use second subset of pixels in the current sub-block in second stage (stage 2), { 3} is with { 4} estimates current sub-block at pattern.

At stages 2 place, intra prediction mode taxon 700 comprises the second subclass generator 745, and its generation is used for suitable diagonal grid 412B, the 522B or 622 (Fig. 5 B, Fig. 6 B or Fig. 6 C) of pixel of current sub-block of the correspondence of I frame 300.This current sub-block is the same sub-piece that is used to produce first subset of pixels.By using the diagonal grid 412B of border 8 * 8 sub-pieces, the second subclass generator 745 produces 16 pixels that (for example) is used for those sub-pieces in border.In addition, produce 16 pixels that (for example) is used for those inner sub-pieces by diagonal grid 522B or 622, the second subclass generators 745 that use 9 * 9 sub-pieces in the frame.

Intra prediction mode taxon 700 further comprises conversion module 750, and it has the lobate conversion of Fourier, for example (but being not limited to) Hadamard transform Hadamard transformation, and it has 16 coefficients, as after a while with more detailed description.Conversion module 750 can separate with conversion module 710 or be identical.The output of the second subclass generator 745 (matrix) is used the lobate conversion of Fourier of conversion module 750 and is stood two dimension (2-D) map function by transformation operator 755 (in one aspect, the matrix multiplication of before comprising and back).The output of transformation operator 755 is sent to the second intra prediction direction sort module 760.Second intra prediction direction classification (IPDC) module 760 produces one among two output Out 3 and the Out 4.If the assessment of the 2nd IPDC module 760 is from the result of transformation operator 755 and determine for { the 3} or { 4} (Fig. 3 A or Fig. 3 B) then can finish operation or processing to that current sub-block of the sub-piece of current assessment predicted vector.Sending mode { 3} or { 4} produces and the relevant bitstream information of coding of being undertaken by encoder apparatus 110 being used to.In one aspect, if { 2} is determined pattern, then also can finish the operation or the assessment of current sub-block.In the stage 2, if pattern { 3} or { 4} or { 2} is determined, then for the probability very low (if not being almost 0) of this mistake of determining.

In view of above description, can detecting pattern 3,4 or the operation of 2} after carry out detecting pattern { 0,1 or the operation of 2}.The estimation of these patterns is functions of the layout of subset of pixels.In one aspect, encoder apparatus can first-selectedly change detecting pattern { 0,1 or 2} and pattern { 3,4 or the order of the operation of 2} when making great efforts to reduce total computing time of coded macroblocks, this accord priority: promptly, move to the phase I, carry out to detect more likely the operation of those patterns of setting up based on the deterministic model of adjacent sub-blocks.

The one IPDC module 720 and the 2nd IPDC module 760 respectively be provided for interpolation from the vector of inclination group or pattern (vector among Fig. 3 B 5}, 6}, 7} and 8}) and in one clue.In the phase III (stage 3), be respectively the calculating of first power and summation (PCS) module 725 and the 2nd PCS module 765 after an IPDC module 720 and the 2nd IPDC module 760.In operation, the result of the first and second PCS modules 725 and 765 is sent to infra-frame prediction interpolation directivity classification (IPIDC) module 730 in stage 3.IPIDC module 730 will compare from the result of the result of a PCS module 725 and the 2nd PCS module 765 with the described result of interpolation with corresponding to be used for the stages 3 place select the pattern of current sub-block 5,6,7 or one the vector of 8} 5,6,7 or 8} in the result of one directivity classify.Vector 5,6,7 and 8 and vectorial 0,1,3 and 4 is 22.5 °.IPIDC module 730 produces output Out 6.

Hadamard transform Hadamard transformation

Hadamard transform Hadamard transformation (Hadamard transform Hadamard transformation is also referred to as Walsh-Hadamard (Walsh-Hadamard) conversion) is an example of the vague generalization classification of Fourier transform.Hadamard transform Hadamard transformation is used for many signal processing and data compression algorithm.In one aspect, conversion module 710 and 750 uses Hadamard transform Hadamard transformation.

Hadamard transform matrix only contains value+1 and-1, and it is all row (row) mutually orthogonal modes.In hadamard transform matrix, only value+1 and-1 existence cause the no multiplication of low-down complexity to be implemented and fast algorithm.Can be based on to comprise exponent number be 2 Hadamard matrix and exponent number for the simple recursion of clo alunite gram (Kronecker) product of the Hadamard matrix of (m-1) produces the hadamard transform matrix that exponent number is m (m 〉=3), it is equal to 2 ^(m-1)* 2 ^(m-1)Size.Clo alunite gram product is the computing to two matrixes of any size that causes block matrix.It is the special circumstances of tensor product.Clo alunite gram product should not obscured with the common matrix multiplication that is entirely nonidentity operation.

