US20020122491A1 - Video decoder architecture and method for using same - Google Patents

Abstract

A decoder and method for using a new picture or frame type is provided. This type is referred to an SP-picture. The temporal redundancies are not exploited in I-frames, compression efficiency of I-frame coding is significantly lower than the predictive coding. A method allows use of motion compensated predictive coding to exploit temporal redundancy in the sequence while still allowing perfect reconstruction of the frame using different reference frames. Methods using this new picture type provide for error resilience/recovery, bandwidth scalability, bitstream switching, processing scalability, random access and other functions.

The SP-type picture provides for, among other functions, switching between different bitstreams, random access, fast forward and fast error-recovery by replacing I-pictures to increase the coding efficiency. As will be demonstrated, SP-pictures have the property that identical SP-frames may be obtained even when they are predicted using different reference frames.

Description

CROSS REFERENCE TO RELATED APPLICATIONS
• This application is related to and claims priority from Provisional Application No. 60/259.529 filed on Jan. 3, 2001, incorporated herein by reference.[0001]
BACKGROUND
• This invention relates generally to the field of the multimedia applications. More particularly, this invention relates to a decoder/decompressor and method for decoding streaming video.[0002]
• Multimedia applications that include audio and streaming video information have come into greater use. Several multimedia groups have established and proposed standards for compressing/encoding and decompressing/decoding the audio and video information. MPEG standards, established by the Motion Picture Expert Group, are the most widely accepted international standards in the field of the multimedia applications. ITU-Telecommunications Standardization have developed video coding standards established by the Video Coding Experts Group (VCEG). Other standards are JPEG and Motion JPEG established by the Joint Photographic Expert Group.[0003]
• The following are incorporated herein by reference:[0004]
• Gisle Bjontegaard, “H.26L Test Model Long Term Number 5 (TML-5) draft0”, document Q15-K-59, ITU-T Video Coding Experts Group (Question 15) Meeting, Oregon, USA Aug. 22-25, 2000. Keiichi Hibi, “Report of the Ad Hoc Committee on H.26L Development”, document Q15-H-07, ITU-T Video Coding Experts Group (Question 15) Meeting, Berlin, Aug. 03-06, 1999. Gary S. Greenbaum, “Remarks on the H.26L Project: Streaming Video Requirements for Next Generation Video Compression Standards”, document Q15-G-11, ITU-T Video Coding Experts Group (Question 15) Meeting, Monterey, Feb. 16-19, 1999. G. Bjontegaard, “Recommended Simulation Conditions for H.26L”, document Q15-I-62, ITU-T Video Coding Experts Group (Question 15) Meeting, Red Bank, N.J., Oct. 19-22, 1999. G. Bjontegaard, “H.26L Test Model Long Term Number 6 (TML-6) draft0”, document VCEG-L45, ITU-T Video Coding Experts Group Meeting, Eibsee, Germany, Jan. 09-12, 2001. ATM & MPEG-2 Integrating Digital Video into Broadband Networks by Michael Orzessek and Peter Sommer (Prentice Hall Upper Saddle River N.J.).[0005]
• The purpose of the video coding is to remove the redundancy in the image sequence so that the encoded data rate is commensurate with the available bandwidth to transport the video sequence while keeping the distortion between the original and reconstructed images as small as possible. The redundancy in video sequences can be categorized into spatial and temporal redundancy. Spatial redundancy refers to the correlation between neighboring pixels in a frame while temporal redundancy refers to correlation between neighboring frames.[0006]
• For every pixel of a image, color information must be provided. Typically, color information is coded in terms of the primary color components red, green and blue (RGB) or using a related luminance/chrominance model, known as the YUV model.[0007]
• Typical video codec employs three types of frames: intra frames (I-frames) and predicted frames (P-frames) and Bi-directional-frame (B-frames). Coding of a frame is performed independently from the others. I-frame, exploits only the spatial correlation of the pixels within the frame. Coding of P-frames exploits spatial as well temporal redundancies between the successive frames. Since in a typical video sequence the objects appearing in a sequence don't change rapidly from one frame to the next frame, i.e., the adjacent frames in a sequence are highly correlated, higher compression efficiencies are achieved when using P-frames. The terms frame and picture have been interchanged in the art. A frame contains all the color and brightness information that is need to display a picture. A picture is divided into a number of blocks, which are grouped into macroblocks. Each block contains a number of lines, with each line holding a number of samples of luminace or chrominance pixel values from a frame.[0008]
• FIGS. 1 and 2 show multimedia coding using MPEG as an example. FIG. 1A is a diagram of an MPEG audio and[0009]video decoder120 that performs decompression of the video and/or audio data which has been compressed and coded according to the MPEG algorithm. Thesystem decoder110 reads the encodedMPEG data stream101 having interlaced compressed video and/or audio data, and generates necessary timing information; Video Presentation Time Stamp (VPTS)104; System Clock Reference (SCR)105 which is also referred to as system time clock (STC); Audio Presentation Time Stamp (APTS)106; and separated video encodedbit streams102 and audio encodedbit streams103. Thevideo decoder111 decompresses thevideo data stream102 and generates adecompressed video signal107. Theaudio decoder112 decompresses theaudio data stream103 and generates thedecompressed audio signal108. Thedecompressed video signal107 is coupled to a display unit, while thedecompressed audio signal108 is coupled to an audio speaker or other audio generation means.
• The MPEG encoded/compressed data stream may contain a plurality of encoded/compressed video data packets or blocks and a plurality of encoded/compressed audio data packets or blocks. An MPEG encoder encodes/compresses the video packets based on video frames, also referred to as pictures. These pictures or frames are source or reconstructed image data consisting of three rectangular matrices of multiple-bit numbers representing the luminance and chrominance signals. For example, H.263+ uses four luminance blocks and two chrominance blocks of 8×8 pixels each. FIGS.[0010]2A-2C illustrate the type of encoded/compressed video frames that are commonly utilized for MPEG standard. FIG. 2A depicts an Intra-frame or I-type frame200. The I-type frame or picture is a frame of video data that is coded without using information from the past or the future and is utilized as the basis for decoding/decompression of other type frames. FIG. 2B is a representation of a Predictive-frame or P-type frame210. The P-type frame or picture is a frame that is encoded/compressed using motion compensated prediction from an I-type or P-type frame of its past, in this case, I.sub.1200. That is, a previous frame is used to encode/compress a present given frame of video data.205arepresents the motion compensated prediction information to create a P-type frame210. FIG. 2C depicts a Bi-directional-frame or B-type offrame220. The B-type frame or picture is a frame that is encoded/compressed using a motion compensated prediction derived from the l-type reference frame (200 in this example) or P-type reference frame in its past and the I-type reference frame or P-type reference frame (210 in this example) in its future or a combination of both. B-type frames are usually inserted between I-type frames or P-type frames. FIG. 2D represents a group of pictures in what is called display order I.sub.1 B.sub.2 B.sub.3 P.sub.4 B.sub.5 P.sub.6. FIG. 2D illustrates the B-type frames inserted between I-type and P-type frames and the direction which motion compensation information flows.
• Motion compensation refers to using motion vectors from one frame to the next to improve the efficiency of predicting pixel values for encoding/compression and decoding/decompression. The method of prediction uses the motion vectors to provide offset values and error data that refer to a past or a future frame of video data having decoded pixel values that may be used with the error data to compress/encode or decompress/decode a given frame of video data.[0011]
• The capability to decode/decompress P-type frames requires the availability of the previous l-type or P-type reference frame and the B-type frame requires the availability of the subsequent I-type or P-type reference frame. For example, consider the encoded/compressed data stream to have the following frame sequence or display order:[0012]
• I.sub.1 B.sub.2 B.sub.3 P.sub.4 B.sub.5 P.sub.6 B.sub.7 P.sub.8 B.sub.9 B.sub.10 P.sub.11 . . . P.sub.n-3 B.sub.n-2 P.sub.n-1 I.sub.n.[0013]
• The decoding order for the given display order is:[0014]
• I.sub.1 P.sub.4 B.sub.2 B.sub.3 P.sub.6 B.sub.5 P.sub.8 B.sub.7 P.sub.11 B.sub.9 B.sub.10 . . . P.sub.n-1 B.sub.n-2 I.sub.n.[0015]
• The decoding order differs from the display order because the B-type frames need future I-type or P-type frames to be decoded. P-type frames require that the previous l-type reference frame be available. For example, P.sub.4 requires I.sub.1 to be decoded such that the encoded/compressed I.sub.1 frame needs to be available. Similarly, the frame P.sub.6 requires that P.sub.4 be available in order to decode/decompress frame P.sub.6. B-type frames, such as frame B.sub.3, require a past and future I-type or P-type reference frames, such as P.sub.4 and I.sub.1 in order to be decoded. B-type frames are inserted frames between l-type, P-type, or a combination during encoding and are not necessary for faithful reproduction of an image. The frames before an I-type frame, such as P.sub.n-1 in the example, are not needed to decode an I-type frame, and no future frames require P.sub.n-1 in order to be decoded/decompressed.[0016]
