This PDF file contains the front matter associated with SPIE Proceedings Volume 7810, including the Title Page, Copyright information, Table of Contents, and the Conference Committee listing.

This work addresses the transmission of pre-encoded video containing meteorological data over JPIP. The primary requirement for the rate allocation algorithm deployed in the JPIP server is the real-time processing demands of the application. A secondary requirement for the proposed algorithm is that it should be able to either minimize the mean squared error (MMSE) of the video sequence, or minimize the maximum distortion (MMAX). The MMSE criterion considers the minimization of the overall distortion, whereas MMAX achieves pseudo-constant quality for all frames. The proposed rate allocation method employs the FAst rate allocation through STeepest descent (FAST) method that was initially developed for video-on-demand applications. The adaptation of FAST in the proposed remote sensing scenario considers meteorological data captured by the European meteorological satellites (Meteosat). Experimental results suggest that FAST can be successfully adopted in remote sensing scenarios.

Compressive sensing (CS) is a new technique for reconstructing essentially sparse signals from a number of measurements smaller than the Nyquist-Shannon criterion. The application of CS to hyperspectral imaging has the potential for significantly reducing the sampling rate and hence the cost of the analog-to-digital sensors. In this paper a novel approach for hyperspectral compressive sensing is proposed where each band of hyperspectral imagery is sampled under the same measurement matrix. It is shown that the correlation between two neighboring band compressive sample values is consistent with that between two neighboring band pixel values. Our hyperspectral compressive sensing experimental results show that the proposed joint reconstruction method yields smaller reconstruction errors than the individual reconstruction method at various sampling rates.

Future CNES high resolution instruments for remote sensing missions will lead to higher data-rates because of the increase in resolution and dynamic range. For example, the ground resolution improvement has induced a data-rate multiplied by 8 from SPOT4 to SPOT5 [1] and by 28 to PLEIADES-HR [2]. Innovative "smart" compression techniques will be then required, performing different types of compression inside a scene, in order to reach higher global compression ratios while complying with image quality requirements. This socalled "selective compression", allows important compression gains by detecting and then differently compressing the regions-of-interest (ROI) and non-interest in the image (e.g. higher compression ratios are assigned to the non-interesting data). Given that most of CNES high resolution images are cloudy [1], significant mass-memory and transmission gain could be reached by just detecting and suppressing (or compressing significantly) the areas covered by clouds. Since 2007, CNES works on a cloud detection module [3] as a simplification for on-board implementation of an already existing module used on-ground for PLEIADES-HR album images [4]. The different steps of this Support Vector Machine classifier have already been analyzed, for simplification and optimization, during this on-board implementation study: reflectance computation, characteristics vector computation (based on multispectral criteria) and computation of the SVM output. In order to speed up the hardware design phase, a new approach based on HLS [5] tools is being tested for the VHDL description stage. The aim is to obtain a bit-true VDHL design directly from a high level description language as C or Matlab/Simulink [6].

A lossy hyperspectral images compression algorithm based on discrete wavelet transform (DWT) and segmented independent component analysis is presented in this paper. Firstly, bands are divided into different groups based on the correlation coefficient. Secondly, maximum noise fraction (MNF) method and maximum likelihood estimation are used to estimate dimensionality of data in each group. Based on the result of dimension estimation, ICA and DWT are deployed in spectral and spatial directions respectively. Finally, SPIHT and arithmetic coding are applied to the transformation coefficients respectively, achieving quantization and entropy coding. Experimental results on 220 band AVIRIS hyperspectral data show that the proposed method achieves higher compression ratio and better analysis capability as compared with PCA and SPIHT algorithms.

Nowadays, it is still a challenge to offer high quality and reliable real-time video services over wireless network. In this paper, we studies a new practical scalable video coding (SVC) multicast scheme based on unequal error protection (UEP) framework with priority transmission. Firstly, we propose a 3D layer-based priority ordering method to determine the importance of scalable video bit stream. Then we present an adaptive packetization algorithm to allocate the source and channel code bit rate for ordered layers and guarantee the layers with highest importance could be recovered first. Simulation results show that the proposed scheme offer better performance than existing methods in preserving video quality for different kinds of sequences under various channel conditions.

