【作者】 吴铮；

【导师】 何明一；

【作者基本信息】 西北工业大学 ， 信号与信息处理， 2001， 硕士

【摘要】 基于小波变换的图像编码技术是近几年来图像压缩领域的研究热点。随着多光谱遥感技术的发展，多光谱图像的压缩也受到了越来越多的关注。特别是在无人机作为载机进行多光谱数据采集的情况下，由于受到信道带宽、飞行高度、无人机载重等多方面因素的限制，在无人机上实现实时的图像压缩是一个棘手的问题。本文结合课题，在对小波理论和图像的小波编码技术进行了较为深入的研究基础上，利用小波编码技术实现了机载多光谱遥感图像的压缩。 本文首先从信号分析和数字信号处理的角度对小波分析理论进行了系统的总结和介绍。然后利用Mallat算法实现了图像的小波变换，并结合图像压缩的应用目的解决了算法中的细节问题，提出从线性相位、正则性、消失矩、紧支性、振铃效应和算法复杂度各方面综合考虑以解决算法中最关键的滤波器选择问题。接着分析了图像小波变换系数的空域、时域和统计分布特性，对现有的小波编码技术进行了回顾。并实现了基于零树量化的SPIHT（Set Partitioning inHierarchical Trees）编码算法，经过实验比较证明了该算法的压缩性能比标准IPEG方法有很大的提高。最后将SPIHT算法用于多光谱图像的谱内压缩，并结合文献资料，提出用一维小波变换和线性预测两种方法去除多光谱图像的谱间相关。又进一步提出了分段DPCM算法以提高预测方法的性能。试验结果表明，分段DPCM＋SPIHT算法实时性好，对64幅图像的总编码时间不超过90秒，同时该算法能取得较好的压缩效果，对于本文中使用的机载多光谱数据，在近无损情况下压缩比可达到6以上，在有损情况下压缩比达到20以上，且图像的纹理和边缘等细节信息仍保持良好，另外由于算法简单，易为硬件实现，因此该算法具有较强的实用性。更多还原

【Abstract】 Image compression based on wavelet transform has attracted many attentions in recent years. With the development of multispectral remote-sensing technology, real- time transmission on limited bandwidth and storing of multispectral image data has become a challenging problem, so compression of multispectral images is of more and more importance. In this paper, theories about wavelet analysis and image compression are reviewed and researches on air-borne multispectral remote-sensing image compression by wavelet transform are discussed. At first, theories about wavelet analysis are systematically reviewed from the viewpoint of signal analysis and digital signal processing. Then Mallat algorithm is used to realize 2-D wavelet transform and details of the algorithm, especially selection criterion of wavelet basis (filter banks), are discussed according to the requirements of image compression. Thirdly, existing image encoding technologies by wavelet transform are reviewed based on the analysis of wavelet transform coefficients characters. SPIHT (Set Partitioning in Hierarchical Trees) encoding algorithm is realized and its performance is testified to be superior to that of JPEG. In the end, SPIHT is used to realize the interband compression of multispectral images and two methods, wavelet transform and one-order predictor, are used to spectrally decorrelate the multispectral data. In order to enhance the performance of predictor, a subgroup DPCM algorithm is proposed. Experiment results show that this subgroup DPCM+SPIHT algorithm is efficient. The nearlossless compression ratio is above 6 for the data used in this paper and details in images are kept well when compression ratio is above 20, while encoding time for all 64 multispectral images is not more than 90 seconds. The algorithm is also easy to implement by hardware, so it is a practicable solution to air-borne multispectral remote-sensing image compression.更多还原