【作者】 王翼；

【导师】 唐娉；

【作者基本信息】 中国科学院研究生院（遥感应用研究所） ， 地图学与地理信息系统， 2006， 硕士

【摘要】 随着遥感技术的不断发展，传感器的分辨率不断提高，遥感卫星越来越多的应用于国民经济和国防军事领域，并发挥着越来越重要的作用。但是，在获得更多的有价值的信息的同时，遥感数据量也开始日益庞大，无论是在存储和传输方面都给星载系统带来了巨大的压力，如果仅仅依靠增加系统的存储能力传输能力则是不够的。因此，在这种情况下，寻找一种适合星载图像特点的高速的压缩方法，并对其进行硬件适应性改造就具有十分重要的实用价值和研究意义。 基于小波变换的嵌入式图像压缩方法是当前研究较多的一类图像编码技术。其具有的高压缩比，嵌入式码流，复杂度低等特点比较符合遥感图像压缩的要求。因此本文重点对这一类图像编码方法进行研究，然后结合遥感图像、编码算法以及星载系统的特点，从中选择了等级树集合分裂算法作为适应性改造的对象，并对这个算法从小波分解、处理模式、容错性和DSP程序优化这四个方面进行了改进，取得了如下研究成果： 1．采用整数小波提升方式代替卷积运算，使算法更易于硬件实现。 2．根据推帚式成像的原理采用条带式的处理流程，减小了运算时的硬件系统的存储量和负荷，并有效的去除了条带效应的问题。 3．提出了一种基于小波空间方向树集合块的编码方式，增强了容错性，有效防止了误码扩散，减小了传输过程中误码造成的影响。 4．根据使用的DSP芯片的特点，采用了多种优化代码的方法对程序进行了优化，仿真试验表明，经过优化后程序在硬件执行时的速度明显加快。更多还原

【Abstract】 With the development of the remote sensing technology and the increase of the spatial resolution, satellites of remote sensing have been widely used in the national economy and the national defense. Although we have obtained much information by processing lots of data, more need to be processed. As a result, the On-board System is enduring a great difficulty on storage and transmission remote sensing image, and this problem can not be solved well by only increasing the ability of the storage and transmission. In this situation, it is important to find a high-speed image compression algorithm which is proper for the characteristics of remote sensing image, and to optimize it basing on specific hardware.The embedded image compression based on the wavelet transform belongs to the image coding techniques which are widely studied in recent years, and its characteristics which include higher-compression-rate, embedded bit-stream, lower computer complexity, satisfy most the requirements in the compression of remote sensing image. In this thesis, we mainly study this kind of image coding technique, and then choose the Set Partition in Hierarchical Trees (SPIHT) to perform the improvement on. The improvement is achieved by taking into account the characteristics of the remote sensing image, the compression algorithms, and the hardware systems. From four aspects, including wavelet transform, processing pattern, error resilient, and optimization of DSP code, we improved SPIHT algorithm and obtained some achievement as bellow:1. Replace the convolution by the integer wavelet lifting, and the calculation of wavelet transform becomes easy to be implemented in hardware.更多还原