Making H is that exponent number is the Hadamard matrix of n (n 〉=1).Then, equation Eq (1)

$\left[\begin{array}{cc}H& H\\ H& -H\end{array}\right]---\mathrm{Eq}\left(1\right)$

The matrix of cutting apart be that exponent number is the Hadamard matrix of (n+1).Can repeatedly use this observation, and cause following serial matrix, its start from the minimum in equation Eq (2) the Hadamard matrix (exponent number be 1 and size be 1 * 1), (size is 2 * 2) the Hadamard matrix of second rank in equation Eq (3) and the 3rd rank (size is 4 * 4) the Hadamard matrix in equation Eq (4A) and Eq (4B):

H ₁＝[1] Eq(2)

$\mathrm{<figure-callout\; id="2"\; label="H"\; filenames="A2007800473640034H1.png"\; state="\{\{state\}\}">H</figure-callout>}2=\left[\begin{array}{cc}1& 1\\ 1& -1\end{array}\right];---\mathrm{Eq}\left(3\right)$

${\mathrm{<figure-callout\; id="3"\; label="H"\; filenames="A2007800473640034H1.png,A2007800473640039H1.png"\; state="\{\{state\}\}">H</figure-callout>}}_3=\left[\begin{array}{cc}\left[\begin{array}{cc}1& 1\\ 1& -1\end{array}\right]& \left[\begin{array}{cc}1& 1\\ 1& -1\end{array}\right]\\ \left[\begin{array}{cc}1& 1\\ 1& -1\end{array}\right]& \left[\begin{array}{cc}1& 1\\ 1& -1\end{array}\right]\end{array}\right];---\mathrm{Eq}\left(4A\right)$

$=\left[\begin{array}{cccc}1& 1& 1& 1\\ 1& -1& 1& -1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\end{array}\right].---\mathrm{Eq}\left(4B\right)$

Because all row (row) are quadrature all, so can at random decide rank to the row (row) of Hadamard matrix.Yet, exist three kinds of use in practice important decide rank, that is, decide rank naturally, decide rank and reverse decide rank continuously.The Hadamard matrix series that is produced by above recurrence is that nature is decided the rank form.

In one aspect, conversion module 710 and 750 uses based on being 2-D 4 * 4 Hadamard transform Hadamard transformations that the size of deciding the rank form continuously is 4 1-D Hadamard transform Hadamard transformation.This decides the number that rank are based on the sign modification in the row, and described change is similar to the frequency of the cosine basic function that increases discrete cosine transform (DCT) and decides rank.Therefore, in the process that increases frequency, decide rank transform data (that is the set of coefficient).

Therefore, provide H by equation Eq (5) ₃Decide continuously the rank form:

$H_3^S=\left[\begin{array}{cccc}1& 1& 1& 1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\\ 1& -1& 1& -1\end{array}\right],---\mathrm{Eq}\left(5\right)$

Wherein, from the top row to the bottom line, there is 0,1,2 and 3 sign modification respectively.More particularly, be displaced to last row by secondary series and respectively the 3rd row of Eq (4B) and the 4th row moved on to the 2nd row and the 3rd and be listed as (it is corresponding to the frequency exponent number of increase) and produce equation Eq (5) the matrix among the equation Eq (4B).

Fig. 8 illustrates the flow chart of the calculating of the first intra prediction direction sort module.To send to an IPDC module 720 by the 2-D conversion that the Hadamard transform Hadamard transformation of equation Eq (5) is carried out from first subset of pixels of the first subclass generator 705.The examples show of the conversion output matrix of gained is for comprising block matrix 722A, 722B, 722C and 722D.In general, if current sub-block has height pattern, then will produce exemplary matrix 722A.Matrix 722A has the coefficient of the top row that is non-zero (NZ) basically.Remaining coefficient is or almost is zero.Therefore, when evaluating matrix 722A, an IPDC module 720 will be observed the pattern of NZ coefficient easily to be used for accurately assessment corresponding to the pattern { 0} of the vertical vector 0 of Fig. 3 B.

Matrix 722B has first (leftmost side) the row coefficient that is non-zero (NZ) basically.Remaining coefficient is or almost is zero.Therefore, when evaluating matrix 722B, an IPDC module 720 will be observed the pattern of NZ coefficient easily to be used for accurately estimation corresponding to the pattern { 1} of the horizontal vector 1 of Fig. 3 B.Matrix 722C has the upper left corner coefficient for non-zero (NZ).Remaining coefficient is or almost is zero.Therefore, when evaluating matrix 722C, an IPDC module 720 will be observed NZ coefficient in the upper left corner easily to estimate the pattern { 2} corresponding to the DC among Fig. 3 B.{ 0} is with { 1} separates 90 ° to pattern.

Matrix 722D have a plurality of for the coefficient of non-zero (NZ) and a plurality of be zero coefficient.In addition, described a plurality of NZ coefficient and described a plurality of value are the position of zero coefficient and are engaged in the structure of discerning among matrix 722A, 722B and the 722C any one.Therefore, when evaluating matrix 722D, an IPDC module 720 will be observed described pattern easily can not be used to the pattern of making { 3,4,5,6,7 and the accurate estimation of 8}.In fact, matrix 722D is provided for the clue of use in the stage 3 (Fig. 7).

Fig. 9 illustrates the flow chart of the calculating of the 2nd IPDC module 760.To send to the 2nd IPDC module 760 by the 2-D conversion that the Hadamard transform Hadamard transformation of equation Eq (5) is carried out from second subset of pixels of the second subclass generator 745.The examples show of the conversion output matrix of gained is for comprising block matrix 762A, 762B, 762C and 762D.In general, if current sub-block has the directivity according to vector 4, then will produce exemplary matrix 762A.Matrix 762A has the coefficient of the top row that is non-zero (NZ) basically.Remaining coefficient is or almost is zero.Therefore, when evaluating matrix 762A, the 2nd IPDC module 720 will be observed the pattern of NZ coefficient easily to be used for accurately estimation corresponding to the pattern { 4} of the vector 4 of Fig. 3 B.The diagonal grid provides the pixel that will produce as the similar 2-D conversion output pattern among Fig. 8 (except the vectorial directivity owing to the diagonal grid of gained is offset) basically.In the case, apart from vertical vector 0 for+45 ° and be identified as vertical and levels respectively for those vectors of-135 ° are existing apart from horizontal vector 1., apply a deviation angle, the feasible vector that can obtain corresponding to 3 or 4 thereafter.{ 3} is with { 4} separates 90 ° to pattern.

