JPEG 2000和JPEG-LS的实现及遥感影像压缩失真分析

JPEG 2000和JPEG-LS是目前应用比较广泛的两种压缩算法。本文介绍了JPEG 2000和JPEG-LS的编码流程、实现方法,通过试验分别分析了遥感影像在不同压缩比下的失真程度大小以及不同的失真类型。结果表明:在相同的中低倍压缩比条件下,JPEG 2000和JPEG-LS在影像细节信息保持方面效果非常接近;在高倍压缩比下,JPEG 2000和JPEG-LS压缩后影像分别出现了明显的模糊失真和间歇性失真现象。本文分析结论对相关算法的使用和改进有一定的借鉴意义。     

JPEG 2000压缩对遥感影像监督分类精度的影响

遥感图像压缩会影响分类精度,是值得研究的问题。以高分辨率遥感影像(Quick Bird)的监督分类精度评定为尺度,采用ER Mapper软件的JPEG 2000图像压缩模块对图像进行压缩,再在eCognition软件中对这9种压缩比图像进行面向对象的监督分类,生成分类精度报告。通过分析分类精度的变化,研究了JPEG 2000压缩对遥感影像分类的影响程度及其在遥感影像压缩方面的应用潜力。     

Effects of lossy compression on remote sensing image classification of forest areas

Lossy compression is being increasingly used in remote sensing; however, its effects on classification have scarcely been studied. This paper studies the implications of JPEG (JPG) and JPEG 2000 (J2K) lossy compression for image classification of forests in Mediterranean areas. Results explore the impact of the compression on the images themselves as well as on the obtained classification. The results indicate that classifications made with previously compressed radiometrically corrected images and topoclimatic variables are not negatively affected by compression, even at quite high compression ratios. Indeed, JPG compression can be applied to images at a compression ratio (CR, ratio between the size of the original file and the size of the compressed file) of 10:1 or even 20:1 (for both JPG and J2K). Nevertheless, the fragmentation of the study area must be taken into account: in less fragmented zones, high CR are possible for both JPG and J2K, but in fragmented zones, JPG is not advisable, and when J2K is used, only a medium CR is recommended (3.33:1 to 5:1). Taking into account that J2K produces fewer artefacts at higher CR, the study not only contributes with optimum CR recommendations, but also found that the J2K compression standard (ISO 15444-1) is better than the JPG (ISO 10918-1) when applied to image classification. Although J2K is computationally more expensive, this is no longer a critical issue with current computer technology.     

Archiving and distribution of 2-D geophysical data using image formats with lossless compression

Certain types of two-dimensional (2-D) numerical remote sensing data can be losslessly and compactly compressed for archiving and distribution using standardized image formats. One common method for archiving and distributing data involves compressing data files using file compression utilities such as gzip and bzip2, which are widely available on UNIX and Linux operating systems. GZIP-compressed files and bzip2-compressed files must first be uncompressed before they can be read by a scientific application (e.g., MATLAB, IDL). Data stored using an image format, on the other hand, can be read directly by a scientific application supporting that format and, therefore, can be stored in compressed form, saving disk space. Moreover, wide use of image formats by data providers and wide support by scientific applications can reduce the need for providers of geophysical data to develop and maintain software customized for each type of dataset and reduce the need for users to develop and maintain or download and install such software. This letter demonstrates the utility of standardized image formats for losslessly compressing, archiving, and distributing 2-D geophysical data by comparing them with the traditional file compression utilities gzip and bzip2 on several types of remote sensing data. The formats studied include TIFF, PNG, lossless JPEG, JPEG-LS, and JPEG2000. PNG and TIFF are widely supported. JPEG2000 and JPEG-LS could become widely supported in the future. It is demonstrated that when the appropriate image format is selected, the compression ratios can be comparable to or better than those resulting from the use of file compression utilities. In particular, PNG, JPEG-LS, and JPEG2000 show promise for the types of data studied.     

