【作者】 李飞鹏；

【导师】 李德仁； 秦前清；

【作者基本信息】 武汉大学 ， 摄影测量与遥感， 2003， 博士

【摘要】 高分辨率遥感对地观测技术是关系到国家安全和可持续发展的重点技术。从二十世纪八十年代起，世界各主要强国纷纷就这项技术的发展提出了一系列庞大的研究计划。面对严峻的国际形势和世界的发展趋势，我国在“十五”863高科技发展计划中明确提出：大力开展高分辨力多维空间信息获取前沿技术研究，突破多维高分辨力集成光学对地观测系统……。当前，遥感对地观测技术存在几个重要的发展趋势，除了影像的空间分辨率、时间分辨率(重复观测时间短)和光谱分辨率不断提高之外，立体成像技术也是一个重要的发展方向，已经发射和计划中的一些遥感卫星都已具备立体观测功能。 海量数据压缩作为信息获取与处理的一项关键技术，在过去的几年内得到了良好的发展，不过，总体而言，数据压缩技术在航天遥感中的应用尚处于初级阶段，有关的一些研究主要局限于高分辨率光学遥感影像，并且所采用的技术也还有进一步发展的余地。随着遥感卫星小型化的发展，空间数据的规模还将持续增大，信息获取和数据传输的矛盾将日益加剧，现有的压缩技术在效率和速度上无法满足应用要求，数据传输的困难将严重影响到影像分辨率的进一步提高， 针对遥感对地观测技术的发展现状，本文以高分辨率遥感影像、遥感立体像对和三线阵CCD立体影像为研究对象，研究海量遥感影像的高保真压缩技术。根据研究对象之间的逻辑关系(单幅——两幅——三幅)，本文研究分为三个部分，高分辨率遥感影像的压缩部分主要解决单幅遥感影像的实时压缩问题；在单幅影像压缩的基础上，进一步研究遥感立体像对的左、右片之间的相关性，设计相应的补偿算法以提高遥感立体像对的压缩效率；三线阵CCD影像的压缩则可视为遥感立体像对压缩研究的进一步扩展。本文的主要工作可以概括为以下几个方面： 1)提出了基于整数小波变换的遥感影像压缩SBC算法； 高分辨率遥感影像压缩具有一般图像压缩所不具备的特点，压缩算法的设计必须考虑到以下限制条件：a)高分辨率推扫式光学CCD像机幅宽达到两万像素以上；b)压缩算法复杂度应尽可能地降低，以便硬件实现；c)星上缓存空间有限；d)数据传数系统的码率固定不变；e)压缩算法的效率必须比现有基于DCT变换的明显更高。整数小波变换的出现为上述条件的满足提供了一个理想的工具。参考其他小波编码算法，本文提出了子带比特平面编码算法Sub-band Bit-plane Coding(SBC)，该算法采用整数小波变换，最优比特平面编码和上下文相关熵编码，可用于实现高分辨率遥感影像的快速编码。SBC算法支持多分辨率、多信噪比的嵌入式码流结构，可实现从无损到有损任意码率或多种质量的图像压缩。和其他压缩算法相比，SBC算法简单高效，很容易移植到硬件，而且可以并行运算，非常适合于压缩高分辨率单色或彩色遥感影像。在压缩质量上，SBC算法在低码率压缩下重建图像的质量略逊于EBCOT算法，在高码率压缩下重建图像的质量则略优于EBCOT算法。 2)提出了基于虚拟分块的图像压缩容错编码算法； 无线通信系统具有相对较高的误码率，在遭受强电磁干扰的情况下，卫星通信系统的误码率将迅速上升。在SBC编码算法的基础上，我们提出了一种简单有效的容错编码方案一一虚拟分块。该算法的基本思想就是对编码图像进行虚拟分块，给每个子块的码流加上一个同步头，使得压缩以后各个子块能够独立的解码，从而提高系统抗噪能力。虚拟分块并不对图像分块，而是对小波变换之后的系数分块，因此不会在重建图像上留下分块的痕迹。加入容错编码以后，sBc算法的抗误码能力有了较大的提高，任何误码只对局部图像发生影响，对其他分块的图像毫无影响。 3)提出了基于视差补偿和辐射补偿的遥感立体像对压缩算法: 遥感立体像对应用于地形量测时总是左、右片联合地发生作用，但是在前人的研究当中，左、右片总是被当做孤立的图像分别处理，丝毫不考虑两幅影像之间的关系。针对遥感立体像对左右片之间存在的较大相关，本文提出了一种基于自适应重叠块立体补偿的遥感立体像对压缩算法。该算法以左图为基准图像，采用自适应分块视差估计方法计算出右图的视差矢量，结合辐射校正和重叠块视差补偿技术得到平滑的右图的预测图像，以右图减去预测图像得到残差图像，然后采用小波压缩算法对残差图像进行压缩。与左、右片独立压缩相比，该算法的单图有损压缩效率可提高30一40%左右，无损压缩的性能提高5%左右， 4)提出了基于全局辐射补偿和双向视差补偿的三线阵CCD立体影像压缩算法; 针对三线阵CCD立体影像的特性，在遥感立体像对压缩算法的基础上提出了三线阵ccD立体影像压缩算法。该算法以正视影像为基准，通过双向立体补偿和全局辐射补偿预测前视影像和后视影像的三线阵CCD立体补偿算法。双向立体补偿是立体补偿技术的扩展，主要用于抵消前视、后视视差和分块辐射差对影像匹配及预测的影响;全局辐射补偿用于校正三幅影像由于拍摄角度不同而形成的辐射差。实验中我们发现三线阵CCD立体补偿算法的预测效率和两幅影像的立体补偿效率相比有所降低，导致预测效率降低的因素主要在于:l)三线阵CcD影像的前更多还原

