New Techniques for Audio, Image, and Video Compression

音频、图像和视频压缩新技术

• 批准号: 9523767
• 负责人: William Pearlman
• 金额: $ 36.91万
• 依托单位: Rensselaer Polytechnic Institute
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1996
• 资助国家: 美国
• 起止时间: 1996-03-01 至 2000-02-29
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9523767&HistoricalAwards=false
• 关键词: New; Techniques; Audio; Image; Video

NCR-9523767 Pearlman, William A. Rensselaer Polytechnic Institute New Techniques for Audio, Image, and Video Compression The goal of this research is to investigate more effective techniques for digital audio, image, and video data compression. For audio, the intention is to employ more efficient entropy-constrained, conditional and unconditional, quantization in the coding of the critical subbands of the MPEG audio standard. The dependence of the subbands is next exploited using set partitioning in hierarchical trees (SPIHT) as done originally in embedded, wavelet, zerotree coding of images. The resulting embedded coding allows delivery of different qualities of service with a single compressed bit stream. A newly developed SPIHT procedure, which obtained probably the best-to-date monochrome, still image results (lossy or lossless) and with extremely fast encoding and decoding, is utilized for audio and also for color image and video coding. The extremely fast execution of the SPIHT coding procedure makes it especially suitable for video, where a requirement of real-time coding becomes an insurmountable obstacle for other procedures. Both lossy and lossless coding are investigated, as they fit naturally into the SPIHT framework. It is expected that the results will surpass those of existing standard systems, which do not produce naturally embedded code, and are far more complex operationally and hence slower in execution. Other investigations involve the use of information theoretic criteria to create optimal adaptive subband compositions and the use of trellis coded quantization (TCQ) in block discrete cosine transform (DCT) and lapped orthogonal transform (LOT) coding. The adaptive subband decompositions pay off in better compression for a given subband coding technique. TCQ coding brings an efficient, but complex technique operating on transform blocks, which preserve regional characteristics of the image. Finally, the SPIHT coding algorithm pr ovokes several theoretical questions which are addressed. Among them are the minimal information rates to determine octave ranges of random variables and convey subset partition numbers and element numbers, and the mutual information between significant bits of dependent random variables and between their respective residual bits. Satisfactory answers to these questions will lead to more efficient coding techniques with decreased complexity.

9523767 Pellman，William A.Rensselaer理工学院音频、图像和视频压缩的新技术本研究的目标是研究更有效的数字音频、图像和视频数据压缩技术。对于音频，其目的是在对MPEG音频标准的关键子带进行编码时使用更有效的受熵约束的、有条件的和无条件的量化。如最初在图像的嵌入、小波、零树编码中所做的那样，接下来使用分层树中的集合分割(SPIHT)来利用子带的相关性。由此产生的嵌入式编码允许用单个压缩比特流传递不同的服务质量。一种新开发的SPIHT过程被用于音频以及彩色图像和视频编码，该过程获得了可能是迄今为止最好的单色静止图像结果(有损或无损)，并且具有极快的编码和解码。SPIHT编码过程的极快执行使其特别适用于视频，其中实时编码的要求成为其他过程不可逾越的障碍。对有损和无损编码都进行了研究，因为它们自然地适合SPIHT框架。预计结果将超过现有标准系统，后者不会产生自然嵌入的代码，并且在操作上要复杂得多，因此执行速度会更慢。其他研究包括使用信息论准则来创建最优自适应子带组成，以及在块离散余弦变换(DCT)和重叠正交变换(LOT)编码中使用网格编码量化(TCQ)。对于给定子带编码技术，自适应子带分解在更好的压缩中是值得的。TCQ编码提供了一种在变换块上操作的高效但复杂的技术，该技术保留了图像的区域特征。最后，对SPIHT编码算法提出的几个理论问题进行了讨论。其中包括确定随机变量的倍频程范围并传递子集分区数和元素数的最小信息率，以及相依随机变量的有效位之间以及它们各自的残差位之间的互信息。对这些问题的满意回答将导致更有效的编码技术和更低的复杂性。