Development and investigation of a hierarchical compression algorithm for storing hyperspectral images

Hyperspectral (HS) image is a three dimensional data image where the 3rd dimension carries the wealth of spectrum information. HS image compression is one of the areas that has attracted increasing attention for big data processing and analysis. HS data has its own distinguishing feature which differs with video because without motion, also different with a still image because of redundancy along the wavelength axis. The prediction based method is playing an important role in the compression and research area. Reflectance distribution of HS based on our analysis indicates that there is some nonlinear relationship in intra-band. The Multilayer Propagation Neural Networks (MLPNN) with backpropagation training are particularly well suited for addressing the approximation function. In this paper, an MLPNN based predictive image compression method is presented. We propose a hybrid Adaptive Prediction Mechanism (APM) with MLPNN model (APM-MLPNN). MLPNN is trained to predict the succeeding bands by using current band information. The purpose is to explore whether MLPNN can provide better image compression results in HS images. Besides, it uses less computation cost than a deep learning model so we can easily validate the model. We encoded the weights vector and the bias vector of MLPNN as well as the residuals. That is the only few bytes it then sends to the decoder side. The decoder will reconstruct a band by using the same structure of the network. We call it an MLPNN decoder. The MLPNN decoder does not need to be trained as the weights and biases have already been transmitted. We can easily reconstruct the succeeding bands by the MLPNN decoder. APM constrained the correction offset between the succeeding band and the current spectral band in order to prevent HS image being affected by large predictive biases. The performance of the proposed algorithm is verified by several HS images from Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) reflectance dataset. MLPNN simulation results can improve prediction accuracy; reduce residual of intra-band with high compression ratio and relatively lower bitrates.

The compression of hyperspectral images (HSIs) has recently become a very attractive issue for remote sensing applications because of their volumetric data. In this paper, an efficient method for hyperspectral image compression is presented. The proposed algorithm, based on Discrete Wavelet Transform and Tucker Decomposition (DWT-TD), exploits both the spectral and the spatial information in the images. The core idea behind our proposed technique is to apply TD on the DWT coefficients of spectral bands of HSIs. We use DWT to effectively separate HSIs into different sub-images and TD to efficiently compact the energy of sub-images. We evaluate the effect of the proposed method on real HSIs and also compare the results with the well-known compression methods. The obtained results show a better performance of the proposed method. Moreover, we show the impact of compression HSIs on the supervised classification and linear unmixing.

Hyperspectral imaging is concerned with the measurement, analysis, and interpretation of spectra acquired from a given scene (or specific object) at a short, medium or long distance by an airbone or satellite sensor. Over the last few years, hyperspectral image data sets have been collected for a great amount of locations over the world, using a variety of instruments for Earth observation. Despite the increasing importance of hyperspectral images in remote sensing applications, there is no common repository of hyperspectral data intended to distribute and share hyperspectral data sets in the community. Quite opposite, the hyperspectral data sets which are available for public use are spread among different storage locations and present significant heterogeneity regarding the storage format, associated meta-data (if any), or ground-truth availability. As a result, the development of a standardized hyperspectral data repository is a highly desired goal in the remote sensing community. In this paper, we take a necessary first step towards the development of a digital repository for remotely sensed hyperspectral data. The proposed system allows uploading new hyperspectral data sets along with meta-data, ground-truth and analysis results, with the ultimate goal of sharing publicly available hyperspectral images within the remote sensing community. The database has been designed in order to allow storing relevant information for the hyperspectral data available through the system, including basic image characteristics (width, height, number of bands, format) and more advanced meta-data (ground-truth information, publications in which the data has been used). The current implementation consists of a front-end to ease the management of images through a web interface, thus containing both synthetic and real hyperspectral images from two highly representative instruments, such as NASAs Airborne Visible Infra-Red Imaging Spectrometer (AVIRIS) over the Cuprite Mining District in Nevada. Most importantly, the developed system includes a spectral unmixing-based content based image retrieval (CBIR) functionality which allows searching for images on the spectral unmixing information (spectrally pure components or endmembers and their associated abundances in the scene). This information is stored as meta-data associated to each hyperspectral image instance, and then used to search and retrieve images based on information content. This paper presents the design of the system and a preliminary validation of the unmixing-based retrieval functionality using both synthetic and real hyperspectral images stored in the database.

