双树复小波域马尔科夫的遥感图像分割方法

针对多尺度高分辨率遥感图像像素分割在降噪时丢失大量高频信息及单一像素孤立性问题,提出了一种双树复小波变换(dual-tree complex wavelet transform,DT-CWT)和模糊马尔科夫随机场(fuzzy markov random field,FMRF)模型相结合的无监督遥感图像分割算法。首先通过DT-CWT遥感图像进行多尺度分解,并采用Bayesian阈值法对分解后的高频分量进行去噪,以增强图像的细节和边缘的表达能力并有效保留图像的主要高频信息;然后采用FMRF分割算法分别对重构后各层分量进行分割,以充分考虑像素分割的模糊性和像素邻域间的相关性;最后根据相似度融合规则融合各层分割结果。对比试验结果表明,该方法在有效去除杂点和噪声的同时能够较好地保留图像细节信息,并且边缘分割更加平滑,具有较高的分割精度和很好的鲁棒性。     

基于自适应分割的多曝光图像融合算法

针对传统多曝光图像融合存在颜色和细节信息保留不完整的问题,提出了一种新的基于自适应分割的多曝光图像融合算法。首先,采用超像素分割将输入图像分割为颜色一致的图像块,再利用结构分解将图像块分解为三个独立分量。根据各分量特点设计不同融合规则,以保留源图像中的颜色和细节信息。然后,采用引导滤波平滑各分量的权重图以及信号强度分量和亮度分量,有效地克服块效应缺陷,保留源图像中的边缘信息,减少伪影。最后,重构融合后的三个分量,得到最终的融合图像。实验结果表明,与传统的融合算法相比,所提算法在互信息（MI）上平均提升了53.6%、标准差（SD）上平均提升了24.0%。该算法能够有效地保留输入图像的颜色和细节纹理信息。     

红外与可见光图像分解与融合方法研究

红外与可见光图像融合旨在生成一幅新的图像,能够对场景进行更全面的描述。本文提出一种图像多尺度混合信息分解方法,可有效提取代表可见光特征分量的纹理细节信息和代表红外特征分量的边缘信息。本文方法将边缘信息进行进一步分割以确定各分解子信息的融合权重,以有效地将多尺度红外光谱特征注入到可见光图像中,同时保留可见光图像中重要的场景细节信息。实验结果表明,本文方法能够有效提取图像中的红外目标,实现在融合图像中凸显红外目标的同时保留尽可能多的可见光纹理细节信息,无论是主观视觉还是客观评价指标都优于现有的图像融合方法。     

迭代离散Shearlet变换异类源遥感图像融合

Shearlet变换是一种多尺度几何分析算法,适用于遥感这类纹理丰富、边缘复杂的图像处理。提出了一种基于迭代离散Shearlet变换的图像融合算法,将源图像进行分解,得到图像的多尺度、多方向树型结构表示;对粗糙分量和带通细节分量分别采用不同的融合规,得到融合图像的树型结构表示;最后进行Shearlet反变换得到融合图像。仿真结果表明,提出的算法比基于Contourlet变换的图像融合算法有更好的效果,更有利于保留纹理和边缘信息。     

一种矿井低照度图像增强算法

针对小波变换在图像边缘保持和细节处理方面无法保持平衡及多尺度Retinex算法易造成图像出现光晕伪影和噪声污染严重等问题,将小波变换与基于多尺度引导滤波的多尺度Retinex算法相结合,提出了一种矿井低照度图像增强算法。该算法首先将低照度图像进行小波分解得到高频分量和低频分量;然后对图像高频分量采用三段式阈值函数进行小波去噪,对图像低频分量采用非线性全局亮度校正以增强图像亮度,同时采用多尺度引导滤波函数代替传统多尺度Retinex算法的高斯滤波函数来估计照射分量,进而求取反射分量,并运用主成分分析法对反射分量与非线性全局亮度校正的图像进行融合,有效提升图像边缘细节保持效果;最后对图像高频分量和低频分量进行小波重构,并对小波重构后的图像进行非线性变换,解决图像泛灰问题。实验结果表明,该算法具有很强的噪声抑制能力,可有效提升图像亮度和对比度,使图像边缘保持性能和细节信息丰富度得到有效平衡,避免了图像出现光晕伪影、颜色失真等现象。     