• One problem with decoding is that the display process may be slower than the decoding process. For example, a 240.times.16 picture requires 3072 clock cycles to decode(76.8 us at 40 Mhz); it takes 200 us to display 16 lines of video data at a 75 Hz refresh rate (13 us.times.16=200 us). The video frames are buffered before being displayed. There is usually a one frame delay between display and decoding. The difference between display and decoding leads to a condition known as tearing. Tearing occurs when the display frame is overwritten by the decoded frame.[0017]
• FIG. 1B depicts tearing. A decoded/decompressed[0018]frame132 of data representing the image of aclosed door133 is currently stored in abuffer135. This decode/decompressed frame is currently being displayed ondisplay unit140. During this display period another decoded/decompressedframe130 with data representing the image of anopen door131 is stored inbuffer135. Thedisplay unit140 will now start displaying using information from the new frame now stored in135. The result is a partial display of the first storedimage141 and partial display of the new storedimage142.
• Video streaming has emerged as one of the essential applications over the fixed internet and- in the near future over 3G multimedia networks. In streaming applications, the server starts streaming the pre-encoded video bitstream to the receiver upon a request from the receiver which plays the stream as it receives with a small delay or no delay. The problem with video streaming is that the best-effort nature of today's networks causes variations of the effective bandwidth available to a user due to the changing network conditions. The server should then scale the bitrate of the compressed video to accommodate these variations. In case of conversational services that are characterized by real-time encoding and point-to-point delivery, this is achieved by adjusting, on the fly, the source encoding parameters, such as quantization parameter or frame rate, based on the network feedback. In typical streaming scenarios when already encoded video bitstream is to be streamed to the client, the above solution can not be applied. A situation similar to tearing as described above would occur.[0019]
• Thus, there is a need to provide bandwidth scalability which will allow a server to dynamically switch between the streams of encoded video in order to accommodate variations of the bandwidth available to the client.[0020]
• The above-mentioned references are exemplary only and are not meant to be limiting in respect to the resources and/or technologies available to those skilled in the art.[0021]
SUMMARY
• A new picture or frame type and method of using same is provided. This type of novel frame type is referred to as a SP-picture. The temporal redundancies are not exploited in I-frames, compression efficiency of I-frame coding is significantly lower than the predictive coding. The proposed method allows use of motion compensated predictive coding to exploit temporal redundancy in the sequence while still allowing perfect reconstruction of the frame using different reference frames. This new picture type provides for error resilience/recovery, bandwidth scalability, bitstream switching, processing scalability, random access and other functions.[0022]
• The SP-type picture provides for, among other functions, switching between different bitstreams, random access, fast forward and fast error-recovery by replacing I-pictures to increase the coding efficiency. As will be demonstrated, SP-pictures have the property that identical SP-frames may be obtained even when they are predicted using different reference frames.[0023]
• These and other features, aspects, and advantages of embodiments of the present invention will become apparent with reference to the following description in conjunction with the accompanying drawings. It is to be understood, however, that the drawings are designed solely for the purposes of illustration and not as a definition of the limits of the invention, for which reference should be made to the appended claims.[0024]
BRIEF DESCRIPTIONS OF THE DRAWINGS
• FIG. 1A is a prior art block diagram of an MPEG decoding system.[0025]
• FIG. 1B is a drawing showing the problem of tearing within prior art devices where frame data from two different frames stored consecutively in frame buffer memory is displayed on a display device.[0026]
• FIGS.[0027]2A-2D are diagrams showing the prior art encoding/compression of video frames.
• FIG. 3 is a block diagram of a generic motion-compensated predictive video coding system (encoder).[0028]
• FIG. 4 is a block diagram of a generic motion-compensated predictive video coding system (decoder).[0029]
• FIG. 5 is an illustration showing switching between[0030]bitstreams1 and2 using SP-pictures.
• FIG. 6 is a block diagram of a decoder in accordance with the preferred embodiment of the invention.[0031]
• FIG. 7 is an illustration of random access using SP-pictures.[0032]
• FIG. 8 is an illustration of a fast-forward process using SP-pictures.[0033]
• FIG. 9 is a set of graphs comparing coding efficiencies of I, P, and SP-pictures.[0034]
• FIG. 10 is a set of graphs comparing performance of SP-pictures and I-pictures when used at fixed one second intervals.[0035]
• FIG. 11 is a set of graphs comparing performance of SP-pictures and I-pictures in fast-forward application, one second Intervals.[0036]
DETAILED DESCRIPTION
• The simplest way of achieving bandwidth scalability in case of pre-encoded sequences is by producing multiple and independent streams of different bandwidth and quality. The server dynamically switches between the streams to accommodate variations of the bandwidth available to the client.[0037]
• Now assume that we have multiple bitstreams generated independently with different encoding parameters, such as quantization parameter, corresponding to the same video sequence. Since encoding parameters are different for each bitstream, the reconstructed frames of different bitstreams at the same time instant will not be the same. Therefore when switching between bitstreams, i.e., starting to decode a bitstream, at arbitrary locations would lead to visual artifacts due to the mismatch between the reference frames used to obtain predicted frame P(x, y). Furthermore, the visual artifacts will not only be confined to the switched frame but will further propagate in time due to motion compensated coding.[0038]
• The main observation is that perfect (mismatch-free) switching between bitstreams, in the current standards, is possible only at the positions where the future frames/regions do not use any information previous to the current switching location and the information at the current location is made available to the client.[0039]
• In prior art solutions, the approach adopted is to insert periodic I-frames during encoding and allow the switch to occur only at these I-frames. Since I-frames may be reconstructed independently from the previous frames, switching at these frames doesn't cause any mismatch. Thus a new type of picture or frame is needed as is architecture and methods of using said new frame type.[0040]
• A new decoder architecture is provided which has the property that identical frames may be obtained even when they are predicted using different reference frames. The picture type obtained using this structure will be called SP-frame also may be referred to as picture.[0041]
• A system for P-frame encoding and decoding is provided and is shown in FIGS. 3 and 4. Referring to FIGS. 3 and 4, a communication system comprising an[0042]encoder300 of FIG. 3 and adecoder400 of FIG. 4 is operable to communicate a multimedia sequence between a sequence generator and a sequence receiver. Other elements of the video sequence generator and receiver are not shown for the purposes of simplicity. The communication path between sequence generator and receiver may take various forms, including but not limited to a radio-link.
• [0043]Encoder300 is shown in FIG. 3 coupled to receive video input on line301 in the form of a current frame to be encoded I(x, y), called the current frame. The video input may be provided to a motion estimation andcoding block370 through305 and to an input of asubtractor307. The motion estimation andcoding block370 may also be coupledframe memory350 to receive indications of a previously coded and transmitted frame R(x, y), called a reference frame. The motion estimation andcoding block370 may also be coupled tomultiplexor380 to provide motion information for bitstream.
• The current frame I(x, y), is partitioned into rectangular regions of M×N pixels. By (x, y) we denote location of the pixel within the frame. These blocks may be encoded using either only spatial correlation (intra-coded blocks) or both spatial and temporal correlation (inter-coded blocks). In what follows, we concentrate on inter blocks. Each of inter-coded blocks may be predicted from one of the previously coded and transmitted reference frame, which at given instant is available in the[0044]Frame Memory350 of theencoder300 and of theFrame Memory450 of thedecoder400 in FIG. 4. Theframe memory350 is coupled to a Motion Compensated (MC)prediction block360. TheMC prediction block360 is operable to generate a prediction frame P(x, y) which is provided to an input ofsubtractor307 andadder345.MC prediction block360 is also coupled to the motion estimation andcoding block370 to receive motion information.
• The prediction information may be represented by two dimensional motion vector (Δx, Δy) where Δx is the horizontal and Δy is the vertical displacement, respectively of the pixels between the current frame and the reference frame. The motion estimation and[0045]coding block370 calculates the motion vectors (Δx, Δy). In the Motion Compensated (MC)Prediction block360, the motion vectors together with the reference frame are used to construct prediction frame P(x, y):
• P(x, y)=R(x+Δx, y+Δy).
• Subsequently, the prediction error E(x, y), i.e., the difference between the current frame and the prediction frame P(x, y) is calculated by:[0046]
• E(x, y)=I(x, y)−P(x, y).
• In[0047]transform block310, the prediction error for each K×L block is represented as weighted sum of a transform basis functions f.sub.ij(x, y),$E\left(X,Y\right)=\sum _{i=1}^K\sum _{J=1}^LC.\mathrm{sub}.\mathrm{err}\left(l,j\right)\mathrm{fsub}.\mathrm{ij}\left(X,Y\right).$