Matrix 762C has the upper left corner coefficient for non-zero (NZ).Remaining coefficient is or almost is zero.Therefore, when evaluating matrix 762C, the 2nd IPDC module 760 will be observed NZ coefficient in the upper left corner easily to estimate the pattern { 2} corresponding to the DC among Fig. 3 B.

Matrix 762D have a plurality of for the coefficient of non-zero (NZ) and a plurality of be zero coefficient.In addition, described a plurality of NZ coefficient and described a plurality of value are that the position of zero coefficient is not engaged in any one in the structure of discerning among matrix 762A, 762B and the 762C.Therefore, when evaluating matrix 762D, the 2nd IPDC module 760 will be observed described pattern easily can not be used to the pattern of making { 5,6,7 and the accurate estimation of 8}.As an alternative, matrix 762D can be used as the clue of the described pattern of interpolation in the inclination group.

Figure 10 illustrates that first power calculates and summation module 725.Matrix 800 is corresponding to matrix 722D and be the 2-D conversion output matrix of gained.To take advantage of the matrix of n matrix and subset of pixels be that n takes advantage of the n matrix if the matrix of Hadamard transform Hadamard transformation is n, and then the 2-D conversion is taken advantage of the preceding and back multiplication of hadamard transform matrix via the subset of pixels matrix and produced n and takes advantage of the n matrix.In one aspect, employed hadamard transform matrix is 4 * 4 matrixes.16 pixels in the subset of pixels also are suitable for being arranged to 4 * 4 matrixes.Therefore, the 2-D conversion output matrix of gained is 4 * 4 matrixes.Yet, can adopt other alternative types, transformation matrix size and corresponding subset of pixels matrix size.

In a PCS module 725, each coefficient in the delegation is through square to derive a power value.Then, for each row, to square result's summation of these coefficients.Equal sign group 812 is illustrative, and for row 802, the end value of summation coefficient equals and R1 through producing and being set at expression.Equally, the coefficient in the remaining row (only showing 804,806 and 808) is sued for peace through quadratic sum, makes each row have final summing value at digital 816 places (with R2 and R3 ... and R4).

In one aspect, via the equal sign 810 of one group of summation to each row through square coefficient summation to produce at digital 814 places and (with C1 and C2 and C3 ... and C4).765 couples of 2-D conversion output matrix 762D (Fig. 9) from transformation operator 755 gained of the 2nd PCS module in stage 2 carry out identical operations.Therefore, for the 2nd PCS module 765, it is unnecessary further discussing.

Figure 11 is to the flow chart of Figure 13 explanation according to the process 900 of the estimation intra prediction direction sexual norm of some embodiment.In following various configurations, with the order flowchart square frame of being described, perhaps can be simultaneously, parallel or carry out these square frames or its part with different order.Process 900 starts from square frame 902, at square frame 902 places, obtains the I frame.Square frame 902 backs are square frame 904A and 904B.At square frame 904A place, obtain the sub-piece in border (that is sub-piece 310).Be square frame 906A behind the square frame 904A,, obtain first subset of pixels of the sub-piece in current border at square frame 906A place.Be square frame 908A behind the square frame 906A,, first subset of pixels of the sub-piece in current border used Hadamard transform Hadamard transformation (that is 2-D Hadamard transform Hadamard transformation) at square frame 908A place.Square frame 908A (with the square frame 908B that describes after a while) implements transformation operator 715, and it constitutes the preceding and back multiplication of subset of pixels matrix and hadamard transform matrix.Be square frame 910A behind the square frame 908A,, IPDC operation take place at square frame 910A place.Be square frame 912A behind the square frame 910A,, determine that current sub-block is whether through last classification or deterministic model { 0,1 or 2} not at square frame 912A place.If the described "Yes" that is defined as, then operation and the assessment to the sub-piece in current border finishes, and process 900 proceeds to square frame 950.If the "No" that is defined as at square frame 912A place, then process 900 proceeds to square frame 920A among Figure 12 with the sub-piece in the current border of further assessment.

Referring now to square frame 904B,,, obtains inner sub-piece (that is sub-piece 320) at square frame 904B place.Be square frame 906B behind the square frame 904B,, obtain first subset of pixels of the sub-piece in current inside at square frame 906B place.Be square frame 908B behind the square frame 906B,, first subset of pixels of the sub-piece in current inside used Hadamard transform Hadamard transformation (that is 2-D Hadamard transform Hadamard transformation) at square frame 908B place.Be square frame 910B behind the square frame 908B,, IPDC operation take place at square frame 910B place.Be square frame 912B behind the square frame 910B,, determine that described sub-piece is whether through last classification or deterministic model { 0,1 or 2} not at square frame 912B place.If the "Yes" that is defined as at square frame 912B place, then operation and the assessment to the sub-piece in current inside finishes, and process 900 proceeds to the square frame of describing after a while 950.