Curvelet变换的多波段遥感图像融合研究

提出一种基于Curvelet变换的多波段遥感图像融合算法。Curvelet变换具有比小波变换更好的边缘表达,因而更适合图像的融合处理。采用具有多尺度、多方向特点的Curvelet变换对多波段遥感图像像进行分解。对于低频系数采用平均融合算法,根据高频子图边缘分布差异,对于方向高频系数采用区域边缘检测和区域谱熵算法实现多波段遥感图像的融合处理。实验结果表明,提出的算法与传统算法相比在保留原始图像边缘和纹理信息同时,可以有效地取得较好的融合视觉效果。     

一种基于低频域边缘增强的小波融合方法

提出了一种基于低频域边缘增强的小波融合方法。首先,对参加融合的两幅图像进行小波多尺度分解,然后对最高层（分辨率最低层）高频细节分量图像进行区域绝对值取大和对其它层高频细节分量图像按区域方差最大化的原则进行融合,而对低频近似分量图像采用尺度系数卷积后区域特征度量的融合方法,增强了低频域的边缘,并采用均方根误差对该方法进行了客观评价。实验结果表明该方法有很好的融合效果,与已有的低频域平均法和尺度系数卷积融合方法相比,能更好地突出低频域边缘细节信息和区域特征。     

基于曲波变换的红外/可见光图像融合

和小波变换相比较,曲波变换能更好地表示图像的边缘信息.在此基础上给出了一种基于曲波变换的图像融合方法,并将其应用于红外和可见光图像融合.首先,对红外图像和可见光图像分别进行曲波变换,得到两幅图像的低频分量和不同尺度的高频分量.在对源图像的各分量融合时,对低频分量采用平均加权进行融合.对高频分量采用取绝对值较大的方法进行融合,得到融合后的低频分量和不同尺度的高频分量,最后对这些融合后的分量进行重构,得到融合图像.仿真结果表明:和基于小波变换的融合算法相比较,该算法较好地保留了源图像的细节信息,提高了融合的效果.     

基于多尺度边缘保持分解与PCNN的医学图像融合

在医学图像融合过程中,传统多尺度分析方法多采用线性滤波器,由于无法保留图像边缘特征导致分解阶段的强边缘处出现模糊,从而产生光晕。为提高融合图像的视觉感知效果,通过结合多尺度边缘保持分解方法与脉冲耦合神经网络（PCNN）,提出一种新的图像融合方法。对源图像进行加权最小二乘滤波分解得到图像的基础层和细节层,采用高斯滤波器对基础层进行二次分解得到低频层和边缘层,将分解过程中每级边缘层和细节层叠加构建高频层,并引入非下采样方向滤波器组进行方向分析。在此基础上,利用改进的空间频率以及区域能量激励PCNN融合高频层和低频层,通过逆变换得到最终的融合图像。实验结果表明,该方法能够突出医学图像的边缘轮廓并增强图像细节,可将更多的显著特征从源图像分离并转移到融合图像中。     

抗噪的多尺度形态学边缘检测算法

为了使图像边缘检测算法的抗噪声能力更强,能检测到更加丰富的边缘信息,在多尺度形态学边缘检测算法的基础上,提出一种抗噪的多尺度形态学边缘检测算法。一方面,用小波变换法替代常用的加权平均法来融合各尺度下获取的边缘图像,对小波分解后得到的低频系数和高频系数分别采取不同的融合策略,从而有效地保留边缘的细节信息,使得融合后获得的图像清晰且细节丰富。另一方面,在用不同尺度的结构元素检测图像边缘时都采用抗噪的检测算法,因此,该算法具有较强的抗噪声能力。仿真结果表明,该算法既能有效地降低噪声对检测结果的影响,又能获得较理想的边缘图像。     

A new hybrid spectral similarity measure for discrimination among Vigna species

The reflectance spectrum of species in a hyperspectral data can be modelled as an n-dimensional vector. A spectral angle mapper (SAM) computes the angle between the vectors that is used to discriminate the species. Spectral information divergence (SID) models the data as a probability distribution so that the spectral variability between the bands can be extracted using stochastic measures. The hybrid approach of the SAM and SID is found to be a better discriminator than the SAM or SID on their own. The spectral correlation angle (SCA) is computed as a cosine of the angle of the Pearsonian correlation coefficient between the vectors. The SCA is a better measure than the SAM as it considers only standardized values of the vector rather than the absolute values of the vector. In the present article, we propose a new hybrid measure based on the SCA and the SID. The proposed method has been compared with the hybrid approach of the SID and SAM for discriminating species belonging to Vigna genus using measures such as relative spectral discriminatory power, relative discriminatory probability and relative discriminatory entropy in different spectral regions. Experimental results using the laboratory spectra show that the proposed method gives higher relative discriminatory power in the 400–700 nm spectral region.     