  
• The weights c.sub.err(l,j), corresponding to the basis functions are called transform coefficients. These coefficients are subsequently quantized in quantization block[0048]320:
• I.sub.err(l, j)=Q(C(l, j),QP)
• where I.sub.err(i, j) are the quantized coefficients. The operation of quantization introduces loss of information—the quantized coefficient can be represented with smaller number of bits. The level of compression (loss of information) is controlled by adjusting the value of the quantization parameter (QP).[0049]
• The quantization block[0050]320 is coupled to both amultiplexor380 and aninverse quantization block330 and in turn aninverse transform block340.Blocks330 and340 provide prediction error which is added to the MC predicted frame P(x, y) by added345 and the result stored inframe memory350.
• Motion vectors and quantized coefficients are further encoded using Variable Length Codes (VLC) which further reduce the number of bits needed for their representation. Encoded motion vectors and quantized coefficients as well as other additional information needed to represent each coded frame of the image sequence constitute a[0051]bitstream415 which is transmitted to thedecoder400 of FIG. 4. Bitstream may be multiplexed380 before transmission.
• FIG. 4 shows the[0052]decoder400 of the communication system.Bitstream415 is received fromencoder300 of FIG. 3.Bitstream415 is demultiplexed viademultiplexor410. Dequantized coefficients d.sub.err(l,j) are calculated in the inverse quantization block420:
• d.sub.err(i, j)=Q⁻¹(I.sub.err(l, j),QP).
• In[0053]inverse transform block430, the dequantized coefficients are used to obtain compressed prediction error:$E.\mathrm{sub}.c\left(X,Y\right)=\sum _{i=1}^K\sum _{J=1}^LC.\mathrm{sub}.\mathrm{err}\left(l,j\right)f.\mathrm{sub}.\mathrm{ij}\left(X,Y\right).$