Low-density parity-check (LDPC) code is a linear block code known to approach the Shannon limit via the iterative sum-product algorithm. LDPC codes have been adopted in most current communication systems such as DVB-S2, WiMAX, WI-FI and 10GBASE-T. LDPC for the needs of reliable and flexible communication links for a wide variety of communication standards and configurations have inspired the demand for high-performance and flexibility computing. Accordingly, finding a fast and reconfigurable developing platform for designing the high-throughput LDPC decoder has become important especially for rapidly changing communication standards and configurations. In this paper, a new graphic-processing-unit (GPU) LDPC decoding platform with the asynchronous data transfer is proposed to realize this practical implementation. Experimental results showed that the proposed GPU-based decoder achieved 271x speedup compared to its CPU-based counterpart. It can serve as a high-throughput LDPC decoder.

The current papers have addressed the issue of synchronization in communication systems. In this paper, a method of code-aided carrier synchronization is proposed, which is based on LDPC-Hadamard code at extremely low signal-tonoise ratios (SNR) .This system consists of two parts: a low-rate LDPC-Hadamard code that is close to Shannon limit at extremely low SNR and an EM-based carrier synchronizer that makes use of the posterior probabilities of the data symbols provided by the iterative decoder. Performance of the proposed synchronizer has been illustrated by simulations in terms of the BER and the root-mean-square-error (RMSE). Simulation results show that the RMSE of the parameters performs very close to the modified Cramer-Rao bound.

In this paper, a semi-fragile watermarking scheme specifically developed for remote sensing images is presented. The method can be tuned to embed the mark depending on the content and the signature to be protected. The suggested method is based on tiling the original three-dimensional images into blocks of different sizes according to the relevance of the area to protect (bigger blocks are used for less relevant areas). For each of these blocks, the discrete Wavelet transform (DWT) is applied to each selected spectral band and the obtained LL DWT sub-bands are used to build a Tree-Structured Vector Quantization tree. This tree is then modified using an iterative algorithm until it satisfies some criterion. Once the target value is reached, the marked block is obtained using the new LL DWT sub-band together with the other original sub-bands (LH, HL and HH) of the block. A secret key produces a different criterion for each block in order to avoid copy-and-replace attacks. The use of the LL DWT sub-band for each spectral band makes it possible to obtain robustness against near-lossless compression attacks and, at the same time, relatively strong modifications are detected as tampering.

Many different approaches have been proposed in recent years for remotely sensed hyperspectral image classification. Despite the variety of techniques designed to tackle the aforementioned problem, the definition of standardized processing chains for hyperspectral image classification is a difficult objective, which may ultimately depend on the application being addressed. Generally speaking, a hyperspectral image classification chain may be defined from two perspectives: 1) the provider's viewpoint, and 2) the user's viewpoint, where the first part of the chain comprises activities such as data calibration and geo-correction aspects, while the second part of the chain comprises information extraction processes from the collected data. The modules in the second part of the chain (which constitutes our main focus in this paper) should be ideally flexible enough to be accommodated not only to different application scenarios, but also to different hyperspectral imaging instruments with varying characteristics, and spatial and spectral resolutions. In this paper, we evaluate the performance of different processing chains resulting from combinations of modules for dimensionality reduction, feature extraction/ selection, image classification, and spatial post-processing. The support vector machine (SVM) classifier is adopted as a baseline due to its ability to classify hyperspectral data sets using limited training samples. A specific classification scenario is investigated, using a reference hyperspectral data set collected by NASA's Airborne Visible Infra-Red Imaging Spectrometer (AVIRIS) over the Indian Pines region in Indiana, USA.

Statistical learning techniques have been used to dramatically speed-up keypoint matching and image registration. However, they are rarely applied to multi-spectral images. Statistical learning techniques regard various intensities as distinctive patterns. Thus, corresponding features extracted from multi-spectral images are recognized as different patterns, because the features have different intensity characteristics. In order to overcome this problem, we propose a novel statistical learning method that can be extended to multi-spectral images. The proposed approach obtains responses from multiple classifiers that are trained with well-registered multi-spectral images, in contrast to earlier approaches using one classifier. The responses of corresponding features can be similarly characterized as being of the same class even though the intensities of the corresponding features are quite different. The experimental results show that our method provides good performance on multi-spectral image registration compared to current methods.