If the "No" that is defined as at square frame 912B place, then process 900 proceeds to square frame 920B among Figure 12 with the sub-piece in the current inside of further assessment.When needs, also make the gained 2-D conversion output matrix that sends to square frame 910A or square frame 910B can be used for square frame 940A.

For exemplary object, show square frame 904A, 906A, 908A, 910A and the 912A that is used for the sub-piece in current border abreast with square frame 904B, the 906B, 908B, 910B and the 912B that are used for the sub-piece in current inside.As understanding, for current sub-block, described sub-piece only is one in sub-piece in border or the inner sub-piece.Therefore, do not carry out described operation abreast, because it relates to arbitrary current sub-block, unless may carry out by processor with parallel processing ability.

If in the "No" that is defined as at square frame 912A or 912B place, then process 900 proceeds to square frame 920A or 920B respectively, it is that the sub-piece in border still is that inner sub-piece is decided on current sub-block.Therefore, for current sub-block, process 900 only need to be associated with the stage 1 those operate the subclass of predicting the pattern that has 90 ° of angles therebetween exactly.Therefore, process 900 reduces on calculating, and the prediction of pin-point accuracy is provided simultaneously.

At square frame 920A place, obtain second subset of pixels of the sub-piece in current border.In this stage, second subset of pixels is from diagonal grid 412B.Be square frame 922A behind the square frame 920A,, second subset of pixels of the sub-piece in current border used Hadamard transform Hadamard transformation (that is 2-D Hadamard transform Hadamard transformation) at square frame 922A place.Square frame 922A (with the square frame 922B that describes after a while) implements transformation operator 755, and it constitutes the preceding and back multiplication of the second subset of pixels matrix and hadamard transform matrix.Be square frame 924A behind the square frame 922A,, the 2nd IPDC operation take place at square frame 924A place.Be square frame 926A behind the square frame 924A,, determine that the sub-piece in current border is whether through last classification or deterministic model { 3,4 or 2} not at square frame 926A place.If the "Yes" that is defined as at square frame 926A place, then operation and the assessment to the sub-piece in current border finishes, and process 900 proceeds to the square frame of describing after a while 950.

If the "No" that is defined as at square frame 926A place, then process 900 proceeds to square frame 940A among Figure 13 and 940B with the sub-piece in current border in further evaluation stage 3.

At square frame 920B place, obtain second subset of pixels of the sub-piece in current inside.In this stage, second subset of pixels is from diagonal grid 522B or 622.Be square frame 922B behind the square frame 920B,, second subset of pixels of the sub-piece in current inside used Hadamard transform Hadamard transformation (that is 2-D Hadamard transform Hadamard transformation) at square frame 922B place.Be square frame 924B behind the square frame 922B,, the 2nd IPDC operation take place at square frame 924B place.Be square frame 926B behind the square frame 924B,, determine that the sub-piece in current inside is whether through last classification or deterministic model { 3,4 or 2} not at square frame 926B place.If the "Yes" that is defined as at square frame 926B place, then operation and the assessment to the sub-piece in current inside finishes, and process 900 proceeds to the square frame of describing after a while 950.

If the "No" that is defined as at square frame 926B place, then process 900 proceeds to square frame 940A among Figure 13 and 940B with the sub-piece in the current inside of further assessment.

For exemplary object, show square frame 920A, 922A, 924A and the 926A that is used for the sub-piece in border abreast with square frame 920B, the 922B, 924B and the 926B that are used for inner sub-piece.As understanding, for current sub-block, described sub-piece only is one in sub-piece in border or the inner sub-piece.Therefore, do not carry out these two groups operations abreast, because it relates to arbitrary current sub-block.Therefore, for current sub-block, process 900 only need to be associated with the stage 2 those operate second subclass of predicting the pattern that has 90 ° of angles therebetween exactly.

If square frame 926A or 926B place are defined as "No", then process 900 proceeds to square frame 940A and 940B.When needs, also make the gained 2-D conversion output matrix of the 924B that sends to the square frame 924A that is used for the sub-piece in current border or be used for the sub-piece in current inside can be used for square frame 940A and 940B.

Now referring to Figure 11 and Figure 13, the matrix from square frame 910A or 910B that will be associated with current sub-block is called " matrix 1 " in this article.Referring to Figure 12 and Figure 13, the matrix from square frame 924A or 924B that will be associated with same current sub-block is called " matrix 2 " in this article.Matrix 1 is showed among Figure 10, and by numeral 800 expressions.Matrix 1 is sent to first power calculate and summation module 725, calculate and summation module 725 places the power of each coefficient in the derivational matrix 1 at first power.At square frame 940A place,, described coefficient square derives described power by being asked.Then, at square frame 942A place, to the coefficient summation of the warp square in each row with the power that obtains 816 places in Figure 10 through summation capable (R1, R2, R3 ..., R4).In addition, at square frame 944A place, to the coefficient summation of the warp square in each row with the power row that obtain 814 places in Figure 10 through summation (C1, C2, C3 ..., C4).