A comparison of error metrics and constraints for multiple endmember spectral mixture analysis and spectral angle mapper

Spectral matching algorithms can be used for the identification of unknown spectra based on a measure of similarity with one or more known spectra. Two popular spectral matching algorithms use different error metrics and constraints to determine the existence of a spectral match. Multiple endmember spectral mixture analysis (MESMA) is a linear mixing model that uses a root mean square error (RMSE) error metric. Spectral angle mapper (SAM) compares two spectra using a spectral angle error metric. This paper compares two endmember MESMA and SAM using a spectral library containing six land cover classes. RMSE and spectral angle for models within each land cover class were directly compared. The dependence of RMSE on the albedo of the modeled spectrum was also explored. RMSE and spectral angle were found to be closely related, although not equivalent, due to variations in the albedo of the modeled spectra. Error constraints applied to both models resulted in large differences in the number of spectral matches. Using MESMA, the number of spectra modeled within the error constraint increased as the albedo of the modeled spectra decreased. The value of the error constraint used was shown to make a much larger difference in the number of spectra modeled than the choice of spectral matching algorithm.     

谐波分析光谱角制图高光谱影像分类

目的 针对光谱角制图(SAM)分类算法对高光谱像元光谱曲线的局部特征和其辐射强度不敏感,而且易受噪声和维数灾难影响,致使分类效率低和精度较差等缺陷,将谐波分析(HA)技术引入到SAM高光谱影像分类中,提出一种基于谐波分析的光谱角制图(HA-SAM)高光谱影像分类算法.方法 利用HA技术将高光谱影像从光谱维变换到能量谱特征维空间,并提取低次谐波分量及特征系数(谐波余项、相位和振幅),用特征系数组成的向量代替光谱向量,对高光谱影像进行SAM分类.结果 将SAM和HA-SAM同时应用于EO-1卫星的Hyperion高光谱影像分类,通过对比和分析,验证了HA-SAM的优越性,再选择AVIRIS(airborne visible infrared imaging spectrometer)高光谱影像对HA-SAM进行验证,结果表明该算法具有较强的普适性.结论 HA-SAM提高了传统SAM高光谱影像分类的效率和精度,而且适用性较强具有良好的应用前景.     

基于多核平台的RS译码器

提出一种基于多核平台的Reed-Solomon(RS)译码器。为提高译码器的数据吞吐率,分析?RS译码算法的特点,在多核层次上进行任务划分,并在SIMD单核层次上进行数据并行处理,以减少存储器访问次数,最小化核间通信,通过多核平台实现RS(255, 239, 8)。实验结果表明,当码率最差时,该译码器的吞吐率仍可达到4.35 Gb/s。     

A novel approach for measuring hyperspectral similarity

Hyperspectral measures are used to capture the degree of similarity between two spectra. Spectral angle mapper (SAM) is an example of such measures. SAM similarity values range from 0 to 1. These values do not indicate whether the two spectra are similar or not. A static similarity threshold is imposed to recognize similar and dissimilar spectra. Adjusting such threshold is a troublesome process. To overcome this problem, the proposed approach aims to develop learnable hyperspectral measures. This is done through using hyperspectral measures values as similarity patterns and employing a classifier. The classifier acts as an adaptive similarity threshold. The derived similarity patterns are flexible, as they are able to capture the specific notion of similarity that is appropriate for each spectral region. Two similarity patterns are proposed. The first pattern is the cosine similarity vector for the second spectral derivative pair. The second pattern is a composite vector of different similarity measures values. The proposed approach is applied on full hyperspectral space and subspaces. Experiments were conducted on a challenging benchmark dataset. Experimental results showed that, classifications based on second patterns were far better than first patterns. This is because first patterns were concerned only with the geometrical features of the spectral signatures, while second patterns combined various discriminatory features such as: orthogonal projections information, correlation coefficients, and probability distributions produced by the spectral signatures. The proposed approach results are statistically significant. This implies that using simple learnable measures outperforms complex and manually tuned techniques used in classification.     

