遥感图像压缩中的编码方法研究和实现

发布时间：2018-01-30 14:32

本文关键词： 遥感图像 JPEG XR 张量分解 3D-SPECK 图像压缩　出处：《哈尔滨工业大学》2014年硕士论文　论文类型：学位论文

【摘要】：遥感图像压缩是信息与信号处理领域的一个重要的研究方向，随着成像光谱技术的快速发展，图像的存储量也急剧增大，所以对遥感图像进行压缩时很有必要的。现在存在很多种压缩遥感图像的方法，但是总体上来说可以归纳为三类，第一类是基于变换的压缩编码算法；第二类是基于预测的压缩编码算法；第三类是基于VQ的压缩编码算法。在基于变换的遥感图像压缩编码算法中，采用张量分解对高光谱图像进行变换是一种比较新颖的变换技术，这种技术不但能去除高光谱图像空间上的相关性，还能去除谱间的相关性，然后结合3D-SPECK这种先进的三维数据编码方法，在不损失的图像信息的情况下进一步对图像进行了压缩。另外本文还研究了最近几年新提出的压缩算法JPEG XR标准中的熵编码部分，它的压缩性能接近JPEG2000，计算复杂度上却要低得多。由于其具有计算复杂度低、内存消耗少等特点，既支持无损压缩，又支持有损压缩，还支持区域解码，HDR等等，不过压缩时间还可以进一步缩短，所以在FPGA上实现JPEG XR的熵编码具有重要的应用前景。首先本文全面介绍了张量分解的原理和JPEG XR标准，重点分析了这两种算法的优缺点，然后提出了两种遥感图像压缩方法，基于张量的高光谱图像压缩和基于JPEG XR的光学图像压缩，从软件和硬件上对遥感图像压缩进行了一个全面的分析。最后对两种压缩方法进行了仿真实验，在基于张量的高光谱图像编码方法中，本文对比了3D-DWT等方法，信噪比提高了大约5dB左右；在基于JPEG XR的光学图像压缩部分，，本文对比了传统的JPEG2000算法，在编码和解码时间，以及重建图像的信噪比上进行了对比，在同等的压缩码率下，JPEG XR的压缩时间是JPEG2000的1/4。
[Abstract]:Remote sensing image compression is an important research direction in the field of information and signal processing. With the rapid development of imaging spectrum technology, the storage capacity of images increases dramatically. So it is necessary to compress remote sensing image. There are many methods to compress remote sensing image, but generally it can be classified into three categories: the first is compression coding algorithm based on transformation; The second kind is the compression coding algorithm based on prediction. The third is the compression coding algorithm based on VQ. In the transform based remote sensing image compression coding algorithm, Zhang Liang decomposition to transform hyperspectral image is a relatively novel transformation technology. This technique can not only remove the spatial correlation of hyperspectral images, but also remove the correlation between spectra, and then combine 3D-SPECK with the advanced 3D data coding method. The image is further compressed without loss of image information. In addition, this paper also studies the entropy coding part of the new compression algorithm JPEG XR proposed in recent years, and its compression performance is close to that of JPEG2000. Because of its low computational complexity and less memory consumption, it supports lossless compression, lossy compression, and region decoding HDR. However, the compression time can be further shortened, so the entropy coding of JPEG XR on FPGA has an important application prospect. Firstly, this paper introduces the principle of Zhang Liang decomposition and JPEG XR standard, analyzes the advantages and disadvantages of the two algorithms, and then proposes two remote sensing image compression methods. Hyperspectral image compression based on Zhang Liang and optical image compression based on JPEG XR. A comprehensive analysis of remote sensing image compression is carried out in software and hardware. Finally, two compression methods are simulated, which are based on Zhang Liang's hyperspectral image coding method. Compared with 3D-DWT and other methods, the signal-to-noise ratio (SNR) is improved by about 5 dB. In the part of optical image compression based on JPEG XR, the traditional JPEG2000 algorithm is compared, and the coding time and decoding time as well as the SNR of reconstructed image are compared. At the same compression rate, the compression time of JPEGXR is 1 / 4 of that of JPEG2000.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.73

【参考文献】