Equally, the matrix from square frame 924A or 924B that will be associated with current sub-block is called " matrix 2 " in this article.The matrix that matrix 2 is similar among Figure 10 to be showed, and by numeral 800 expressions.Matrix 2 is sent to second power calculate and summation module 765, calculate and summation module 765 places the power of each coefficient in the derivational matrix 2 at second power.At square frame 940B place,, described coefficient square derives described power by being asked.Then, at square frame 942B place, capable to obtain through the power of summation to the coefficient summation of the warp square in each row.In addition, at square frame 944B place, to the coefficient summation of the warp square in each row to obtain power row through summation.

Square frame 944A and 944B back be a square frame 946, at square frame 946 places, in comparator matrix 1 and the matrix 2 with.Square frame 946 carry out in matrixes 1 and the matrix 2 and the power that presents by each matrix with identification of comparative analysis distribute and the degree of closeness that had before distributed by the characteristic power of corresponding matrix (722A, 722B, 722C) or (762B, 762A, 762C) mode designated (0,1,2) or pattern (3,4,2).In one aspect, if in matrix 1 and R1 more much bigger than (R2+R3+R4), perhaps comparably, if ratio R 1/ (R2+R3+R4) is greater than the threshold value of value 20.0, then square frame 946 can determine the predictive mode of current sub-block is categorized as " perpendicular " or is pattern 0.In general, square frame 946 from matrix 1 and matrix 2 both each and between carry out similarly relatively.Square frame 946 backs are square frame 948, and at square frame 948 places, the interpolation of execution result is to determine directivity and the predictive mode of pattern (5,6,7 or 8) is classified.Square frame 948 back is a square frame 950, at square frame 950 places, will be used in the frame from stage 1, stage 2 or stages 3 recognized patterns _ 4x4 predictive mode or be used for other predictive mode of current sub-block.Formerly accurately between the vector of prediction or in the pattern any one, vector 5,6,7 and 8 is 22.5 ° (perhaps half of angular distance).

Square frame 950 backs are square frame 952, at square frame 952 places, determine whether current sub-block is the last sub-piece of frame 300.If be defined as "Yes", then process 900 finishes.Yet if in the "No" that is defined as at square frame 952 places, process 900 turns back to square frame 904A or 904B to obtain next the height piece in the frame 300.

As finding out from the above description, process 900 is with the calculating effective and efficient manner and with predictability in the frame of high accuracy assessment macro block, and it is the important component part by the mode decision algorithm of encoder/transcoder execution.In conjunction with realizing above two targets (that is, computational efficiency and accuracy), process 900 also is provided for the descriptive guiding of tool of 8 * 8 sub-pieces, the input (clue) that its complexity that can be used as the final mode decision stage of determining final pattern reduces.In the final mode decision stage, encoder apparatus is jointly checked the rate-distortion advantage of all different macroblock encoding types that can allow the current macro that is used for just being encoded.For instance, for the macro block in the I type coding frame, admissible macroblock encoding type comprises in the frame at luminance channel _ 4x4 or frame in _ 16x16 and frame at chrominance channel in _ 8x8.For the macro block in P or the category-B type encoded picture, this tabulation of the macroblock coding option that can permit will be with various interframe encode options.

As seeing easily, first subset of pixels and second subset of pixels are two subset of pixels that do not comprise mutually, and it is identified in 8 * 8 sub-pieces of macro block in brightness (Y) passage.Explanation two pairs of these different a little collection (each is made up of 16 pixels) among Fig. 5 A and Fig. 5 B.Based on neighbor's availability, can utilize other subclass that for example is showed among Fig. 6 A and Fig. 6 B or Fig. 6 C.The subset of pixels of introducing is to defining two subsample grids using 4 * 42-D Hadamard transform Hadamard transformation.

Output place in the 2-D Hadamard transform Hadamard transformation, in the coefficient subclass that defines in advance, based on through square coefficient value and comparison (this is for needing the situation of division of integer), perhaps in one aspect in, based on the absolute coefficient value and comparison, transformation results (the directivity structure of Dui Ying grids of pixels perhaps equivalently) can be categorized as one in the following kind: perpendicular; Near vertical; Substantial horizontal; Near level; Smoothly; Other person at square frame 946 places (promptly; In the non-above kind one).

Based on this classification of two Hadamard transform Hadamard transformation results (from rectangular grid, and another person is from the diagonal grid), table 1 explanation can with respect to its in the frame _ controllability of 4x4 prediction and to 8 * 8 modes of classifying.In this table, final classification results " texture (Texture) " shows bad infra-frame prediction performance, and all other final digital sort results show good infra-frame prediction performance, and wherein digital value is described in the frame that cooperates most _ the 4x4 predictive mode.

For the complexity that makes process 900 remains lowly, the decision tree that can implement light-duty pruning is with the approximate classification of being implemented by table 1.

Table 1:8 * 8 block sorts and directivity decision-making

Via transformation results, for example vertical pattern of vertical edge will finally be disclosed, and the gained classification will be " perpendicular " about rectangular grid.In the analysis about the diagonal grid, this vertical pattern will be neither vertical configuration be also non-horizontal, and therefore, its classification will be " other ".As seeing easily from table 1, this sample situation is corresponding to last row in first row of table 1, and its most final classification mode is labeled as { 0}.As second example, the directional patterns that consideration is aimed at the vector 6 among Fig. 3 B will cause the estimation of " near level " classification about the transformation results of rectangular grid.As result about the conversion of diagonal grid, the estimation that the same slant pattern of aiming at the vector 6 among Fig. 3 B will cause " near vertical " to classify.As seeing easily from table 1, this second sample situation is corresponding to the 4th row and the 2nd row of table 1, and its most final classification mode is labeled as { 6}.