  
• The pixels of the current coded frame are reconstructed by finding the prediction pixels in the reference frame R(x,y) using the received motion vectors and then adding to the compressed prediction error in[0054]adder435 resulting in decoded video:
• I.sub.c(x, y)=R(x+Δx, y+Δy)+E.sub.c(x, y).
• These values can be further normalized and filtered.[0055]
• One of the key requirements for video streaming is to scale the transmission bitrate of the compressed video according to the changing network conditions. In case of conversational services that are characterized by real-time encoding and point-to-point delivery, this is achieved by adjusting on the fly the source encoding parameters, such as quantization parameter or frame rate based on the network feedback. In typical streaming scenarios when already encoded video sequence is to be streamed to the client the above solution can not be applied.[0056]
• The simplest way of achieving bandwidth scalability in case of pre-encoded sequences is by producing multiple and independent streams of different bandwidth and quality. The server dynamically switches between the streams to accommodate variations of the bandwidth available to the client. Since the encoding algorithms employ motion-compensated prediction, switching between bitstreams at arbitrary P-type pictures, although possible, would lead to visual artifacts due to the mismatch between the reconstructed frames at the same time instant of different bitstreams. The visual artifacts will further propagate in time.[0057]
• In the current video encoding standards, perfect (mismatch-free) switching between bitstreams is possible only at the positions where the future frames/regions do not use any information previous to the current switching location, i.e., at I-frames. Furthermore, by placing l-frames at fixed (e.g. 1 sec) intervals, VCR functionalities, such as random access or “Fast Forward” and “Fast Backward” (increased playback rate) for streaming video content, are achieved. User may skip a portion of video and restart playing at any I-frame location. Similarly, increased playback rate, i.e., fast-forwarding, can be achieved by transmitting only I-pictures.[0058]
• It is, however, well known that I-frames require a lot more bits than the motion-compensated predicted frames. An embodiment of the present invention provides a novel picture type, SP-picture, which allows switching from one bitstream to another, enables VCR like functionalities without introducing any mismatch while still utilizing motion compensated prediction. One of the properties of SP-pictures is that identical SP-frames may be obtained even when different reference frames are used.[0059]
• Bitstream Switching[0060]
• An example of how to utilize SP-frames to switch between different bitstreams is illustrated in the FIG. 5. FIG. 5 shows two bitstreams corresponding to the same sequence encoded at different bitrates—[0061]bitstream1510 andbitstream2520. Within each encoded bitstream, SP-pictures may be placed at the locations at which one wants to allow switching from one bitstream to another (pictures S.sub.1513 and S.sub.2523). When switching frombitstream1 tobitstream2, another SP-picture will be transmitted. This is show in FIG. 5 by picture S.sub.12550. Pictures S.sub.2523 and S.sub.12550 are represented by different bitstreams, i.e., S.sub.2 (S.sub.12) uses the previously reconstructed frames frombitstream2 as the reference frames, however their reconstructed values are identical.
• Random Access[0062]
• Application of SP-pictures to enable random access is depicted in FIG. 7. SP-pictures are placed at fixed intervals within[0063]bitstream1720 (e.g. picture S.sub.1 (730)) which is being streamed to the client. To each one of these SP-pictures there is a corresponding pair of pictures generated and stored as another bitstream (bitstream2 (740)):
• I-picture, I.sub.2 ([0064]750), at the temporal location preceding SP-picture.
• SP-picture[0065]710, S.sub.2, at the same temporal location as SP-picture.
• Bitstream[0066]1 (720) may then be accessed at a location corresponding to an I-picture in bitstream2 (740). For example to accessbitstream1 at frame I.sub.2, first the pictures I.sub.2, S.sub.2 frombitstream2 are transmitted and then the following pictures frombitstream1 are transmitted.
• Fast-forward[0067]
• FIG. 8 is an illustration of a fast-forward process using SP-pictures. If the[0068]bitstream2 constitutes of only SP-pictures predicted from each other but at larger temporal intervals (e.g. 1 sec) as illustrated, SP-pictures can be used to obtain “Fast Forward” functionality. Furthermore, “Fast Forward” can start and stop at any location in the bitstream. Similarly, “Fast Backward” functionality can be obtained.
• Video Redundancy Coding[0069]
• SP-pictures have other uses in applications in which they do not act as replacements of I-pictures. Video Redundancy Coding can be given as an example (VRC). “The principle of the VRC method is to divide the sequence of pictures into two or more threads in such a way that all camera pictures are assigned to one of the threads in a round-robin fashion. Each thread is coded independently. In regular intervals, all threads converge into a so-called sync frame. From this sync frame, a new thread series is started. If one of these threads is damaged because of a packet loss, the remaining threads stay intact and can be used to predict the next sync frame. It is possible to continue the decoding of the damaged thread, which leads to slight picture degradation, or to stop its decoding which leads to a drop of the frame rate. Sync frames are always predicted out of one of the undamaged threads. This means that the number of transmitted I-pictures can be kept small, because there is no need for complete re-synchronization.” For the sync frame, more than one representation (P-picture) is sent, each one using a reference picture from a different thread. Due to the usage of P-pictures these representations are not identical. Therefore, mismatch is introduced when some of the representations cannot be decoded and their counterparts are used when decoding the following threads. Usage of SP-pictures as sync frames eliminates this problem.[0070]
• Error Resiliency/Recovery[0071]
• Multiple representations of a single frame in the form of SP-frames predicted from different reference pictures, e.g., the immediate previously reconstructed frames and a reconstructed frame further back in time, can be used to increase error resilience. Now, consider the case when an already encoded bitstream is being streamed and there has been a packet loss leading to a frame loss. The client signals the lost frame(s) to the sender which responds by sending the next SP-frame in the representation that uses frames that have been already received by the client.[0072]
• We have described the application of SP-pictures in different application/functionality scenarios. Note that the bitstreams in the applications discussed above could have different bitrates, frame sizes and frame rates. Depending on the client's available bandwidth, decoding and viewing capabilities, the appropriate streams can be streamed and moreover, the streams could be dynamically changed to accommodate any changes in these. In the following, we provide a detailed description of SP-picture encoding/decoding within the context of H.26L.[0073]
• SP-frame comprises blocks encoded using only spatial correlation among the pixels (intra blocks) and blocks encoded using both spatial and temporal correlation (inter blocks).[0074]
• For each inter block:[0075]
• The prediction of this block, P(x,y), is formed using received motion vectors and a reference frame.[0076]
