【正文】 ? 容錯(cuò)能力強(qiáng)：提供了解決在不穩(wěn)定網(wǎng)絡(luò)環(huán)境下容易發(fā)生的丟包等錯(cuò)誤的必要工具。 ? 在 MPEG7推薦的應(yīng)用中蘊(yùn)涵了許多值得研究的課題。 ? 幾個(gè)查詢例子： 音樂：在鍵盤上彈幾個(gè)音符就能得到包含（或近似）要求曲調(diào)的音樂作品列表，或以某種方式匹配音符的圖像。 ” 中間抽象層也可能存在。 MPEG7正式名稱叫作 “ 多媒體內(nèi)容描述接口 ” ，為各種類型的多媒體信息規(guī)定一種標(biāo)準(zhǔn)化的描述方法，這種描述與多媒體信息的內(nèi)容本身一起，支持用戶對(duì)其感興趣的各種 “ 資料 ” 進(jìn)行快速、有效的檢索。這種數(shù)字聲頻傳播之所以可能實(shí)現(xiàn)是因?yàn)樗恍枰s 16kbit/s的寬帶。因此， MPEG4提供了一種嶄新的交互方式－基于內(nèi)容的交互（ Contentbased Interactivity） ? 在視頻編碼方面， MPEG４ 與現(xiàn)有標(biāo)準(zhǔn)相比也有了重要突破。 MPEG4是目前視頻壓縮技術(shù)的最新發(fā)展水平。 MPEG2 ? Generic coding of moving pictures and associated audio ? Part 1 Systems ? Part 2 Video ? Part 3 Audio ? Part 4 Conformance testing ? Part 5 Software simulation ? Part 6 System extensions DSMCC ? Part 7 Audio extension NBC mode ? Part 8 VOID (withdrawn) ? Part 9 System extension RTI ? Part 10 Conformance extension DSMCC ? Part 11 IPMP on MPEG2 Systems ? 它是一個(gè)直接與數(shù)字電視廣播有關(guān)的高質(zhì)量圖像和聲音編碼標(biāo)準(zhǔn)。其中應(yīng)用最廣泛的是 VCD光盤。然后 P幀和兩個(gè) I幀可以被緩存起來，這樣接下來收到的 B幀就可以在觀看端進(jìn)行解碼。新老情景間的差別很大，這時(shí)很可能不得不發(fā)送新的場(chǎng)景。由于 JPEG只壓縮一幅單獨(dú)圖像的信息，所以 MPEG必須處理時(shí)間冗余。 the AC coefficients are ignored. ? D frames contain only the lowfrequency ponents of an image. ? A Dframe always consists of one type of macro block and only the DCcoefficients of the DCT are coded. ? D frames are used for fastforward display. This could also be realized by a suitable placement of I frames. ? P幀和 B幀圖像采用幀間編碼方式，即同時(shí)利用了空間和時(shí)間上的相關(guān)性。 “ 準(zhǔn)宏塊 ” 的邊界不是 I圖中的 16?16的宏塊，是 I圖中的一個(gè)類似塊，這一個(gè)復(fù)制過程稱為 “ 運(yùn)動(dòng) ” 。 ? I幀主要用于接收機(jī)的初始化和信道的獲取，以及節(jié)目的切換和插入， I幀圖像的壓縮倍數(shù)相對(duì)較低。一幅靜態(tài)圖像就是由許多這樣的宏塊組成。 ? 1999年 12月發(fā)布了 MPEG4。 ? JPEG和 MPEG都是在 ISO領(lǐng)導(dǎo)下的專家小組，其成員也有很大的交疊。 JPEG的目標(biāo)集中于靜止圖像壓縮，而 MPEG的目標(biāo)是針對(duì)活動(dòng)圖像的數(shù)據(jù)壓縮，但靜止圖像與活動(dòng)圖像有密切關(guān)系。 ? MPEG7為多媒體內(nèi)容描述接口標(biāo)準(zhǔn)。對(duì)于分辨率為 352?240的 NTSC制式的一幅圖像，有 22?15=330個(gè)宏塊組成。 ? I幀圖像周期性地出現(xiàn)在圖像序列中的，出現(xiàn)頻率可由編碼器選擇。由于 P是在 I的將來，所以稱為 “ 前向預(yù)測(cè) ” 。 ? P幀圖像只采用前向時(shí)間預(yù)測(cè)，可以提高壓縮效率和圖像質(zhì)量。 ? 從根本上講，這屬于差分編碼的技術(shù)。 ? 不同的幀類型在一個(gè)幀序列中應(yīng)按什么形式排列？ 要保證 I幀必須在任何幀序列中周期性地出現(xiàn)。 Quantization ? ACcoefficients of B and P frames are usually very large values, whereas those of I frames are very small. ? MPEG quantization adjusts itself accordingly. If the data rate increases too much, quantization bees more coarse. If the data rate falls, then quantization is performed with finer granularity. 3 語音編碼 ? MPEG audio coding is patible with the coding of audio data used for Compact Disc Digital Audio (CDDA) and Digital Audio Tape (DAT). ? The most important criterion is the choice of sample rate of or 48kHz (additionally 32kHz) at 16bits per sample value. Each audio signal is pressed to either 64, 96, 128, or 192Kbit/s. ? Three quality levels (layers) are defined with different encoding and decoding plexity. ? An implementation of a higher layer must be able to decode the MPEG audio signals of lower layers ? FFT is applied for audio, and the spectrum is divided into 32 nonoverlapping subbands ? noise level in each subband is determined using a psychoacoustic model. ? In the first and second layers, the appropriately quantized spectral ponents are simply PCMencoded. ? The third layer additionally performs Huffman coding. ? MPEG provides for two types of stereo sound. Two channels are processed pletely independently. In the joint stereo mode, MPEG achieves a higher pression ratio by exploiting redundancies between the two channels ? The minimal value is always 32Kbit/s. ? The layers support different maximal bit rates: layer 1 allows for a maximum of 448Kbit/s layer 2 for 384Kbit/s layer 3 for 320Kbit/s. ? For layers 1 and 2, not all binations of bit rate and mode are allowed, and a decoder is not required to support a variable bit rate. ? In layer 3, a variable bit rate is specified by allowing the bit rate index to be switched. 4 數(shù)據(jù)流 ? An audio stream is prised of frames, which are made up of audio access units, which in turn are divided into slots. ? An audio access unit is the smallest pressed audio sequence that can be pletely decoded independently of all other data. Video Stream ? A video stream is