Those skilled in the art will appreciate that, can use in multiple different technologies and the skill any one to come expression information and signal.For instance, data, instruction, order, information, signal, position, symbol and chip that may reference in describing more than whole can be represented by voltage, electric current, electromagnetic wave, magnetic field or magnetic particle, light field or light particle or its any combination.

The those skilled in the art will further understand, and various illustrative components, blocks, module and the algorithm steps described in conjunction with example disclosed herein can be embodied as electronic hardware, firmware, computer software, middleware, microcode or its combination.For this interchangeability of hardware and software clearly is described, above substantially according to its functional descriptions various Illustrative components, piece, module, circuit and step.With this functional hardware that is embodied as still is that software is decided on application-specific and the design constraint of forcing on whole system.The those skilled in the art can implement described functional by different way for each application-specific, but these a little implementation decisions should not be interpreted as causing breaking away from the scope of the method that is disclosed.

Can use general processor, digital signal processor (DSP), application-specific integrated circuit (ASIC) (ASIC), field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components or its to implement or carry out various illustrative components, blocks, assembly, module and the circuit of describing in conjunction with example disclosed herein with any combination of carrying out function described herein through design.General processor can be microprocessor, but in replacement scheme, processor can be arbitrary conventional processors, controller, microcontroller or state machine.Processor also can be implemented as the combination of calculation element, for example, and the associating of the combination of DSP and microprocessor, the combination of a plurality of microprocessors, DSP core and one or more microprocessors or any other this type of configuration.

In one or more software modules that the method for describing in conjunction with example disclosed herein or the step of algorithm can directly be included in the hardware, carried out by one or more treatment elements, or in both combinations.Software module can reside in the combination of the medium of arbitrary other form known in RAM memory, flash memory, ROM memory, eprom memory, eeprom memory, register, hard disk, removable disk, CD-ROM or this technology or medium.The example medium can be coupled to processor, make processor to write medium from read information with information.In replacement scheme, medium can be incorporated into processor.Processor and medium can reside in the application-specific integrated circuit (ASIC) (ASIC).ASIC can reside in the radio modem.In replacement scheme, processor and medium can be used as discrete component and reside in the radio modem.

The those skilled in the art provides the previous description of revealed instance so that can make or use the method and apparatus that is disclosed.The those skilled in the art will understand the various modifications to these examples easily, and the principle that is defined herein can be applicable to other example, and can add additional element.

Claims (43)

1. device, it comprises: processor, it can be operated with first subset of pixels based on the current sub-block of frame in the frame and estimate the first spatial prediction mode subclass, and estimate the second spatial prediction mode subclass based on second subset of pixels of described current sub-block, the described first predictive mode subclass is different with the described second predictive mode subclass; And memory, it is coupled to described processor.

2. device according to claim 1, wherein said processor further can operate with based on described first subset of pixels with second subset of pixels and the interpolation three spatial prediction mode subclass different with the described second spatial prediction mode subclass with the described first spatial prediction mode subclass.

3. device according to claim 2, the wherein said first spatial prediction mode subclass are based on first pair of vector of 90 ° of displacements; The described second spatial prediction mode subclass is based on second pair of vector of 90 ° of displacements; And described the 3rd spatial prediction mode subclass is based on 22.5 ° a plurality of vectors of any one displacement from the described second pair of vector of described first pair of vector sum.

4. device according to claim 1, wherein said processor further can be operated to change the size of described current sub-block and the layout of described first subset of pixels and described second subset of pixels automatically based on the left side of described current sub-block and top neighbor's availability.

5. device according to claim 1, wherein said first subset of pixels is arranged in first grid arrangement, and the selected pixel of the vicinity in wherein said first subset of pixels does not contact; And described second subset of pixels is arranged in second grid arrangement, and wherein only two or more angles of Lin Jin selected pixel contact in pattern of diagonal lines.

6. device according to claim 1, wherein when estimating the described first spatial prediction mode subclass, described processor can operate with the map function of carrying out described first subset of pixels and Hadamard transform Hadamard transformation with derivation have with discern the described first spatial prediction mode subclass in the first gained matrix of the coefficient pattern that is associated of a spatial prediction mode; And when estimating the described second spatial prediction mode subclass, described processor can be operated the second gained matrix that has the described coefficient pattern that is associated with a spatial prediction mode in the described second identification space predictive mode subclass with the described map function of carrying out described second subset of pixels and described Hadamard transform Hadamard transformation with derivation.

7. device according to claim 1, wherein said first subset of pixels are arranged in the rectangle sampling grid arrangement; Described second subset of pixels is arranged in the diagonal sampling grid arrangement; And described rectangle sampling grid forms the sampling grid arrangement that tilts with described diagonal sampling grid arrangement.

8. device according to claim 1, the pixel of the pixel of wherein said first subset of pixels and described second subset of pixels does not comprise mutually.

9. device according to claim 1, the pixel portion of the pixel of wherein said first subset of pixels and described second subset of pixels is overlapping.

10. multimedia system, it comprises:

Encoder, it can be operated with first subset of pixels based on the current sub-block of frame in the frame and estimate the first spatial prediction mode subclass, and estimate the second spatial prediction mode subclass based on second subset of pixels of described current sub-block, the described first predictive mode subclass is different with the described second predictive mode subclass; And

Decoder, it can be operated based on described first spatial prediction mode subclass and the described second spatial prediction mode subclass frame in the described frame is decoded.