【Abstract】 The earth observation Technique of high-resolution remote sensing is critical to nation security and sustainable development. From 1980’s on, all the developed country in the world hold up a serious of research plan in this fields. Faced on such international situation, China proposed definetly in "tenth-five" 863 high-tech development plan: Pay more attention to research the advanced technique of acquiring high-resolution information in Multi-dimension space. At present, not only the increase in spatial, spectral and time resolution but also stereo image technique, which has been primarily carried out by the sent remote sensing satellites, become main tendencies in the fields of the earth observation Technique of remote sensing.As an important technique to acquire and process information, the data compression has achieved great progress in past years, however, generally it is still the beginning of wide application in space-flight remote sensing because some related research just located in optical high resolution remote sensing image, and the current adopted techniques exist space to develop further. With the development of small size remote sensing satellite, the number of space data will augment unremittingly, and the conflict between acquiring information and transmitting data will be more and more sharp. The present available compression technique can’t meet the demands of application in the aspects of efficiency and speed so that it is difficult to transmit the data, which in turn cause not to improve further image resolution.In this paper, we aim to research high fidelity compression of high-resolution remote sensing image, stereo image pair and three-lined CCD image. According to the logic relationship among research objects (one breadth-two breadth-three breadth), this paper is separated to three parts: the first part, high-resolution remote sensing image compression, focus on real time compression of single breadth; the second part, stereo image pair compression based on the first part, is the estimation of disparity between the left frame and the right frame, a problem similar to the estimation of motion vectors in video coding; the last part, three-lined CCD image compression, is the extension of the second part. The main work of this paper is concluded as following several aspects. 1) Propose the remote sensing image compression scheme (SBC) based on integerwavelet transformationSince high-resolution remote sensing compression possesses the special character that the other type image compression have not, the design of compression scheme is restricted within the following limits: a) the swath of high-resolution push-broom optional CCD is beyond 20,000 pixels; b) the compression scheme complexity is aspossible as low so as to be realized in the way of hardware; c) the on-board memory is limited; d) the code rate of data transmitting is fixed; e) the compression efficiency is much higher than that based on DCT. Integer wavelet transformation provides an ideal tool for meeting the above limits. Referencing the other wavelet coding scheme, this paper presents Sub-band Bit-plane Coding (SBC), which adopt integer wavelet transformation, most priority Bit-plane Coding and related context entropy coding, and be applied to realize the fast compression coding of high-resolution image. SBC supports embedded coding enabling arbitrary Rate-distortion control. Contrasted with the other compression approaches, SBC is low complexity, efficient, easy to transplant to hardware, and can parallel computer. In low code rate, the reconstructed image quality of SBC is lightly worse than that of EBCOT, but in high code rate, the result is reverse.2) Propose error-resilient image compression coding scheme based on virtual partition .Wireless communication system holds relatively high rate of error code, so error code rate of satellite communication system rise greatly under the condition of intensive electromagnetism interruption. On the basis of SBC, we propose a simple and efficient error-resilient imag更多还原