Hyperspectral image (HSI) denoising is an effective image processing technology to improve the quality of HSI. By exploiting the spectral-spatial characteristics of HSI, this paper proposes a HSI denoising method based on the Graph-structured Low Rank model and Non-Local constraint (GLRNL). Graph-structured low rank model can reveal the spectral similarity in HSI and preserve the distribution characteristics of spectral bands. The graph structure is used to ensure that the denoised HSI has similar spectral characteristics as the real HSI, and reduce the spectral distortions. Moreover, the non-local constraint in GLRNL adopts the spatial-spectral self-similarity and provides an effective side prior information for low rank calculation. Comparing with some traditional low rank and sparse model based HSI denoising methods, the proposed method achieves the competitive results on different simulated and real HSI data with various types of noise.

The compression of hyperspectral images becomes recently very attractive issue for remote sensing applications because of the volumetric data. In this paper, an efficient method for hyperspectral image compression is presented based on Tucker Decomposition (TD) and Discrete Wavelet Transform (DWT). The core idea behind our proposed technique is to apply TD on the DWT coefficients of spectral bands of hyperspectral images. Our method not only exploits redundancies between bands but also uses spatial correlation of every image band. Simulation results applied on the real hyperspectral images show a remarkable compression ratio and quality.

International audience

Hyperspectral image compression has become increasingly important in data exploitation because of enormous data volumes and high redundancy provided by hundreds of contiguous spectral channels. Since a hyperspectral image can be viewed as a 3-dimensional (3D) image cube, many efforts have been devoted to extending 2D image compression techniques to perform 3D image compression on hyperspectral image cubes. Unfortunately, some major issues generally encountered in hyperspectral data exploitation at low or very low-bit rate compression, for example, subpixels and mixed pixels which do not occur in traditional pure pixel-based image compression are often overlooked in such a 2D-to-3D compression. Accordingly, a direct application of 2D-to-3D compression techniques to hyperspectral image cubes without taking precaution may result in significant loss of crucial spectral information provided by subtle substances such as small objects, anomalies during low bit-rate lossy compression. This paper takes a rather different view by investigating lossy hyperspectral compression from a perspective of exploring spectral information, referred to as exploitation-based lossy compression and further develops spectral/spatial hyperspectral image compression to effectively preserve crucial and vital spectral information of objects which are generally missed by commonly used mean-squared error (MSE) or signal-to-noise ratio (SNR)-based compression techniques when lossy compression is performed at low bit rates. In order to demonstrate advantages of the proposed spectral/spatial compression approach applications of subpixel target detection and mixed pixel analysis are used for experiments for performance evaluation.

Super-resolution (SR) is significant for hyperspectral image (HSI) applications. In single-frame HSI SR, how to reconstruct detailed image structures in high resolution (HR) HSI is challenging since there is no auxiliary image (e.g., HR multispectral image) providing structural information. Wavelet could capture image structures in different orientations, and emphasis on predicting high-frequency wavelet sub-bands is helpful for recovering the detailed structures in HSI SR. In this study, we propose a multi-scale wavelet 3D convolutional neural network (MW-3D-CNN) for HSI SR, which predicts the wavelet coefficients of HR HSI rather than directly reconstructing the HR HSI. To exploit the correlation in the spectral and spatial domains, the MW-3D-CNN is built with 3D convolutional layers. An embedding subnet and a predicting subnet constitute the MW-3D-CNN, the embedding subnet extracts deep spatial-spectral features from the low resolution (LR) HSI and represents the LR HSI as a set of feature cubes. The feature cubes are then fed to the predicting subnet. There are multiple output branches in the predicting subnet, each of which corresponds to one wavelet sub-band and predicts the wavelet coefficients of HR HSI. The HR HSI can be obtained by applying inverse wavelet transform to the predicted wavelet coefficients. In the training stage, we propose to train the MW-3D-CNN with L1 norm loss, which is more suitable than the conventional L2 norm loss for penalizing the errors in different wavelet sub-bands. Experiments on both simulated and real spaceborne HSI demonstrate that the proposed algorithm is competitive with other state-of-the-art HSI SR methods.