11. system according to claim 10, wherein said encoder further can operate with based on described first subset of pixels with second subset of pixels and the interpolation three spatial prediction mode subclass different with the described second spatial prediction mode subclass with the described first spatial prediction mode subclass.

12. system according to claim 11, the wherein said first spatial prediction mode subclass are based on first pair of vector of 90 ° of displacements; The described second spatial prediction mode subclass is based on second pair of vector of 90 ° of displacements; And described the 3rd spatial prediction mode subclass is based on 22.5 ° a plurality of vectors of any one displacement from the described second pair of vector of described first pair of vector sum.

13. system according to claim 10, wherein said encoder further can be operated to change the size of described current sub-block and the layout of described first subset of pixels and described second subset of pixels automatically based on the left side of described current sub-block and top neighbor's availability.

14. system according to claim 10, wherein said first subset of pixels is arranged in first grid arrangement, and the selected pixel of the vicinity in wherein said first subset of pixels does not contact; And described second subset of pixels is arranged in second grid arrangement, and wherein only two or more angles of Lin Jin selected pixel contact in pattern of diagonal lines.

15. system according to claim 10, wherein when estimating the described first spatial prediction mode subclass, described encoder can operate with the map function of carrying out described first subset of pixels and Hadamard transform Hadamard transformation with derivation have with discern the described first spatial prediction mode subclass in the first gained matrix of the coefficient pattern that is associated of a spatial prediction mode; And when estimating the described second spatial prediction mode subclass, described encoder can operate with the described map function of carrying out described second subset of pixels and described Hadamard transform Hadamard transformation with derivation have with discern the described second spatial prediction mode subclass in the second gained matrix of the described coefficient pattern that is associated of a spatial prediction mode.

16. system according to claim 10, wherein said first subset of pixels are arranged in the rectangle sampling grid arrangement; Described second subset of pixels is arranged in the diagonal sampling grid arrangement; And described rectangle sampling grid forms the sampling grid arrangement that tilts with described diagonal sampling grid arrangement.

17. system according to claim 10, the pixel of the pixel of wherein said first subset of pixels and described second subset of pixels does not comprise mutually.

18. system according to claim 10, the pixel portion of the pixel of wherein said first subset of pixels and described second subset of pixels is overlapping.

19. a multimedia system, it comprises:

Be used for coming the device of first subclass of estimation space predictive mode based on first subset of pixels of the current sub-block of frame in the frame;

Be used for estimating based on second subset of pixels of described current sub-block the device of the second spatial prediction mode subclass, the described first predictive mode subclass is different with the described second predictive mode subclass; And

Be used for the device of frame in the described frame being decoded based on the described first spatial prediction mode subclass and the described second spatial prediction mode subclass.

20. system according to claim 19, it further comprises and is used for based on described first subset of pixels with second subset of pixels and the device of the interpolation three spatial prediction mode subclass different with the described second spatial prediction mode subclass with the described first spatial prediction mode subclass.

21. system according to claim 20, the wherein said first spatial prediction mode subclass are based on first pair of vector of 90 ° of displacements; The described second spatial prediction mode subclass is based on second pair of vector of 90 ° of displacements; And described the 3rd spatial prediction mode subclass is based on 22.5 ° a plurality of vectors of any one displacement from the described second pair of vector of described first pair of vector sum.

22. system according to claim 19, it further comprises the device of the layout of the size that is used for changing automatically based on the left side of described current sub-block and top neighbor's availability described current sub-block and described first subset of pixels and described second subset of pixels.

23. system according to claim 19, the wherein said device that is used for estimating the described first spatial prediction mode subclass comprises and is used to carry out the map function of described first subset of pixels and Hadamard transform Hadamard transformation has the first gained matrix of the coefficient pattern that is associated with a spatial prediction mode of the described first spatial prediction mode subclass of identification with derivation device; The described device that is used for estimating the described second spatial prediction mode subclass can be operated the second gained matrix that has the described coefficient pattern that is associated with a spatial prediction mode of the described second spatial prediction mode subclass of identification with derivation with the described map function of carrying out described second subset of pixels and described Hadamard transform Hadamard transformation.

24. a computer program, it comprises the computer-readable media that comprises the instruction that is used to handle multi-medium data, and wherein said instruction causes computer:

First subset of pixels based on the current sub-block of frame in the frame is estimated the first spatial prediction mode subclass;

Estimate the second spatial prediction mode subclass based on second subset of pixels of described current sub-block, the described first predictive mode subclass described second is different with the predictive mode subclass; And

Based on described first spatial prediction mode subclass and the described second spatial prediction mode subclass frame in the described frame is encoded.

25. computer program according to claim 24, it further comprises the instruction that causes described computer to carry out following operation: based on described first subset of pixels with second subset of pixels and the interpolation three spatial prediction mode subclass different with the described second spatial prediction mode subclass with the described first spatial prediction mode subclass.

26. computer program according to claim 25, the wherein said first spatial prediction mode subclass are based on first pair of vector of 90 ° of displacements; The described second spatial prediction mode subclass is based on second pair of vector of 90 ° of displacements; And described the 3rd spatial prediction mode subclass is based on 22.5 ° a plurality of vectors of any one displacement from the described second pair of vector of described first pair of vector sum.