• The transform coefficients C.sub.pred for P(x,y) corresponding to basis functions f.sub.ij(xy) are calculated.[0077]
• The quantized values of c.sub.pred are denoted as I.sub.pred and the dequantized values of I.sub.pred are denoted as d.sub.pred.[0078]
• Quantized coefficients I.sub.err, for the prediction error are received from the decoder. The dequantized values of these coefficients will be denoted as d.sub.err.[0079]
• Value of each pixel S(x,y) in the inter block is decoded as a weighted sum of the basis functions f.sub.ij(x,y) where the weigh values d.sub.rec will be called dequantized reconstruction image coefficients. The values of d.sub.rec, have to be such that coefficients c.sub.rec, exist by which quantization and dequantization d.sub.rec can be obtained. In addition, values d.rec have to fulfill one of the following conditions:[0080]
• d.sub.rec=d.sub.pred+d.sub.err; or
• C.sub.rec=C.sub.pred+d.sub.err.
• Values S(x,y) can be further normalized and filtered.[0081]
• How to utilize SP-frames to switch between different bitstreams is explained in FIG. 5. Within each encoded bitstream, SP-pictures should be placed at locations at which one wants to allow switching from one bitstrearm to another (pictures S.sub.1 ([0082]513), and S.sub.2 (523) in FIG. 5). When switching from bitstream1 (510) to bitstream2 (520), another picture of this type will be transmitted (FIG. 5 picture S.sub.12 (550) will be transmitted instead of S.sub.2 (523)). Pictures S.sub.2 (523) and S.sub.12 (550) in FIG. 5 are represented by different bitstreams. However, their reconstructed values are identical.
• The invention is described in view of certain embodiments. Variations and modification are deemed to be within the spirit and scope of the invention. The following describes the preferred implementation of the invention as illustrated in FIG. 6.[0083]
• In the preferred mode of implementation coefficients d.sub.rec are calculated as follows:[0084]
• Form prediction of current block, P(x,y), using received motion vectors and the reference frame.[0085]
• Calculate transform coefficients c.sub.pred for P(x,y) corresponding to basis functions f.sub.ij(x,y) (Transform block[0086]660). Quantize these coefficients (Quantization block670). The quantized values will be referred to as quantized prediction image coefficients and denoted as I.sub.pred.
• Obtain quantized reconstruction image coefficients I.sub.rec by adding the received quantized coefficients for the prediction error I.sub.err to I.sub.pred, i.e., I.sub.rec=I.sub.pred−I.sub.err.[0087]
• Dequantise I.sub.rec. The dequantised coefficients, output of the Inverse Quantization block, are equal to d.sub.rec.[0088]
• In the following, we describe the encoding of SP-frames for the decoder structure described as the preferred embodiment of the invention.[0089]
• As can be observed from FIG. 5, there are two types of SP-frames, specifically, the SP-frames; placed within the bistream, e.g., S.sub.1 ([0090]513) and S.sub.2 (523) in FIG. 5, and the SP-frames (S.sub.12 in FIG. 5) that will be sent when there is a switch between bitstreams (frombitstrearn1 to bitstream2). The encodings of S.sub.2 (523) and S.sub.12 (550) are such that their reconstructed frames are identical although they use different reference frames as described below.
• First, we describe the encoding of SP-frames placed within the bitstream, e.g., S.sub.1 ([0091]513) and S.sub.2 (523) in FIG. 5. The original frame is partitioned into blocks and each inter coded block is predicted from one of the earlier reconstructed frames. The prediction frame is formed as described in above. The transform coefficients c.sub.orig and c.sub.pred are calculated for the original frame I(x,y) and prediction frame P(x,y), respectively. C.sub.orig and C.sub.pred are quantized to obtain I.sub.orig and I.sub.pred, respectively. The quantized prediction error coefficients I are then obtained by subtracting I.sub.orig from I.sub.pred, i.e., I.sub.err,=I.sub.orig−I.sub.pred. Motion vectors and the quantized prediction error coefficients are encoded using VLC and the corresponding bitstream is transmitted to the decoder.
• Let I[0092]².sub.err and I².sub.pred denote the quantized coefficients of the prediction error and the prediction frame, respectively, obtained from encoding of S.sub.2. With the procedure described above. Note that in this case, quantized reconstruction image coefficients are given by I².sub.rec=I².sub.err−I².sub.pred. Assume that there will be a switch from bitstream1 (510) to bitstream2 (520) at S.sub.2 (523). The encoding of S.sub.12 (550) follows the same procedures as in the encoding of S.sub.2 (523) with the following exceptions: The first difference is that the reference frames are the reconstructed frames obtained from the decoding of thebitstream1 up to the current frame. Secondly, the quantized prediction error coefficients are calculated as follows: 1¹².sub.err=I².sub.rec−I¹².sub.pred where and I¹².pred denotes the quantized prediction image coefficients. The updated quantized prediction error coefficients and the motion vectors are transmitted to the decoder.
• When decoding frame S.sub.12 ([0093]550), using the reconstructed frames frombitstream1 before the switch, coefficients I¹².sub.pred are constructed and added to the received quantized prediction error coefficients I¹².sub.err as described above, i.e., I¹².sub.rec=I¹².sub.err+I¹².sub.pred=I¹².sub.rec+I¹².sub.pred−I¹².pred=I².sub.rec. Note that I¹².rec and I².sub.rec are identical, i.e., S.sub.2 and S.sub.12 are identical. In summary, although S.sub.12 (550) and S.sub.2 (523) have different reference frames, they have identical reconstruction values.
• The changes required in H.26L in order to implement this embodiment of the present invention are described. Although H.26L is used an example standard, embodiments of the present invention and any variations and modifications therefrom are deemed to be within the spirit of scope of the invention.[0094]
• SP-Picture Decoding[0095]
• Additional picture types Ptype Code_number=5 is added to H.26L for signaling SP-picture:[0096]
• SP-picture has the same syntax as P-picture. However, interpretation of some of the syntax element differs for Inter and Copy type macroblocks. The macroblocks with type “Inter” are reconstructed as follows:[0097]
• 1. Decode levels (both a magnitude and a sign) of the prediction error coefficients, L.sub.err, and motion vectors for the macroblock.[0098]
• 2. After motion compensation, for each 4×4 block in the predicted macroblock, perform forward transform.[0099]
• An example of a forward transform is provided by Gisle Bjontegaard, “H.26L Test Model Long Term Number 5 (TML-5) draft0”, document Q15-K-59, ITU-T Video Coding Experts Group (Question 15) Meeting, Oregon, USA Aug. 22-25, 2000. Instead of DCT, an integer transform with basically the same coding property as a 4×4 DCT is used. The transformation of some pixels a,b,c,d (say) into 4 transform coefficients A,B,C,D may be defined by:[0100]
• A=13a+13b+13c+13d
• B=17a+7b−7c−17d
• C=13a−13b−13c+13d
• D=7a−17b+17c−7d
• The inverse transformation of transform coefficients A,B,C,D into 4 pixels a′,b′,c′,d′ is defined by:[0101]
• a′=13A+17B+13C+7D
• b′=13A+7B−13C−17D
• c′=13A−7B−13C+17D
• d′=13A−17B+13C−7D
• Due to the fact that the expressions above are not normalized, a′=676a. Normalization may be performed in the quantization/dequantization process and a final shift after inverse quantization.[0102]
• The transform/inverse transform may be performed both vertically and horizontally in the same manner as in H.263.[0103]
• For chroma component, an additional 2×2 transform for the DC coefficients may be performed. The 2 dimensional 2×2 transform procedure is illustrated below. DC0,1,2,3 are the DC coefficients of 2×2 chroma blocks.[0104]