遥感影像压缩的构像质量评价

本文分析了JPEG2000压缩算法的基本原理,对某地区的数字航空遥感影像进行了压缩试验,通过比较不同压缩比下各种评价指标的变化,来研究JPEG2000压缩算法对重建影像构像质量的影响,评价结果表明压缩比的增加,重建影像灰度平均值、标准方差和信息熵在一定范围内波动;同时,随着压缩比的增加,重建影像纹理越来越粗,视觉效果降低,并且重建影像与原始影像的一致性程度降低,差别越来越大。     

平坦地区DSM到DEM的试验

讨论了平坦地区根据影像匹配的结果将DSM修正成DEM的方法,针对平坦地区一般都有建筑物和树木,利用其本身的地形特点,提出附近最小值方法自动修正DSM以获得近似正确DEM的方法。并对试验结果的精度进行了检查,发现试验的结果比影像匹配后的结果要精确许多,将建筑物和树木等植被的高程去掉,得到比较精确的DEM数据。     

高分辨率遥感影像DSM的改进半全局匹配生成方法

提出一种基于改进半全局匹配算法的高分辨率遥感影像数字表面模型（digital surface model,DSM）生成方法。首先利用影像间连接点几何约束关系对有理函数模型进行系统误差补偿,在补偿模型的基础上对影像进行分块,利用投影轨迹法逐块得到核线影像对;在密集匹配阶段,对影像建立金字塔后逐层进行半全局匹配,匹配过程中引入顾及影像纹理信息的视差图膨胀腐蚀算法约束每一层视差搜索范围,增加了视差图边缘处的有效像素数,同时减少了算法所需的内存开销和计算时间;在视差图后处理阶段,利用加权中值滤波算法保护了视差图的边缘信息;最后基于前方交会获取DSM。选取WorldView 3和资源三号立体影像进行试验,结果表明,本文方法获取的DSM精度在高程方向上接近于1.5倍GSD,并且较好地保持了地物的边缘特性,在计算效率和内存开销方面也具有较好的平衡。     

基于AMMGC三线阵影像多视匹配模型的DSM自动提取

提出了适用于三线阵数字影像多视匹配的AMMGC模型,该模型综合利用多张线阵影像,减少信息盲区,提高匹配成功率,较大程度上克服了传统双像匹配模型的不足。AMMGC模型通过金字塔影像传递匹配初值,利用匹配检查策略检验匹配结果,非常适用于特征点与格网点匹配。运用ADS40三线阵数字影像对AMMGC匹配模型进行系统试验,结果表明:该模型能够从影像上获取大量同名像点,生成密集DSM,在DSM自动采集方面具有较强的可行性与实用性。     

Fractal compression of satellite images

Fractal geometry provides a means for describing and analysing the complexity of various features present in digital images. In this paper, characteristics of Fractal based compression of satellite data have been tested for Indian Remote Sensing (IRS) images (of different bands and resolution). The fidelity and efficiency of the algorithm and its relationship with spatial complexity of images is also evaluated. Results obtained from fractal compression have been compared with popularly used compression methods such as JPEG 2000, WinRar. The effect of bands and pixel resolution on the compression rate has also been examined. The results from this study show that the fractal based compression method provides higher compression rate while maintaining the information content of RS images to a great extent than that of JPEG. This paper also asserts that information loss due to fractal compression is minimal. It may be concluded that fractal technique has many potential advantages for compression of satellite images.     

多尺度字典学习的高光谱图像压缩算法

结合小波变换及字典学习提出了一种针对高光谱图像的压缩算法。该算法首先通过小波变换构建多尺度样本集,在小波域使用K-均值奇异值分解(K-SVD)方法学习得到原子尺寸不同的多尺度字典,然后在稀疏表示的过程中,定义一个原子使用频次筛选因子,通过统计局部最优波段稀疏表示时原子使用情况,结合筛选因子对字典原子进行优化筛选,使用精简后的字典对其余波段进行稀疏求解,最后针对不同尺度的表示系数采用自适应的量化编码。实验结果表明,与目前常用的3D-SPIHT和其他的多尺度字典学习算法相比,本文算法在中低比特率下,具有更好的重建性能。     

