Scalable Compression and Transmission of Internet Multimedia

提案#0104834德州工程协会这个雄心勃勃的项目在具有挑战性的多媒体通信系统的算法设计和架构方面探索了几个新的方向。采用统一的方法在异构网络（例如Internet）和基于可扩展源编码的噪声信道（例如无线局域网）上传输多媒体数据。本课题涉及的主题有：1)基于变换系数位平面的高效算术编码的可扩展多媒体数据压缩。在图像编码方面，采用了与JPEG-2000标准相同的小波变换。重点是在传统的非渐进式网格编码量化方案中引入渐进式。采用三维小波变换进行全可伸缩视频编码。将一种新的非均匀调制复叠变换（NMCLT）应用于音频编码。为了获得较高的算术编码性能，在NMCLT或小波系数的高阶上下文建模上投入了大量的精力。此外，研究了音频编码中NMCLT系数的掩码量化噪声和三维小波视频编码中运动信息挖掘的新方法。2)通过最优分组、联合拥塞和差错控制实现互联网多媒体的可扩展传输。研究了一种基于动态规划的可扩展多媒体比特流最优分组方案。次优但低复杂度的分组方案也进行了研究。拥塞控制用于确定可用的传输速率和端到端最优错误控制方案，以减轻包丢失的影响。这是通过最优地权衡可伸缩的源代码和不相等的错误保护来实现的。

PROPOSAL #0104834TEXAS ENGINEERING EXP. STA.XIONG, ZIXIANGThis ambitious project explores several new directions in the algorithmic design and architecture of challenging multimedia communication systems. A unified approach is taken for transmission of multimedia data over heterogeneous networks (e.g., the Internet) and noisy channels (e.g., a wireless LAN) based on scalable source coding. Topics covered in the project are:1) Scalable multimedia data compression based on efficient arithmetic coding of bit planes of transform coefficients. For image coding, the wavelet transform is used like in the new JPEG-2000 standard. The focus is on introducing progressiveness in the traditionally non-progressive trellis coded quantization scheme. The 3-D wavelet transform is adopted for fully scalable video coding. A new nonuniform modulated complex lapped transform (NMCLT) is applied for audio coding. To achieve high arithmetic coding performance, a significant amount of effort is invested in high-order context modeling of the NMCLT or wavelet coefficients. In addition, new methods are studied for masking quantization noise of the NMCLT coefficients for audio coding and exploring the motion information for 3-D wavelet video coding.2) Scalable transmission of Internet multimedia via optimal packetization and joint congestion and error control. A dynamic programming based optimal packetization scheme for scalable multimedia bit streams isstudied. Suboptimal but low complexity packetization schemes are also investigated. Congestion control is used to determine the available transmission rate and an end-to-end optimal error control scheme devisedto mitigate the effects of pack loss. This is achieved by optimally trading off scalable source coding and unequal error protection.