  DC0	DC1	Two dimensional 2 × 2	=> DDC (0, 0)	DDC (1, 0)
  		transform
  DC2	DC3		DDC (0, 1)	DDC (1, 1)

• Definition of transform:[0105]
• DCC(0,0)=(DC0+DC1+DC2+DC3)/2
• DCC(1,0)=(DC0−DC1+DC2−DC3)/2
• DCC(0,1)=(DC0+DC1−DC2−DC3)/2
• DCC(1,1)=(DC0−DC1−DC2+DC3)/2
• Definition of inverse transform:[0106]
• DC0=(DCC(0,0)+DCC(1,0)+DCC(0,1)+DCC(1,1))/2
• DC1=(DCC(0,0)−DCC(1,0)+DCC(0,1)−DCC(1,1))/2
• DC2=(DCC(0,0)+DCC(1,0)−DCC(0,1)−DCC(1,1))/2
• DC3=(DCC(0,0)−DCC(1,0)−DCC(0,1)+DCC(1,1))/2
• Quantize obtained coefficients K L.sub.pred=(K×A(QP)+0.5×2[0107]²⁰)/2²⁰where A(QP) is defined below in the following example from the art.
• An example of quantizing is provided by Gisle Bjontegaard, “H.26L Test Model Long Term Number 5 (TML-5) draft0”, document Q15-K-59, ITU-T Video Coding Experts Group (Question 15) Meeting, Oregon, USA Aug. 22-25, 2000.[0108]
• The quantization/dequantization process may perform ‘normal’ quantization/dequantization as well as take care of the above transform process which did not contain normalization of transform coefficients. 32 different QP values may be used.[0109]
• The QP signaled in the bitstream applies for luma quantization/dequantization referred to as QP.sub.luma. For chroma quantization/dequantization a different value—QP.sub.chroma—is used. The relation between the two is:[0110]
• QP.sub.luma[0111]
• 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31[0112]
• QP.sub.chroma[0113]
• 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 17 18 19 20 20 21 22 22 23 23 24 24 25 25[0114]
• When QP is used in the following we mean QP.sub.luma or QP.sub.chroma depending on what is appropriate.[0115]
• Two arrays of numbers are used for quantization/dequantization.[0116]
• A(QP=0, . . , 31)[0117]
• 620, 553, 492, 439, 391, 348, 310, 276, 246, 219, 195, 174, 155, 138, 123, 110, 98, 87, 78, 69, 62, 55, 49, 44, 39, 35, 31, 27, 24, 22, 19, 17[0118]
• B(QP=0, . . ,31)[0119]
• 3881,4351,4890,5481,6154,6914,7761,8718,9781,10987,12339,13828,15523,17435,19561, 21873,24552,27656,30847,34870,38807,43747,49103,54683,61694,68745,77615,89113,10 0253,109366,126635,141533[0120]
• The relation between A( ) and B( ) is: A(QP)×B(QP)×676[0121]²=2⁴⁰.
• It is assumed that a coefficient K is quantized in the following way:[0122]
• LEVEL=(K×A(QP)+fx2[0123]²⁰)/2²⁰f is in the range (0-0.5) and f has the same sign as K.
• Dequantization:[0124]
• K′=LEVEL×B(QP)
• After inverse transform this results in pixel values that are 2[0125]²⁰too high. A shift of 20 bits (with rounding) is therefore needed on the reconstruction side. The definition of transform and quantization is designed so that no overflow will occur with use of 32 bit arithmetic.
• The coefficients of K is obtained by:[0126]
• L.sub.pred=(K×A(QP)+0.5×2²⁰)/2²⁰
• Add the quantized prediction image coefficients, L.sub.pred, to the prediction error coefficients levels, i.e., L.sub.rec=L.sub.err+L.sub.pred.[0127]
• 3. The coefficients, L.sub.rec, are dequantized and inverse transform is performed for these dequantized levels. Dequantization and inverse transform are performed. The reconstructed values are equal to the result of the inverse transformation shifted by 20 bit (with rounding) as described above.[0128]
• For Copy type macroblocks, only steps 2 and 3 are performed. While applying the deblocking filter, both Inter and Copy macroblocks are treated as Intra macroblocks with coefficients represented by L.sub.rec.[0129]
• SP-Picture Encoding[0130]
• SP-picture placed within a single bitstream (pictures S.sub.1 ([0131]513 of FIG. 5 and S.sub.2 (523 of FIG. 5). The encoding and decoding of SP-picture which is transmitted when switching from one bitstream to another (picture S.sub.12 (550) and picture S.sub.2 (523)) are described below.
• When encoding an SP-picture placed within a bitstream, the prediction error coefficients for luminance may be obtained as follows:[0132]
• 1. After motion compensation, for each 4×4 block in the predicted macroblock and in the original image, forward transform is performed. For chroma component an additional 2×2 transform for DC coefficients is performed. The transform coefficients for the original image are denoted as K.sub.orig and for the predicted image as K.sub.pred.[0133]
• 2. Transform coefficients for the predicted blocks are quantized. Obtained levels are denoted as L.sub.pred.[0134]
• 3. The prediction error coefficients are obtained by K.sub.err=K.sub.orig−L.sub.pred×2[0135]²⁰/A(QP) and can be quantized.
• Let as assume that we want to encode the SP-picture, denoted as S.sub.12, to switch from[0136]bitstream1 tobitstream2. The reconstructed values of this picture have to be identical to the reconstructed values of SP-picture inbitstream2, denoted as S.sub.2, to which we are switching. The bitstream of the Intra macroblocks in frame S.sub.2are copied to S.sub.12. The encoding of Inter macroblocks is performed as follows:
• 1. Form the predicted frame for S.sub.12 by performing motion estimation with the reference frames being pictures preceding S.sub.1 in[0137]bitstream1.
• 2. Perform for each 4×4 block in the predicted macroblock 4×4 forward transform. An additional 2×2 transform for DC coefficients of the chroma component is performed.[0138]
• 3. Quantize the obtained coefficients and subtract the quantized coefficient levels from the corresponding L.sub.rec of S.sub.2-picture. The resulting levels are the levels of the prediction error which will be transmitted to the decoder.[0139]
• FIG. 9 illustrates a comparison of the coding efficiency of each picture type, namely I, P and SP frames in terms of their PSNR as a function of bitrate performances for the selected sequences (container and hall sequences). These results are generated by encoding every frame in the sequence with the same picture type, i.e., I, P or SP, except the first frame which is always an I-frame. As can be observed from FIG. 8, the coding efficiency of an SP-picture is worse than P-frames while it is significantly much better than that of I-frames. Although the coding efficiency of each picture type is important, it is important to note that the SP-frames provide functionalities that are usually achieved only with I-frames.[0140]
• In the following, we illustrate the simulation results when SP and I-frames are introduced at fixed intervals. FIG. 10 illustrates the results obtained with the following conditions: The first frame is encoded as an I-Picture and at fixed intervals, in this[0141]case 1 sec, the frames are encoded as I or SP-pictures while the rest of the frames are encoded as P-pictures.
• Also included in FIG. 10 is the performance achieved when all the frames are encoded using P-frames. Note that in this case, none of the functionalities mentioned earlier can be obtained while this provides a benchmark for comparison of both SP and I-picture cases. As can be observed from FIG. 10, SP-pictures, while providing the same functionalities as an I-picture, yield significantly better performance in terms of PSNR as a function of bitrate. For example for the Hall sequence around 40 Kbps, there is 2-2.5 dB improvement when SP-frames are used instead of I-frames while there is 0.5 dB penalty over the benchmark all P-frame conditions.[0142]