Virtual dimensionality (VD) was introduced as a definition of the number of spectrally distinct signatures in hyperspectral data where a method developed by Harsanyi-Farrand-Chang, referred to as HFC method was used to estimate the VD. Unfortunately, some controversial issues occur due to misinterpretation of the VD. Since the non-literal (spectral) information is the most important and critical for hyperspectral data to be preserved, the VD is particularly defined to address this issue as the number of spectrally distinct signatures present in the data where each spectral dimension is used to accommodate one specific signature. With this interpretation the VD is actually defined as the minimum number of spectral dimensions used to characterize the hyperspectral data. In addition, since hyperspectral targets of interest are generally insignificant and their occurrences have low probabilities with small populations, their contributions to 2nd order statistics are usually very limited. Consequently, the HFC method using eigenvalues to determine the VD may not be applicable for this purpose. Therefore, this paper revisits the VD and extends the HFC method to high-order statistics HFC method to estimate the VD for such a type of hyperspectral targets present in the data.

This paper proposes a new regularization algorithm combining the wavelet-based and contourlet-based regularization items based on the Compressive Sensing (CS) theorem. The new algorithm aims at gaining maximum benefit by combining the multiscale and multiresolution properties common to both wavelet and contourlet schemes, while simultaneously incorporating their individual properties of point singularity and line singularity respectively. CS is applied to remote sensing image deblurring. It has great practical significance due to saving the hardware cost and aiding fast transmission. Experimental results show the method achieves improvement in peak-signal-noise-ratio and correlation function as compared to traditional regularization algorithms.

Increasing the spatial resolution of panchromatic images and multispectral images is a classical problem in remote sensing. However, it is still in its infancy to spatially enhance the resolution of hyperspectral imageries. In this paper, we proposed a new method for increasing the spatial resolution of a hyperspectral data cube by using an iterative back projection (IBP) based method. Also, we developed a new metric to measure the visual quality of the enhanced images. This metric is good at measuring the visual quality of an image whose full-reference image is not available whereas the low spatial resolution image is available. Experimental results confirm the superiority of the proposed method.

Spectral unmixing is an important task for remotely sensed hyperspectral data exploitation. The spectral signatures collected in natural environments are invariably a mixture of the pure signatures of the various materials found within the spatial extent of the ground instantaneous field view of the imaging instrument. Spectral unmixing aims at inferring such pure spectral signatures, called endmembers, and the material fractions, called fractional abundances, at each pixel of the scene. A standard technique for spectral mixture analysis is linear spectral unmixing, which assumes that the collected spectra at the spectrometer can be expressed in the form of a linear combination of endmembers weighted by their corresponding abundances, expected to obey two constraints, i.e. all abundances should be non-negative, and the sum of abundances for a given pixel should be unity. Several techniques have been developed in the literature for unconstrained, partially constrained and fully constrained linear spectral unmixing, which can be computationally expensive (in particular, for complex highdimensional scenes with a high number of endmembers). In this paper, we develop new parallel implementations of unconstrained, partially constrained and fully constrained linear spectral unmixing algorithms. The implementations have been developed in programmable graphics processing units (GPUs), an exciting development in the field of commodity computing that fits very well the requirements of on-board data processing scenarios, in which low-weight and low-power integrated components are mandatory to reduce mission payload. Our experiments, conducted with a hyperspectral scene collected over the World Trade Center area in New York City, indicate that the proposed implementations provide relevant speedups over the corresponding serial versions in latest-generation Tesla C1060 GPU architectures.

This work aims to determine an efficient procedure (balanced between quality and compression ratio) for compressing multispectral remote sensing time series images in a 4-dimensional domain (2 spatial, 1 spectral and 1 temporal dimension). The main factors studied were: spectral and temporal aggregation, landscape type, compression ratio, cloud cover, thermal segregation and nodata regions. In this study, the authors used three-dimensional Discrete Wavelet Transform (3d-DWT) as the compression methodology, implemented in the Kakadu software with the JPEG2000 standard. This methodology was applied to a series of 2008 Landsat-5 TM images that covered three different landscapes, and to one scene (19-06-2007) from a hyperspectral CASI sensor. The results show that 3d-DWT significantly improves the quality of the results for the hyperspectral images; for example, it obtains the same quality as independently compressed images at a double compression ratio. The differences between the two compression methodologies are smaller in the Landsat spectral analysis than in the CASI analysis, and the results are more irregular depending on the factor analyzed. The time dimensional analysis for the Landsat series images shows that 3d-DWT does not improve on band-independent compression.