基于FFT的星载SAR复数图像压缩及其码率控制算法

分析了成像后的星载SAR复数图像压缩的必要性及复数图像特点, 对已有压缩算法做了简单分析。在此基础上, 提出一种基于FFT变换和加权量化的星载SAR复数图像压缩算法, 同时建立了适合SAR复数图像压缩的码率控制模型, 提出一种精确码率控制方法。实验结果表明, 以幅值PSNR和平均结构相似度以及平均相位误差为性能指标, 提出的SAR复数图像压缩算法优于现有压缩算法(JPEG2000、H.264等); 同时, 提出的码率控制算法能够实现精确码率控制。     

不同格网密度的DSM/DEM对影像分类精度的影响研究

在无控制点的卫星影像正射校正中,大多采用DSM/DEM数据作为辅助数据来消除或限制因地形起伏引起的形变,然而经不同格网密度的DSM/DEM正射校正后的影像对后续处理会产生不同程度的影响,如对地物分类精度产生影响。针对这一问题,本文分别采用不同的DSM/DEM数据(China DSM 15 m、ASTER GDEM 30 m和SRTM 90 m)对资源三号影像进行正射校正,然后对正射校正后影像利用支持向量机进行分类,比较正射校正后影像结果的分类精度。结果表明:在相同重采样方法下,影像经China DSM 15 m DSM正射校正后结果的分类精度优于ASTER GDEM 30 m DEM和SRTM 90 m DEM。     

参考地理数据的大比例尺影像无控定位

针对传统摄影测量影像定位通过人工方式获取控制点导致的效率低、费用高等问题,本文提出了一种参考地理数据的大比例尺影像无控定位方法。首先利用经典SIFT算法从参考地理数据（DEM和DOM）提取带有控制信息的特征点;然后将通过POS获取的DSM配准到DEM地理坐标上,并利用形态学滤波对DSM进行处理,以提取并剔除建筑物、树木等高程剧烈变化区域的匹配点,降低校正误差;最后进行影像校正和定位。试验结果表明,本文方法能够满足大比例尺航空影像生产定位精度要求,具有高效生产、节约资源等优势。     

Segmented Principal Component Analysis for Parallel Compression of Hyperspectral Imagery

Principal component analysis (PCA) is widely used for spectral decorrelation in the JPEG2000 compression of hyperspectral imagery. However, due to the data-dependent nature of principal components, the principal component transform matrix is stored in the JPEG2000 bitstream, constituting an overhead that is often negligible if the spatial size of the image is large. However, in parallel compression in which the data set is partitioned to multiple independent processing nodes, the overhead may no longer remain negligible. It is shown that a segmented approach to PCA can greatly mitigate the detrimental effects of transform-matrix overhead and can outperform wavelet-based decorrelation which entails no such overhead.     

利用投影影像精化数字表面模型

随着摄影测量技术的成熟和人们对摄影测量产品应用的要求不断提高,真正射影像在工程与人们生活中扮演越来越重要的角色。生成高质量真正射影像的关键在于如何获取高精度的DSM。目前修正和改善DSM误差的方法普遍存在费力费时、精度缺乏保证及任意性等问题。本文针对DSM精度的修正问题提出了一种新的方法,即基于投影影像的概念,将空间点位在投影影像上的投影轨迹线作为几何约束条件,直接对现有DSM的高程进行修正的方法。通过获取高精度的DSM,进而生成更为精确的真正射影像。本文通过理论推导、单点高程和多点高程修正试验,验证了提出的新方法直观简单,适用性强,具有一定的实际应用价值。     

基于立体补偿的遥感立体像对压缩方法

提出研究遥感立体像对的压缩问题。主要讨论了左右影像的视差补偿和辐射补偿。针对遥感立体像对视差分布不均以及左右影像存在辐射差的特点,提出了一种基于立体补偿的遥感立体像对压缩算法。该算法以左片为基准图像,采用自适应视差估计计算出右片的视差矢量,结合辐射校正和重叠块视差补偿技术得到平滑的右片的预测图像,以右片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。实验结果表明,该算法能显著提高遥感立体像对的压缩性能。     

Effects of Jpeg Compression on the Accuracy of Digital Terrain Models Automatically Derived from Digital Aerial Images