• FIG. 11 demonstrates the performance improvement using SP-pictures instead of I-frames for “Fast-forward”. We also include the performance achieved by using only P-frames. Note again, in this case, restarting playing is not possible without a mismatch. Nevertheless, this provides another benchmark for comparison of the other schemes. As can be observed from FIG. 11, there is significant improvement with SP-pictures over I-pictures. For container sequence at 10 Kbps, an improvement of 5.5 dB can be obtained.[0143]
• In another embodiment of the present invention, the coding efficiency of SP-frames is improved by using a separate quantization value for the predicted frame than the prediction error coefficients. The changes required in H.26L in order to implement this embodiment of the present invention are described. Although H.26L is used an example standard, embodiments of the present invention and any variations and modifications therefrom are deemed to be within the spirit of scope of the invention.[0144]
• Another embodiment for SP-Picture decoding[0145]
• Picture types Ptype Code_number=5 is added to H26.L standard to signal SP-picture. If the Ptype indicates SP-picture, an additional codeword SPQP(5 bits) follows PQP codeword. Otherwise SP-picture has the same syntax as P-picture; however, interpretation of some of the syntax element may differ for Inter and Copy type macroblocks.[0146]
• Additional array of numbers is used when decoding SP-picture:[0147]
• F(QP)=2²⁰/A(QP)
• where constant A(QP) is defined above in the section on quantization.[0148]
• Decoding of the luma component is described first. The macroblocks with type “Inter” are reconstructed as follows:[0149]
• 1. Decode levels (both a magnitude and a sign) of the prediction error coefficients, L.sub.err, and motion vectors for the macroblock.[0150]
• 2. After motion compensation, for each 4×4 block in the predicted macroblock, perform forward transform as described above. Using the resulting prediction image coefficients K.sub.pred and prediction error coefficient levels L.sub.err calculate coefficients[0151]
• K.sub.rec=(K.sub.pred+L.sub.err×F(QP.sub.1))
• and quantize them:[0152]
• L.sub.rec=(K.sub.rec×A(QP.sub.2)+0.5×2²⁰)/2²⁰.
• The value of QP.sub.1 is given by PQP and QP.sub.2 by SPQP.[0153]
• 3. The coefficients, L.sub.rec, are dequantized using QP.sub.2 and the inverse transform is performed for these dequantized levels. Dequantization and inverse transform are performed as described in Gisle Bjontegaard, “H.26L Test Model Long Term Number 5 (TML-5) draft0”, document Q15-K-59, ITU-T Video Coding Experts Group (Question 15) Meeting, Oregon, USA Aug. 22-25, 2000. The reconstructed values are equal to the result of the inverse transformation shifted by 20 bit (with rounding).[0154]
• For Copy type macroblocks, only the[0155]steps 2 and 3 are performed. While applying deblocking filter, both Inter and Copy macroblocks are treated as Intra macroblocks with coefficients represented by L.sub.rec.
• Dequantization of the chroma component is performed in a similar manner with the following differences. In[0156]step 2 additional 2×2 transform for the DC coefficients is performed after 4×4 transform. Values of QP.sub.1 and QP.sub.2 are changed according to the relation between QP values used for luma and chroma specified above.
• Another embodiment for SP-Picture Encoding[0157]
• When encoding an SP-picture placed within a bitstream, the prediction error coefficients for luminance can be obtained as follows:[0158]
• 1. After motion compensation, for each 4×4 block in the predicted macroblock and in the original image, forward transform is performed. The transform coefficients for the original image are denoted as K.sub.orig and for the predicted image as K.sub.pred.[0159]
• 2. Transform coefficients for the predicted blocks are quantized using QP=QP.sub.2 as specified above with f=0.5. The resulting levels are denoted as L.sub.pred.[0160]
• 3. The prediction error coefficients K.sub.err=K.sub.orig−L.sub.pred×F(QP.sub.2) can be quantized using one of the methods described in G. Bjontegaard, “H.26L Test Model Long Term Number 6 (TML-6) draft0”, document VCEG-L45, ITU-T Video Coding Experts Group Meeting, Eibsee, Germany, Jan. 09-12, 2001, with QP=QP.sub.1.[0161]
• The same procedure is used to calculate coefficients for chrominance with the following differences: In[0162]step 1, an additional 2×2 transform for DC coefficients is performed after 4×4 transform. Values of QP₁and QP₂are changed according to the relation between QP values used for luma and chroma specified in above.
• Let as assume that we want to encode the SP-picture, denoted as S.sub.12, to switch from bitstream[0163]1 (510) to bitstream2 (520) in FIG. 5. The reconstructed values of this picture have to be identical to the reconstructed values of SP-picture in bitstream2 (520), denoted as S.sub.2 (523), to which we are switching. The bitstream of the Intra macroblocks in frame S.sub.2 (523) are copied to S.sub.12 (550). The encoding of Inter macroblocks is performed as follows:
• 1. Form the predicted frame for S.sub.12 by performing motion estimation with the reference frames being pictures proceeding S.sub.1 ([0164]513) in bitstream1 (510).
• 2. Perform for each 4×4 block in the predicted macroblock 4×4 forward transform followed by an additional 2×2 transform for DC coefficients for chroma component.[0165]
• 3. Quantise obtained coefficients and subtract the quantized coefficient levels from the corresponding levels L.rec of S.sub.2 picture. Use QP equal to value specified by SPQP codeword of frame S.sub.2. The resulting levels are the levels of the prediction error which will be transmitted to the decoder through a communication system. Notice that for frame S.sub.12 QP values specified by PQP and SPQP are same and equal to the value given by SPQP codeword of frame S.sub.2.[0166]
• A novel SP-picture encoding/decoding method which uses different quantization values for the predicted block and the prediction error coefficients has been provided. The use of two different values of OP allows to trade off between coding efficiency of SP-pictures placed within a single bitstream and SP-pictures used when switching from one bitstream to another. The lower the value of SPQP with respect to PQP, the higher the coding efficiency of SP-picture placed within a single bitstream, while, on the other hand, the larger number of bits is required when switching to this picture. The choice of the SPQP value can be application dependent. For example, when SP-pictures are used to facilitate random access one can expect that SP frames placed within a single bitstream will have the major influence on compression efficiency and therefore SPQP value should be small. On the other hand, when SP pictures are used for streaming rate control, SPQP value should be kept close to PQP since SP-pictures sent during switching from one bitsream to another will have large share of the overall bandwidth.[0167]
• While the preferred embodiment and various alternative embodiments of the invention has been disclosed and described in detail herein, it will be obvious to those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope thereof.[0168]