Aiming at the problems of the context-dilution and complicated context-quantization of high-order context, this paper proposes a new context arithmetic coding using weighted-based context modeling. By classifying the weights with non-uniform quantization, conventional high-order context-based arithmetic coding method can be approximated as low-order arithmetic coding. Compared with the existing high-order context modeling, the proposed method not only decreases the complexity of computation but also effectively improves the performance of entropy coding. Experimental results show that the algorithm with proposed weighted-based arithmetic coding method performs better than SPECK, SPIHT and JPEG2000.

The paper presents a context-based arithmetic coder's VLSI architecture used in SPIHT with reduced memory, which is used for high speed real-time applications. For hardware implementation, a dedicated context model is proposed for the coder. Each context can be processed in parallel and high speed operators are used for interval calculations. An embedded register array is used for cumulative frequency update. As a result, the coder can consume one symbol at each clock cycle. After FPGA synthesis and simulation, the throughput of our coder is comparable with those of similar hardware architectures used in ASIC technology. Especially, the memory capacity of the coder is smaller than those of corresponding systems.

This paper extends clustered differential pulse code modulation (C-DPCM) lossless compression method for hyperspectral images. In C-DPCM method the spectra of a hyperspectral image is clustered, and an optimized predictor is calculated for each cluster. Prediction is performed using a linear predictor. After prediction, the difference between the predicted and original values is computed. The difference is entropy-coded using an adaptive entropy coder for each cluster. The proposed use of adaptive prediction length is shown have lower bits/pixel value than the original C-DPCM method for new AVIRIS test images. Both calibrated are uncalibrated images showed improvement over fixed prediction length.

In multiple-input multiple-output(MIMO) wireless systems, combining good channel codes(e.g., Non-binary Repeat Accumulate codes) with adaptive turbo equalization is a good option to get better performance and lower complexity under Spatial Correlated Frequency Selective(SCFS) Channel. The key of this method is after joint antennas MMSE detection (JAD/MMSE) based on interruption cancelling using soft information, considering the detection result as an output of a Gaussian equivalent flat fading channel, and performing maximum likelihood detection(ML) to get more correct estimated result. But the using of ML brings great complexity increase, which is not allowed. In this paper, a low complexity method called list sphere decoding is introduced and applied to replace the ML in order to simplify the adaptive iterative turbo equalization system.

In order to improve the reconstructed quality of video sequence, a Random Linear Network Coding (RLNC) based video transmission scheme for Scalable Video Coding (SVC) is proposed in wireless broadcast scenario. A packetization model for SVC streaming is introduced to transmit the scalable bit streams conveniently, on the basis of which the RLNC based Unequal Error Protection (RUEP) method is proposed to improve the efficiency of video transmission. The RUEP's advantage lies in the fact that the redundancy protection of UEP can be efficiently determine by the capacity of broadcast channel. Simulation results show that RUEP can improve the reconstructed quality of video sequence compared with the traditional Store and Forward (SF) based transmission schemes.

Automatic target and anomaly detection are considered very important tasks for hyperspectral data exploitation. These techniques are now routinely applied in many application domains, including defence and intelligence, public safety, precision agriculture, geology, or forestry. Many of these applications require timely responses for swift decisions which depend upon high computing performance of algorithm analysis. However, with the recent explosion in the amount and dimensionality of hyperspectral imagery, this problem calls for the incorporation of parallel computing techniques. In the past, clusters of computers have offered an attractive solution for fast anomaly and target detection in hyperspectral data sets already transmitted to Earth. However, these systems are expensive and difficult to adapt to on-board data processing scenarios, in which low-weight and low-power integrated components are essential to reduce mission payload and obtain analysis results in (near) real-time, i.e., at the same time as the data is collected by the sensor. An exciting new development in the field of commodity computing is the emergence of commodity graphics processing units (GPUs), which can now bridge the gap towards on-board processing of remotely sensed hyperspectral data. In this paper, we describe several new GPU-based implementations of target and anomaly detection algorithms for hyperspectral data exploitation. The parallel algorithms are implemented on latest-generation Tesla C1060 GPU architectures, and quantitatively evaluated using hyperspectral data collected by NASA's AVIRIS system over the World Trade Center (WTC) in New York, five days after the terrorist attacks that collapsed the two main towers in the WTC complex.

