[ Pobierz całość w formacie PDF ]
.10.2.6 MPEG-4This standard [23–27] is explicitly designed to meet the problems of sendingmoving images and videoconferencing over low bit rate, error-prone net-works such as the PSTN and mobile networks.Initially it was expected that only low resolution images, such as QCIFand sub-QCIF, would be supported on account of the low bandwidth, butthe success of the standard has led to its use at higher speeds for much higher quality sources, such as 4CIF and 16CIF, which have 4 and 16 times thenumber of pixels as CIF, respectively.MPEG-4 works with audiovisual objects (AVO) together with a scenedescription procedure for assembling them into the full display rather thantransmitting complete frames as a whole.AVOs can either be primitiveobjects, such as the sound of somebody talking or the image of a person,or else compound objects, such as a person talking.An instance of a visualobject at a specific time is called a video object plane (VOP).Binary format for scenes (BIFS) is the scene description language, which in addition toreducing latency also provides a quick means of updating scenes.Specific Applications273Like MPEG-2 it has multiple profiles and levels, the main differencebeing the addition of the QCIF and sub-QCIF levels.The profiles determinewhich of the various MPEG-4 object types are supported, while the levelsindicate the constraints placed on parameter values.The feature types inMPEG-4 are as follows: (1) basic I- and P-VOP, (2) error resolution, (3)short headers, (4) B-VOP, (5) temporal scalability of P-VOP, (6) binaryshapes, (7) sprites, and (8) N-Bit.The related profile types for natural images are shown Tables 10.13 and 10.14 [28] where “y” indicates support.Table 10.13MPEG-4 ProfilesObject TypesProfile Name(1)(2)(3) (4) (5) (6) (7) (8)SimpleyyySimple scalable yyyCoreyyyyyyMainyyyyyyyN-bityyyyyyyTable 10.14MPEG-4 LevelsProfile NameLevelBit RateMaximum ObjectsImage FormatSimpleL028.8 Kbps2Sub-QCIFSimpleL164.0 Kbps4QCIFSimpleL2128.0 Kbps4CIFSimpleL3384.0 Kbps4CIFCoreL1384.0 Kbps4QCIFCoreL22.0 Mbps 16CIFMainL22.0 Mbps 16CIFMainL315.0 Mbps 32R 601MainL438.4 Mbps 321,920 × 1,088274QoS in Integrated 3G NetworksThere are also several profiles for specific artificial features, notablytypes of animation, and a hybrid profile that has both the natural and syn-thetic features.The basic technique for coding the images is essentially that ofJPEG-2000 using wavelet transforms, and one of the key features is the useof VOPs.A video frame is described as a layered structure in which eachmain object has its own VOP, with the VOPs superimposed on each otherand on a background field.For example, in the broadcast of a game of base-ball or cricket, the main features, such as the ball and key players, could have their own VOPs (but depending on profile and level), while the sportsground would be the background VOP or sprite.Each VOP is treated inde-pendently and its evolution depicted through its own sequence of I-, P-, andB-frames.The background VOP is treated as a sprite in some profilesand coded much less frequently (with information on camera angles usedto update the display) than the main moving objects, and so it contributesmuch less network traffic in MPEG-4 than in MPEG-2 or H.263.Packetsbelonging to the different VOPs can be treated as different logical channelsover the network and be provided with differing qualities of service ifrequired.This applies particularly in video-conferencing and films for multilingual presentation.A key factor for ease of user interpretation in each case is lip synchronization of the speakers—so lips can be given their own VOP tobe superimposed on that of the face and be provided with a higher degree oferror protection on the radio link.In the case of the multilingual films, separate VOPs for the lip movements can be dubbed in addition to the soundto provide a much more convincing presentation than traditional sounddubbing.Use of the MPEG time-stamps for the lip VOPs and the audiostream allows precise lip synchronization.Another major feature of MPEG-4 is the use of visual texture coding(VTC) to provide high-quality synthetic images.VTC uses different wavelettransform and quantization techniques to the main video module, and differsin its resilience features also.MPEG-4 has three sets of audio procedures as follows:1.Natural sound can be sampled at up to 8 kHz and parameter codedby harmonic vector excitation coding (HVXC) at bit rates from 1.2to 4 Kbps.These algorithms have low MOS ratings of 3 or less.2.CELP algorithms can also code natural sound with sampling ratesin the range 8 to 16 kHz and bit rates of 6 to 24 Kbps.The precisealgorithms differ from those used for the ITU G.Series of recom-mendations and those for mobile networks.Specific Applications2753.High-quality sound can be provided using AAC algorithms withsampling rates from 8 to 96 kHz and bit rates from 16 to at least64 Kbps.Audio scaling is possible by using a CELP base with AAC enhancements [ Pobierz całość w formacie PDF ]

  • zanotowane.pl
  • doc.pisz.pl
  • pdf.pisz.pl
  • hanula1950.keep.pl

    		•