Claims (9)

What is claimed is:

1. A decoder for decoding encoded data wherein identical frames may be obtained even when they are predicted using different reference frames, said decoder comprising:

means for forming a prediction of a current block of data using a plurality of motion vectors and a reference frame;

means for calculating a plurality of transform coefficients for said current block of data corresponding to a set of basis functions;

means for quantizing said coefficients creating a plurality of quantized prediction image coefficients;

means for obtaining a plurality of quantized reconstruction image coefficients by adding said received quantized coefficients for the prediction error to said plurality of quantized prediction image coefficients; and

means for dequantizing said plurality of quantized reconstruction image coefficients.

2. A decoder for decoding a block of encoded data wherein identical frames may be obtained even when they are predicted using different reference frames, said decoder comprising:

frame memory for storing a reference frame;

demultiplexor for receiving and demultiplexing said encoded data into motion information and a current frame;

motion compensation predictor coupled to said demultiplexor and said frame memory for receiving said motion information and constructing a prediction of the current block based on said motion information and reference frame;

transformer coupled to said motion compensation predictor for creating a plurality of transform coefficients;

quantisationizer coupled to said transformer for quantizing said plurality of coefficients; and

adder coupled to said quantisationizer and said demultiplexor for adding current frame information and said quantized plurality of coefficients to form a reconstructed frame.

3. The decoder ofclaim 2, further comprising:

inverse quantizationizer coupled to said adder; and

inverse transformer coupled to said inverse quantizationizer.

4. A method for encoding a frame of video data, comprising the steps of:

forming a prediction of a current block of data using a plurality of motion vectors and a reference frame;

calculating a plurality of transform coefficients for said current block of data corresponding to a set of basis functions;

quantizing said coefficients creating a plurality of quantized prediction image coefficients;

obtaining a plurality of quantized reconstruction image coefficients by adding said received quantized coefficients for the prediction error to said plurality of quantized prediction image coefficients; and

dequantizing said plurality of quantized reconstruction image coefficients.

5. A method for switching between a plurality of bitstreams in a data communication system, wherein said bitstreams correspond to a same data sequence but are encoded at different bitrates, said method comprising the steps of:

placing a first picture within each of said plurality of bitstreams in locations at which switching from one of said plurality of bitstreams to another one of said plurality of bitstreams is desired;

transmitting a second picture wherein said first picture and said second picture are represented by different bitstreams, but wherein said first picture and said second picture reconstructed values are identical.

6. A method for enabling access in a data stream, said method comprising the steps of:

placing a plurality of SP-pictures at fixed intervals within a first bitstream;

generating an I-picture and an SP-picture for each one of said plurality of SP-pictures in said first bitstream;

storing said I-picture in a second bitstream at a temporal location preceding said each one of said plurality of SP-pictures in said first bitstream; and

storing said SP-picture in said second bitstream at same temporal locations as each of said SP-pictures in said first bitstream.

7. The method ofclaim 6, wherein said second bitstream comprises only SP-pictures predicted from each other, but at longer temporal periods.

8. A method for providing Video Redundancy Coding (VRC), comprising the steps of:

dividing a sequence of pictures into a plurality of threads wherein all pictures are assigned to one of said plurality of threads in a round-robin fashion;

coding each of said plurality of threads independently;

creating a frame, wherein all of said threads converge; and

starting a second plurality of threads from said frame.

9. A method for providing error control in a data stream between a sender and a client in a communication system, said method comprising:

creating a plurality of representations of a frame in the form of a plurality of SP-pictures predicted from different reference pictures;

signaling said sender information regarding lost frames and a one of said plurality of representations received by said client; and

sending said client a SP-picture which is the next picture in said one of plurality of representations received by client.

Images