Satellite remote sensing images can be interpreted to provide important information of large-scale natural resources, such as lands, oceans, mountains, rivers, forests and minerals for Earth observations. Recent advances of remote sensing technologies have improved the availability of satellite imagery in a wide range of applications including high dimensional remote sensing data sets (e.g. high spectral and high spatial resolution images). The information of high dimensional remote sensing images obtained by state-of-the-art sensor technologies can be identified more accurately than images acquired by conventional remote sensing techniques. However, due to its large volume of image data, it requires a huge amount of storages and computing time. In response, the computational complexity of data processing for high dimensional remote sensing data analysis will increase. Consequently, this paper proposes a novel classification algorithm based on semi-matroid structure, known as the parallel k-dimensional tree semi-matroid (PKTSM) classification, which adopts a new hybrid parallel approach to deal with high dimensional data sets. It is implemented by combining the message passing interface (MPI) library, the open multi-processing (OpenMP) application programming interface and the compute unified device architecture (CUDA) of graphics processing units (GPU) in a hybrid mode. The effectiveness of the proposed PKTSM is evaluated by using MODIS/ASTER airborne simulator (MASTER) images and airborne synthetic aperture radar (AIRSAR) images for land cover classification during the Pacrim II campaign. The experimental results demonstrated that the proposed hybrid PKTSM can significantly improve the performance in terms of both computational speed-up and classification accuracy.

In this paper, we present a new panchromatic sharpening method based on quality parameter optimization. Traditionally, quality metrics such as UIQI, CORR, and ERGAS have been used to assess the quality of panchromatic sharpening. Generally, HPF (high pass filtering) based panchromatic sharpening methods produce good performance. However, one problem with these methods is the peak noise that arises due to a small denominator value when the mean shift problem is addressed. In order to address this problem, we introduce an offset value that was optimized based on a quality metric. We assumed that the offset value was invariant with respect to the spatial scale, and it was used to enhance the resolution of the original multispectral images by using a high-resolution panchromatic image. The experimental results demonstrate that the proposed method showed better performance than some existing panchromatic sharpening methods.

The imagery resolution of imaging systems for remote sensing is often limited by image degradation resulting from unwanted motion disturbances of the platform during image exposures. Since the form of the platform vibration can be arbitrary, the lack of priori knowledge about the motion function (the PSF) suggests blind restoration approaches. A deblurring method which combines motion estimation and image deconvolution both for area-array and TDI remote sensing has been proposed in this paper. The image motion estimation is accomplished by an auxiliary high-speed detector and a sub-pixel correlation algorithm. The PSF is then reconstructed from estimated image motion vectors. Eventually, the clear image can be recovered by the Richardson-Lucy (RL) iterative deconvolution algorithm from the blurred image of the prime camera with the constructed PSF. The image deconvolution for the area-array detector is direct. While for the TDICCD detector, an integral distortion compensation step and a row-by-row deconvolution scheme are applied. Theoretical analyses and experimental results show that, the performance of the proposed concept is convincing. Blurred and distorted images can be properly recovered not only for visual observation, but also with significant objective evaluation increment.

The Consultative Committee for Space Data Systems (CCSDS) recommends the use of a two-stage strategy for lossless data compression in space. At the core of the second stage is the Rice coding method. The Rice compression ratio rapidly decreases in the presence of noise and outliers, since this coder is specially conceived for noiseless data following geometric distributions. This, in turn, makes the CCSDS recommendation too sensitive in front of outliers in the data, leading to non-optimal ratios in realistic scenarios. In this paper we propose to substitute the Rice coder of the CCSDS recommendation by a subexponential coder. We show that this solution offers high compression ratios even when large amounts of noise are present in the data. This is done by testing both compressors with synthetic and real data. The performance is actually similar to that obtained with the FAPEC coder, although with slightly higher processing requirements. Therefore, this solution appears as a simple improvement that can be done to the current CCSDS standard with an excellent return.

The FORMOSAT-5 is an optical remote sensing satellite with PAN 2m and MS 4m resolution, which is under development by the National Space Organization (NSPO) in Taiwan. The payload consists of one Panchromatic (PAN) band with 12,000 pixels and four Multi-Spectrum (MS) bands with 6000 pixels in the remote sensing instrument. The image data compression method complies with the Consultative Committee for Space Data Systems (CCSDS) standards^1,2. The compression ratio can be 1.5 for lossless compression, 3.75 or 7.5 for lossy compression. The Xilinx Virtex-5Q FPGA, XQR5VFX130 is used to achieve near real time compression.

