【作者】 马舜峰；

【导师】 金龙旭；

【作者基本信息】 中国科学院研究生院（长春光学精密机械与物理研究所） ， 光学工程， 2011， 博士

【摘要】 随着遥感成像技术的发展,星载遥感相机的覆盖宽度和分辨率不断提高,使得获取的图像数据量急剧增加。在卫星下行信道传输带宽有限的条件下,遥感图像数据必须经过压缩编码处理。星载遥感图像压缩是空间技术应用中的关键技术之一,在保证良好的图像压缩性能前提下,如何设计实现高效且复杂度适中的压缩编码器,是目前研究的热点问题。研究适用于卫星遥感图像的压缩技术具有重要的理论意义及工程应用价值。论文以某“星载CCD遥感相机图像压缩技术研究”项目为依托,对遥感图像压缩技术展开研究,寻找满足工程应用需求的遥感图像压缩解决方案。论文首先分析了CCD遥感相机的特点,对满足工程应用的CCD相机图像压缩系统进行了分析。在深入研究JPEG2000标准算法基础上,设计了基于ADV212的遥感相机图像压缩系统。考虑到相机轻型化的需求,提出了将CCD采集单元及图像压缩模块集成于一体的设计方案,使用单片FPGA作为它们的核心控制单元。最终在硬件平台上实现了图像压缩系统的功能,并对系统的实际性能进行了测试。实验结果表明,系统符合预期的技术指标,满足工程应用需求。尽管JPEG2000算法性能优秀,但是复杂度过高,对于低功耗等有特殊要求的卫星平台并不是最适用的解决方案;另外,考虑到专用集成芯片ADV212硬件结构固定,实现图像压缩算法有一定的局限性。因此,接下来,论文主要对复杂度适中且性能高的CCSDS图像压缩算法进行了研究。为了提高了小波变换模块的处理速度,针对推扫式CCD遥感相机的成像特点,提出了一种二维小波变换的流水线处理结构,改进后的结构节省了大量缓存资源。另外,针对码率截断方法的不足,提出了一种改进的码率控制算法,提高了重构图像的质量。实验结果表明,与改进前的算法相比,峰值信噪比最大可以提高0.67dB。最后,针对小波变换处理图像方向特征的不足,探讨了Bandelet变换在遥感图像压缩中的应用,主要研究了Bandelet变换结合算术熵编码的压缩方案。实验结果表明,该方案对遥感图像主观质量的保持能力优于小波变换编码。更多还原

【Abstract】 With the increasing development of high spatial resolution and wide swath, an extensive amount of imaging data acquired from satellite remote sensing camera increases dramatically. As the bandwidth of satellite downlink channel is limited, remote sensing image must be compressed first. The remote sensing image compression is one of key technologies for spatial application. How to balance the complexity and performance of image compression is the hot issue. Research on image compression technology used for spatial remote sensing image is very important and valuable for theory and engineering application. Based on the project of image compression research for on-board satellite CCD remote sensing camera, the objective of this dissertation is to search the solution for remote sensing image compression.First, the dissertation analyses the characteristics of the CCD remote sensing camera, and also the image compression system used for engineering application. After research on JPEG2000 standard algorithm, the image compression system based on ADV212 is proposed. Taking into account the demand for lightness, the design with CCD imaging data collection and image compression integrated is proposed, and a single FPGA is their core control unit. The image compression system is implemented on hardware platform finally. The experiment results show that the proposed system meets the expected performance index, and satisfies to engineering application.Despite the excellent performance of JPEG2000 standard, it is too complicated. The JPEG2000 is not the best solution for satellite application with limited power demand. In addition, the use of ASIC is restricted. Therefore, secondly, the image compression algorithm proposed by CCSDS is described, with moderate complexity and excellent performance. To increase data processing speed of the discrete wavelet transform module, for push-broom type sensors imaging application, the pipeline structure for two-dimension discrete wavelet transform is proposed. Through the proposed pipeline structure, huge memory resource is saved. For the shortage of rate truncation, an improved rate control algorithm is proposed, and the quality of the reconstructed image is enhanced. The experimental results show that the PSNR can be increased at 0.67dB, compared to the original algorithm.However, wavelet transform is not optimal for representing directional characteristics for image. Finally, to solve the shortage of wavelet transform, the Bandelet transform used in remote sensing image compression is described. The scheme with Bandelet transform and arithmetic entropy is mainly analyzed. The result proves that the proposed scheme can preserve better subjective performance than compression measure based on wavelet transform.更多还原