Combining magnitude and shape features for hyperspectral classification

The spectral angle mapper (SAM) and maximum likelihood classification (MLC) are two traditional classifiers for hyperspectral classification. This paper presents two methods to combine magnitude and shape features, one for each classifier. As the magnitude and shape features are complementary, combining both features can improve the classification accuracy. First, magnitude features are represented by the spectral radiance vector, whereas shape features are represented by the spectral gradient vector. Then, in SAM, each feature vector generates a spectral angle for each class. The two generated angles are added together to obtain a single similarity, which is used for the final classification. Similarly, in MLC, after the dimensionality reduction using Fisher's linear discriminant (FLD), each feature vector in the new feature space generates a likelihood. The two generated likelihoods are multiplied to obtain a single value, which is adopted for the final classification. Experimental results on an Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) data set demonstrate that the proposed methods outperform the methods with a single feature set.     

高光谱遥感图像的单形体分析方法

将n个波段的高光谱图像像元与n维空间里的散点联系起来，结合凸体几何中单形体概念研究高光谱遥感图像纯净像元提取方法，实现图像的地物精确分类识别及像元波谱分解。寻找高光谱遥感图像n维空间里的单形体并认知分析单形体是该研究方法的重要环节。通过MNF(minimum noise fraction)变换和PPI(pixel purity index)计算技术寻找到单形体，基于单形体进行像元分解分析单形体，并结合应用实例和SAM(spectral angle mapper)分类技术完成高光谱图像地物精确分类制图，验证了该研究方法的可操作性。该研究方法的优点在于不需要用户提供地物波谱信息，用于制图和波谱分解的终端单元可由图像本身得到，并由用户控制分类制图和波谱分解的详细程度。     

基于SAM与SVM的高光谱遥感蚀变信息提取

高光谱遥感技术的发展,提高了遥感技术的定量化水平,要求人们从光谱维去理解地物在空间维的变换。提出了一种光谱角匹配技术（Spectral Angle Mapper,SAM）与支持向量机（Support Vector Machine,SVM）相结合的高光谱遥感蚀变信息提取模型,在光谱维提取地表的蚀变信息。鉴于SAM算法仅考虑波谱矢量方向,忽略辐射亮度大小的缺点,利用SVM算法对SAM的提取结果进行二次分类,利用网格搜索法并结合分类精度评估进行参数寻优。通过AVIRIS高光谱数据实验证明,提取的蚀变信息分类精度为78.172 6%,Kappa系数为0.712 5。该模型计算方便,对于解决光谱维的地物分类及相似矿物的蚀变信息提取具有一定的实际意义。     

Comparison of multi-temporal ASTER images for hydrothermal alteration mapping using a fractal-aided SAM method

The shortwave infrared (SWIR) spectral bands of four multi-temporal images acquired by the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) on NASA’s Terra platform were analysed for evaluating the effects of acquisition properties and atmospheric pre-processing levels on the resulting hydrothermal alteration maps a using the fractal-aided Spectral Angle Mapper (SAM) method. Three ASTER level-1B products covering the Sar Cheshmeh area in Iran were used for hydrothermal alteration mapping. These images were converted to surface reflectance using the Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes (FLAASH) method. The low reflectance of band 5 of the level-1B products was compensated for by using the spectra of collected rock samples. Level-2 (AST2B05S) SWIR ASTER images that had already been processed were also used. Reference spectra of the main hydrothermal alteration types were extracted for each product. The threshold angles were determined using the real value–area (RV–A) fractal technique. Then, SAM classification was carried out to map hydrothermal alteration for every product. It is concluded that the level-1B products that had been converted to reflectances have a better spectral contrast than the AST2B05S product. Summer images with lower tilt angle and higher solar elevation should be used to increase the accuracy of the image classification and minimize the effect of vegetation on the spectra of index minerals. By comparing the resulting hydrothermal alteration maps with known alteration types using a confusion matrix, it was shown that the application of the RV–A fractal technique to produce less biased threshold angles increases the accuracy of SAM classification.     

谱流形快速学习算法研究

谱流形学习算法的目标是发现嵌入在高维数据空间中的低维表示,其近年来得到了广泛的应用。虽然已经取得了许多令人骄傲的成绩,但是却存在一个很大的瓶颈--计算复杂度太高,这严重阻碍了算法在实际中的应用。提出了谱流形快速学习算法,该算法包括两个降低算法复杂度的技术：（1）通过随机选择或者k-means方法从n个样本点中选出 p个锚点,把每个样本点表达为由锚点的邻域点线性组合的形式,从而设计了邻接矩阵的新形式,降低了邻接图的计算复杂度;（2）利用线性化的流形学习算法有效地计算高维数据到低维数据的映射,从而降低了优化特征值的计算复杂度。该算法在3个常用人脸数据集（Yale、ORL、Extended Yale B）上得到了验证,进一步证明了算法的有效性。