One of the most challenges in the remote sensing applications is Hyperspectral image classification. Hyperspectral image classification accuracy depends on the number of classes, training samples and features space dimension. The classification performance degrades to increase the number of classes and reduce the number of training samples. The increase in the number of feature follows a considerable rise in data redundancy and computational complexity leads to the classification accuracy confusion. In order to deal with the Hughes phenomenon and using hyperspectral image data, a hierarchical algorithm based on SVM is proposed in this paper. In the proposed hierarchical algorithm, classification is accomplished in two levels. Firstly, the clusters included similar classes is defined according to Euclidean distance between the class centers. The SVM algorithm is accomplished on clusters with selected features. In next step, classes in every cluster are discriminated based on SVM algorithm and the fewer features. The features are selected based on correlation criteria between the classes, determined in every level, and features. The numerical results show that the accuracy classification is improved using the proposed Hierarchical SVM rather than SVM. The number of bands used for classification was reduced to 50, while the classification accuracy increased from 73% to 80% with applying the conventional SVM and the proposed Hierarchical SVM algorithm, respectively.

Convolutional neural network (CNN) has shown superior classification results for hyperspectral image classification. Due to high dimensionality of the hyperspectral image, usually a feature reduction method is applied to the hyperspectral data before feeding it to the CNN model. Principal component analysis is the most widely used approach to this end. However, it has two main disadvantages. It uses the mean square error measure and ignores the class separability. Moreover, it is a pixel-based method, which does not utilize the spatial information. To deal with these weaknesses, the patch-based nonparametric discriminant analysis method is proposed for hyperspectral image feature reduction in this work. However, due to the complex and heterogeneous nature of the hyperspectral images, the use of fixed square patches is not so appropriate. To deal with this difficulty, a shaping approach is introduced for the generation of adaptive patches with consistent shapes in each local region. The input patches of the CNN model are also weighted according to the spectral similarity and class boundaries information. Moreover, a simple residual CNN model is suggested for the classification of the extracted shaped input patches. The efficiency of the proposed framework is assessed in different cases using various evaluation measures on six real hyperspectral images.

Many remote sensing applications require a high-resolution hyperspectral image. However, resolutions of most hyperspectral imagers are limited to tens of meters. Existing resolution enhancement techniques either acquire additional multispectral band images or use a pan band image. The former poses hardware challenges, whereas the latter has limited performance. In this paper, we present a new resolution enhancement method that only requires a color image. Our approach integrates two newly developed techniques in the area: (1) A hybrid color mapping algorithm, and (2) A Plug-and-Play algorithm for single image super-resolution. Comprehensive experiments using real hyperspectral images are conducted to validate and evaluate the proposed method.

Spectral uncertainty is one of the most prominent spectral characteristics of hyperspectral images. Compared to the process of remote sensing hyperspectral imaging, hyperspectral imaging under land-based imaging conditions has the characteristics of variable detection directions, random imaging times, and complex environmental conditions, resulting in increased spectral uncertainty of targets in land-based hyperspectral images. The spectral uncertainty of the target mainly refers to the phenomenon of “Same spectral reflectance but different objects” and “Same object but different spectral reflectance” which poses significant difficulties for subsequent hyperspectral image target detection and recognition. In order to analyze the spectral uncertainty of hyperspectral images in land-based applications and address the issue of spectral uncertainty in similar targets, a spectral uncertainty evaluation index based on standard deviation vector was proposed. For the overall spectral separability between different types of targets, a quantitative index based on Jaccard Distance (JD-SSI) is proposed to measure the spectral separability of different kinds of targets. The experiment focused on grassland and its four typical camouflage materials, analyzing the spectral intra class differences and inter class separability of each target with grassland. It is a fundamental work for studying the spectral characteristics of land-based hyperspectral images, providing a new approach for subsequent spectral band extraction, hyperspectral image classification, and target detection tasks.