27. computer program according to claim 24, it further comprises the instruction that causes described computer to carry out following operation: change the size of described current sub-block and the layout of described first subset of pixels and described second subset of pixels automatically based on the left side of described current sub-block and top neighbor's availability.

28. computer program according to claim 24, the wherein said instruction that causes described computer to estimate the described first spatial prediction mode subclass comprise that the map function that causes described computer to carry out described first subset of pixels and Hadamard transform Hadamard transformation have the instruction of the first gained matrix of the coefficient pattern that is associated with a spatial prediction mode in the described first spatial prediction mode subclass of identification with derivation.

29. computer program according to claim 28, the wherein said instruction that causes described computer to estimate the described second spatial prediction mode subclass comprise that the described map function that causes described computer to carry out described second subset of pixels and described Hadamard transform Hadamard transformation have the instruction of the second gained matrix of the described coefficient pattern that is associated with a spatial prediction mode in the described second spatial prediction mode subclass of identification with derivation.

30. computer program according to claim 29, it further comprises the instruction that causes described computer to carry out following operation: based on the described first gained matrix with the described second gained matrix and the interpolation three spatial prediction mode subclass different with the described second spatial prediction mode subclass with the described first spatial prediction mode subclass.

31. an encoder device, it comprises:

Encode processor, its can operate with when carrying out intraframe coding based on frame in first subset of pixels of current sub-block of frame estimate the first spatial prediction mode subclass, and estimate the second spatial prediction mode subclass based on second subset of pixels of described current sub-block, the described first predictive mode subclass is different with the described second predictive mode subclass; And

Memory, it is coupled to described processor.

32. encoder device according to claim 31, wherein said encode processor further can operate with based on described first subset of pixels with second subset of pixels and the interpolation three spatial prediction mode subclass different with the described second spatial prediction mode subclass with the described first spatial prediction mode subclass.

33. encoder device according to claim 32, the wherein said first spatial prediction mode subclass are based on first pair of vector of 90 ° of displacements; The described second spatial prediction mode subclass is based on second pair of vector of 90 ° of displacements; And described the 3rd spatial prediction mode subclass is based on 22.5 ° a plurality of vectors of any one displacement from the described second pair of vector of described first pair of vector sum.

34. encoder device according to claim 31, wherein said encode processor further can be operated to change the size of described current sub-block and the layout of described first subset of pixels and described second subset of pixels automatically based on the left side of described current sub-block and top neighbor's availability.

35. encoder device according to claim 31, wherein said first subset of pixels is arranged in first grid arrangement, and the selected pixel of the vicinity in wherein said first subset of pixels does not contact; And described second subset of pixels is arranged in second grid arrangement, and wherein only two or more angles of Lin Jin selected pixel contact in pattern of diagonal lines.

36. encoder device according to claim 31, wherein when estimating the described first spatial prediction mode subclass, described encode processor can operate with the map function of carrying out described first subset of pixels and Hadamard transform Hadamard transformation with derivation have with discern the described first spatial prediction mode subclass in the first gained matrix of the coefficient pattern that is associated of a spatial prediction mode; And when estimating the described second spatial prediction mode subclass, described processor can operate with the described map function of carrying out described second subset of pixels and described Hadamard transform Hadamard transformation with derivation have with discern the described second spatial prediction mode subclass in the second gained matrix of the described coefficient pattern that is associated of a spatial prediction mode.

37. encoder device according to claim 36, wherein said encode processor further can operate with based on the described first gained matrix with the described second gained matrix and the interpolation three spatial prediction mode subclass different with the described second spatial prediction mode subclass with the described first spatial prediction mode subclass.

38. a method, it comprises:

First subset of pixels based on the current sub-block of frame in the frame is estimated the first spatial prediction mode subclass;

Estimate the second spatial prediction mode subclass based on second subset of pixels of described current sub-block, the described first predictive mode subclass is different with the described second predictive mode subclass; And

Based on described first spatial prediction mode subclass and the described second spatial prediction mode subclass frame in the described frame is encoded.

39. according to the described method of claim 38, it further comprises: based on described first subset of pixels with second subset of pixels and the interpolation three spatial prediction mode subclass different with the described second spatial prediction mode subclass with the described first spatial prediction mode subclass.

40. according to the described method of claim 38, it further comprises: change the size of described current sub-block and the layout of described first subset of pixels and described second subset of pixels automatically based on the left side of described current sub-block and top neighbor's availability.

41. according to the described method of claim 38, the described first spatial prediction mode subclass of wherein said estimation comprises: carry out the map function and the Hadamard transform Hadamard transformation of described first subset of pixels; And described method further comprises: derive have with discern the described first spatial prediction mode subclass in the first gained matrix of the coefficient pattern that is associated of a spatial prediction mode.

42. according to the described method of claim 41, the described second spatial prediction mode subclass of wherein said estimation comprises: carry out the described map function and the described Hadamard transform Hadamard transformation of described second subset of pixels; And described method further comprises: derive have with discern the described second spatial prediction mode subclass in the second gained matrix of the described coefficient pattern that is associated of a spatial prediction mode.

43. according to the method for claim 42, it further comprises: based on the described first gained matrix with the described second gained matrix and the interpolation three spatial prediction mode subclass different with the described second spatial prediction mode subclass with the described first spatial prediction mode subclass.

Images