【作者】 王青海；

【导师】 张兆扬； 莫玉龙；

【作者基本信息】 上海大学 ， 通信与信息系统， 2004， 博士

【摘要】 数字图象压缩在多媒体信息的存储和传输中扮演着非常重要的角色。变换编码技术是目前最成熟、应用最广泛的主流技术,离散余弦变换和离散小波变换是最常用的变换手段。基于传统香农信息理论的图象变换编码方法目前已经发展到极限,在数字图象的编码中根据人眼视觉特性发展新的编码方法是打破香农信息理论框架,提供更有效压缩方法的途径之一,从而进一步推动数字图象压缩编码技术的发展。 小波变换作为一种良好的时频分析工具,在信号处理和图象压缩编码领域获得广泛应用。本文首先对小波变换理论进行研究,详细地分析小波变换在新一代数字图象压缩标准JPEG2000中的应用。以JPEG2000的校验模型为基础讨论了JPEG2000中小波变换的分解级数,不同小波基对压缩性能和编码算法复杂度的影响,提出一种适用于多种小波基的统一提升格式实现方法,使得根据不同的图象内容选择不同的小波基成为可能。 小波变换与离散余弦变换相比,运算复杂度高,如何解决小波变换的快速实现对小波变换的广泛应用具有重要的意义。文中讨论了小波变换卷积计算和提升格式计算具体实现方法和二者之间的关系,设计了一个JPEG2000硬件实现的开发平台。该开发平台采用FPGA和ARM微处理器构成,可以方便的对算法的硬件实现和仿真进行调试分析。本文对小波变换的硬件实现进行了详细的讨论并提出相应的解决方法,并提出一种在频域上利用自由参数构造小波基的方法,采用该方法构造的小波基具有易于硬件实现的特点,从而降低小波变换硬件实现的复杂性。 结合JPEG2000的校验模型和人眼视觉特性,本文提出三种在数字图象编码中利用人眼视觉特性来改善重建图象视觉效果的方法。1) 在多级小波变换编码系统中,人眼最敏感的信息集中在第三和第四分解级和低频子带中,通过加权的方法增强这些分解级的信息,抑制其他分解级对人眼不敏感的信息可以有效地改善重建图象的视觉效果;2) 利用人眼的视觉掩蔽效应掩蔽重建图象中的编码噪声从而改善重建图象的视觉效果,视觉掩蔽效应主要体现为不同幅度的信号对编码噪声的掩蔽能力不同和相邻象素对自身编码噪声的掩蔽:3) 采用具有视觉特性的小波基也可以在相同的压缩比下有效地改善重建图象的视觉效果:采用不同的小波基进行小波变换,变换后小波系数的能量集中特性也不同,具有视觉特性更多还原

【Abstract】 The technologies of the compression of digital image act key role to storing and transferring of multimedia information. Transform coding is the most extensive mainstream coding technology, and the discrete cosine transform and discrete wavelet transform were used widely. The coding method based Shannon information theory had developed to the limit. For the development of digital image compression, finding another coding method is efficient approach, so it will promote the development of digital image compression technologies.As perfect time-frequency analyzing tool, wavelet transform is used widely in signal processing and image compression. In this paper we first research the theory of wavelet transform then describe the JPEG2000 standard that was developed in recently years in detail. Based on the verification model of JPEG2000 analyzing numbers and choosing of wavelet base in wavelet transform was discussed in detail in image compression and an uniform implementation method of lifting scheme was put forward, so we can use different wavelet base with the image context in digital image compression.Comparing to the DCT(discrete cosine transform), the realizing of DWT(discrete wavelet transform) is more complex, so solving the fast calculating of DWT is vej-y significative to use widely DWT. With discussing the relationship of lifting scheme and convolution of DWT, we design a hardware platform for JPEG2000 standard. The platform uses ARM micro-processor and FPGA chip. The hardware implementation of JPEG2000 algorithm can be debugged and analyzed on it, specially the hardware implementation of DWT is researched. We propose a biorthogonal wavelet base constructed method on frequency domain based free factor and construct a series of biorthogonal wavelet bases. These wavelet bases are low computation complexity and easily implemented on hardware. "Combining the verification model of JPEG2000 and human visual system (HVS), we propose three methods that using HVS to the digital image compression for improving the reconstruction image visual quality. 1) In multi-level wavelet transform of digital image the important coefficients information to human eyes is concentrated in the third, forth and low frequency grades. Through weighting the wavelet coefficients, the更多还原