It is well recognised that data volume represents a huge overhead for softcopy photogrammetry. For example, a file size of 100 Mbytes will be generated from a black and white aerial photograph if digitised with a resolution of 20μm. Large data volumes not only create storage problems but also affect the speed of image processing. As a consequence, data compression of image data is a matter of great significance. This paper describes an investigation into the effects of image compression on the accuracy of digital terrain models (DTMs) extracted from the compressed images. The JPEG system implemented in the Z/I Imaging ImageStation digital photogrammetric workstation (DPW) was used in the study. A systematic test has been carried out on the effect of different levels of JPEG compression (with Q-factors from 1 to 100) on the resulting DTM, which is automatically generated by the DPW using Match-T software. An analysis of the results from the two sites tested shows that image compression tends to cause more significant degradation when the image texture is richer, but that recommendations on Q-factors for use with the ImageStation appear to err on the side of caution. This analysis leads to some tentative conclusions and recommendations both for future investigation and for photogrammetric practice.     

以交叉累积剩余熵为准则的星载激光测高仪大光斑波形数据与地形匹配法

提出了一种基于交叉累积剩余熵的星载激光测高仪大光斑波形数据与地形匹配方法。根据星载激光测高仪大光斑回波波形信号包含地形结构信息的特性,将激光回波波形数据和数字表面模型(DSM)投影到统计特征空间,建立数据的统计特征向量,消除数据间维度差异,以交叉累积剩余熵为相似性测度匹配波形数据与地形的统计特征。试验结果表明,本方法能够较好地实现激光回波波形数据与地形的匹配,匹配精度达到一个像素以内。     

TerraSAR-X立体雷达数据提取东北典型林区DSM

森林垂直结构信息缺乏严重制约森林地上生物量估算精度的提高。立体雷达具备对森林垂直结构探测的能力,但早期雷达图像分辨率低。TerraSAR-X等高分辨率雷达数据的出现为利用立体雷达数据进行森林垂直结构探测提供了新的契机。本研究尝试采用立体雷达处理技术,利用TerraSAR-Xstripmap模式数据,以内蒙古根河大兴安岭林区和长白山自然保护区为研究区,重点探讨金字塔分层匹配策略中涉及的关键匹配参数对DSM提取精度的影响规律。结果发现:(1)影像匹配采用的金字塔层数对DSM提取精度影响明显,在大兴安岭研究区,采用5、6、7层金字塔匹配得到的DSM误差在±10m内的像元百分比分别为81.8%、77.7%和77.1%,5层金字塔能以较高的初始分辨率减少林区纹理信息的损失;在长白山研究区,采用7层金字塔得到的DSM能够抑制明显的误匹配点,提高地形提取精度;(2)影像匹配窗口对同名点识别的质量影响明显,采用25×25的匹配窗口,与9×9匹配窗口相比,大兴安岭和长白山研究区影像的平均相关性分别由0.43、0.40提高到0.49、0.45,林区纹理信息欠丰富,宜采用较大匹配窗口;(3)大兴安岭和长白山研究区提取的DSM与参考DSM线性回归的RMSE分别为6.682 m和10.384 m,DSM误差主要存在于坡度变化剧烈的地区,透视收缩和叠掩等几何畸变导致匹配结果不可靠。     

多基线机载合成孔径雷达影像匹配的SANCC法

为了有效利用合成孔径雷达(synthetic aperture radar,SAR)多基线影像的几何信息和辐射信息,提高匹配精度,提出了一种适用于多基线SAR幅度影像的自适应归一化互相关系数和(sum of adaptive normalized cross-correlation,SANCC)影像匹配方法。该方法首先利用SAR成像参数、平台参数和物方高程范围构建匹配方向线;然后引入Gestalt原理的接近性和相似性原则对窗口内的像素加权,计算获得沿匹配方向线的SANCC值;最后采用winner-take-all(WTA)优化策略获取多基线影像的匹配结果和物方三维信息。利用国产机载SAR系统获取的多基线影像进行匹配试验,结果表明与常规的相关匹配方法相比,该方法可以获得更为密集、精确的匹配点,有效减少了由重复纹理造成的误匹配,并能较好地解决由纹理匮乏导致的匹配难题。