Hyperspectral image provides a way to analyze component and the fraction by spectral unmixing. The classical method is Least Squares algorithm. But it is defined without any constraint on the abundance. The Fully constrained Least Squares (FCLS) algorithm allows resulted abundances satisfying sum-to-one and non-negative, which is practical. It has been widely used till now because it may work quickly with a small unmxing error. But when the number of endmembers is large, it gets much slow. And unmxing error is high if the endmembers are not found out completely. As for improving the speed of the algorithm, there have been many research works. But fewer algorithms are developed for increasing accurateness. Here in this paper, a primal-dual interior-point method is tested for the fully constraint unmixing with the linear unmixing model being modified slightly. A dual problem is established firstly, and then the process of the algorithm is designed to decrease distance between original problem and the dual problem based on result of FCLS. In order to promise high unmxing accurateness, the unmxing error of resulted abundances is added into the optimizing goal. Several algorithms of fully constraint abundance estimation are tested on both simulation and real hyperspectral images. The experimental results show that the prompted algorithm may find the optimal abundance for existing endmembers even though the endmember set is not complete. Accurateness of unmxing is increased, especially on the real image.

Dimensionality reduction based on random projection (RP) includes two problems, namely, the dimensionality is limited by the data size and the class separability of the dimensionality reduction results is unstable due to the randomly generated projection matrix. These problems make the RP algorithm unsuitable for large-size hyperspectral image (HSI) classification. To solve these problems, this paper presents a new partitioned RP (PRP) algorithm and proves its rationality in theory. First, a large-size HSI is evenly divided into multiple small-size sub-HSIs. Afterwards, the projection matrix that maximizes the class separability is selected from multiple samplings in which the class dissimilarity measurement is defined as large inter-class distance and small intra-class variance. By using the same projection matrix, each small-size sub-HSI is projected to generate a low dimensional sub-HSI, thereby generating a low dimensional HSI. Next, the minimum distance (MD) classifier is utilized to classify the low dimensional HSI obtained by the PRP algorithm. Finally, four real HSIs are used for experiments, and three of the most popular classification algorithms based on RP are selected as comparison algorithms to validate the effectiveness of the proposed algorithm. The classification performance is evaluated with the kappa coefficient, overall accuracy (OA), average accuracy (AA), average precision rate (APR), and running time. Experimental results indicate that the proposed algorithm can obtain reliable classification results in a very short time.

Due to the multidimensional nature of the hyperspectral image (HSI), multi-way arrays (called tensor) are one of the possible solutions for analyzing such data. In tensor algebra, CANDECOMP/PARAFAC decomposition (CPD) is a popular tool which has been successfully applied for the HSI data processing. However, on the one hand, CPD requires large memory for temporal variables. As a result, the memory usually overflows during the process for a real HSI whose size is large. On the other hand, so far no finite algorithm can well-determine the rank of the tensor to be decomposed. An inappropriate number of the rank may over-fit/under-fit the information provided by the tensor. To deal with these problems, this paper proposes an improved CPD with spectral and spatial partitioning for the HSI anomaly detection. First, the original HSI is divided into a set of smaller-sized sub-tensors. Second, CPD is applied onto each sub-tensor. Then, an anomaly detection algorithm is implemented and the detection results are fused along the spectral direction. Experiments with a real HSI data set reveals that the proposed method outperforms the CPD with no partition and the traditional RX anomaly detector with better detection performance.