【作者】 李昌坤；

【导师】 何小海；

【作者基本信息】 四川大学 ， 通信与信息系统， 2005， 硕士

【摘要】 JPEG2000标准是国际标准组织(ISO)和国际电信联盟(ITU)联合制定的新一代静止图像压缩标准。与以往的压缩标准相比,JPEG2000标准具有很多的优点,如良好的低比特压缩性能、良好的复合图像压缩性能、信噪比和分辨率的可分级特性、感兴趣区域编码、更好的错误稳健性等。JPEG2000标准凭借良好的压缩性能和高度灵活的嵌入码流,广泛地应用于图像压缩和网络传输,逐渐成为静止图像压缩的主流标准。本文深入地研究了JPEG2000标准采用的主要算法,并对其中的两个部分进行了改进。 本文首先研究了JPEG2000标准的组织结构及其第一部分的编解码流程。其完整的编码流程包括预处理、小波变换、量化和熵编码几个部分。预处理包括片分割、直流电平平移和分量变换。然后再对每一个片进行小波变换,最后对量化后的小波系数进行熵编码,输出嵌入式码流。 小波变换是JPEG2000标准的主要特点之一。本文回顾了小波变换的发展历程,分析了多分辨分析理论,然后详细研究了JPEG2000标准采用的提升小波变换,分析其实现的基本方法,并对各种传统小波基转化为提升实现的问题作了详细讨论。 具有最优截断的嵌入式块编码算法(EBCOT)是JPEG2000中熵编码的核心部分。本文对它作了深入的研究,并改进了其量化部分。EBCOT采用分位数更多还原

【Abstract】 JPEG2000 is a new standard for the compression of still image established by the International Organization for Standardization (ISO) and the International Telecommunications Union (ITU). Compared with former standards, JPEG2000 provides a set of features, such as superior low bit-rate performance, excellent compound image performance, progressive transmission by pixel accuracy and resolution, Region of Interests coding, robustness of bit errors. Because of these features,JPEG2000 standard applies to many fields such as image compression and network transmission and gradually become major standard of still image compression. In this paper, we analyzed its major algorithms and improve two parts of them.Firstly, we studied the structure of the standard and codec procedure of its part one. The encoding procedure includes several parts: preprocessing, wavelet transform, quantization and entropy coding. Preprocessing includes tiling, DC level shifting and component transformation. After that, wavelet transform is used for the analysis of the tile components into different decomposition levels. Then quantizedwavelet coefficients are compressed using entropy coding and embedded bit stream is acquired.Wavelet transform is one feature of JPEG2000 standard. We have reviewed the history of wavelet transform and analyzed multi-resolution theory firstly, then we studied basic method of lifting scheme and discussed how to realize traditional wavelet transform using lifting scheme.EBCOT is the kernel part of entropy coding in JPEG2000. We proposed a new algorithm which can find the reconstruction point self-adaptively which based on EBCOT. Some features such as SNR and resolution scalability and "random access" are achieved through bit-plane coding, and excellent compression performance through PCRD-opt In order to overcome the disadvantage that the reconstruction point is fixed to the middle of quantization section in the step of anti-quantization, a new method is proposed. In this method, the reconstruction point can be selected self-adaptively at the cost of trivial bits increase so that reconstruction coefficients are closer to original coefficients. The result indicates that this method can increase PSNR and improve the quality of reconstructed image.At last, we focused on the ROI method in JPEG2000 standard: maxshift method and scaling based method and a new ROI method is proposed. The two traditional methods implements ROI through shifting background coefficients, which waste lots of time and resources. To overcome this problem, "S-twice coding" are proposed. In this method, the highest bit-planes are encoded twice instead of shifting the background coefficients, implementation complexity are reduced, and the decoder is simpler. The result shows that the encoding time shortens 5%-10%.更多还原