基于小波变换的航空遥感图像压缩方法研究与实现

【类型】学位论文

【作者】马文波 

【关键词】 航拍遥感图像,小波变换,HVS,快速算法,图像压缩,实时压缩系统

【摘要】航空航天遥感图像在人类生活和军事领域中的应用日益广泛，然而海量的遥感数据对信息的传输、存储以及处理都带来很大压力，对数据进行压缩是解决这一问题的一个行之有效的办法。本文系统介绍了作者几年来在航空遥感图像实时编码领域的主要研究工作及成果。本文的具体研究内容如下：在对提升小波变换(第二代小波变换)进行分析的基础上，利用C语言编制了样条5/3小波提升格式的DSP程序。对设计小波滤波器时需要考虑的因素进行了详细的分析。在对CDF9/7和样条5/3滤波器组的优缺点进行研究和分析的基础上，利用它们各自的优势，提出了一种适合航空遥感图像压缩的小波滤波器选择方案。对多幅航拍遥感图像小波变换系数的特性进行了详细的研究。分别从小波系数的空频局部化特性、方向选择特性、小波域能量聚集特性和衰减特性、高频系数的统计分布以及小波域系数的相关特性等诸多方面着手研究和试验，给出了图像小波系数的分布规律。提出了一种新的基于图像先验知识和小波系数统计特性的快速静态压缩算法。该算法利用大量实拍图像的先验知识以及待压缩图像的统计特性，可以使各高频子带的量化值快速收敛到最优，从而大大降低了程序运行的时耗。提出一种新的解决图像边界赝像问题的方法，它采用最优冗余重叠处理部分图像数据的途径，使得图像分块编码过程中产生的块边界赝像问题基本消失，并将此方法成功应用在下述的实时压缩系统中，达到了最佳压缩效果。提出了一种新的结合小波频带划分和人眼视觉系统CSF带通特性的图像质量评价方法，并将其应用于本文对航拍遥感图像压缩算法的评价中。采用三片TMS320C6416、FPGA以及SDRAM等器件设计实现了大数据量、高分辨率图像实时压缩系统。本系统已成功应用于某实际机载航拍CCD相机项目中。Recently, much more attentions have been focused on remote sensing image which is extensively applied to human life and military region. However, large number of remote data will bring great tension for information transformation, storage and procession. Compression is a feasible technique to solve these problems. In this paper, important research results of airborne remote sensing image encoding are outlined. The main contributions can be summarized as follows:Based on lifting scheme wavelet transform (the second generation wavelet transform), DSP program of 5/3 wavelet lifting scheme is realized by using C language.The properties of several airborne remote sensing images are specially analyzed in detail. The localization in space or frequency domain, the selectivity of different orientation, the collection and attenuation of wavelet energy, the statistical distribution characteristics and the correlation of wavelet coefficients are studied and simulated in this paper.A new fast still compression algorithm is proposed based on image priori knowledge and statistical property of wavelet coefficients. In the algorithm, the quantified value of each high frequency sub-band can realize its optimization quickly using priori knowledge and statistical characteristics of original image. In this way, the consumed time is reduced largely.A new method dealing with boundary artifact is proposed in virtue of optimal overlapping part of images. In this way, the problem of boundary artifact because of blocking compression does not exist any more.Combined with wavelet frequency band sharing and the CSF performance of human eye vision system, a new assessment method of image quality is investigated in this paper. Three pieces of TMS320C6416, FPGA and SDRAM etc are first adopted to design real-time compression system with large data and high resolution image.

【学位名称】博士

【学位授予单位】北京理工大学

【学位授予年度】2016

【导师姓名】毛二可

【读秀链接】读秀链接