Multicomponent data have become popular in several scientific fields such as forest monitoring, environmental studies, or sea water temperature detection. Nowadays, this multicomponent data can be collected more than one time per year for the same region. This generates different instances in time of multicomponent data, also called 4D-Data (1D Temporal + 1D Spectral + 2D Spatial). For multicomponent data, it is important to take into account inter-band redundancy to produce a more compact representation of the image by packing the energy into fewer number of bands, thus enabling a higher compression performance. The principal decorrelators used to compact the inter-band correlation redundancy are the Karhunen Loeve Transform (KLT) and Discrete Wavelet Transform (DWT). Because of the Temporal Dimension added, the inter-band redundancy among different multicomponent images is increased. In this paper we analyze the influence of the Temporal Dimension (TD) and the Spectral Dimension (SD) in 4D-Data in terms of coding performance for JPEG2000, because it has support to apply different decorrelation stages and transforms to the components through the different dimensions. We evaluate the influence to perform different decorrelators techniques to the different dimensions. Also we will assess the performance of the two main decorrelation techniques, KLT and DWT. Experimental results are provided, showing rate-distortion performances encoding 4D-Data using KLT and WT techniques to the different dimensions TD and SD.

In order to further improve the coding performance of Block-Based and Quadtree-Based Adaptive Loop Filter (BQ_ALF), the Fast Multi-Symmetry Adaptive Loop Filter Algorithm (FMS_ALF) is proposed. Firstly, this algorithm determines the optimal symmetry filter according to area symmetry and average sum of absolute difference. Then the filter areas are obtained through the block-based and quadtree-based method in I frame and through motion vector and Rate Distortion Optimization model A (RDO_A) in P or B frame .Finally the obtained areas are filtered by the optimal symmetry filter. Simulation results show that compared with BQ_ALF, the proposed algorithm reduces the coding time greatly while retains the reconstructed picture quality.

The Chinkuashih Benshen mine was the largest gold-copper mine in Taiwan before the owner had abandoned the mine in 1987. However, even the mine had been closed, the mineral still interacts with rain and underground water and flowed into the sea. The polluted sea surface had appeared yellow, green and even white color, and the pollutants had carried by the coast current. In this study, we used the optical satellite images to monitoring the sea surface. Several image processing algorithms are employed especial the subpixel technique and linear mixture model to estimate the concentration of pollutants. The change detection approach is also applied to track them. We also conduct the chemical analysis of the polluted water to provide the ground truth validation. By the correlation analysis between the satellite observation and the ground truth chemical analysis, an effective approach to monitoring water pollution could be established.

In this paper, we describe a new system for unsupervised and supervised classification of satellite images from Google Maps. The system has been developed using the SwingX-WS library, and incorporates functionalities such as unsupervised classification of image portions selected by the user (at the maximum zoom level) using ISODATA and k-Means, and supervised classification using the Minimum Distance and Maximum Likelihood, followed by spatial post-processing based on majority voting. Selected regions in the classified portion are used to train a maximum likelihood classifier able to map larger image areas in a manner transparent to the user. The system also retrieves areas containing regions similar to those already classified. An experimental validation of the proposed system has been conducted by comparing the obtained classification results with those provided by commercial software, such as the popular Research Systems ENVI package.

In this paper, we propose an unsupervised segmentation method for hyperspectral images using mean shift filtering. One major problem of traditional mean shift algorithms is the difficulty of determining kernel bandwidths. We address this problem by using efficient clustering methods. First, PCA (Principal Component Analysis) was applied to hyperspectral images and the first three eigenimages were selected. Then, we applied mean shift filtering to the selected images using a kernel with a small bandwidth. This procedure produced a large number of clusters. In order to merge the homogeneous clusters, we used the Bhattacharyya distance. Experiments showed promising segmentation results without requiring user input.

We develop a Graphics Processing Unit (GPU)-based high-performance RTTOV-7 forward model. The RTTOV forward model performs the fast computation of the radiances, brightness temperatures, overcast radiances, surface to space transmittances, surface emissivities and pressure level to space transmittances for a given profile vector. A special optimized high performance CUDA kernel was used for multi-profile processing. The difference between single-profile and multi-profile kernels is that in a multi-profile kernel each thread is responsible for computing the results for a single channel in several profiles. Multi-profile processing gave over two fold increase in processing speed compared to singleprofile processing. Using a GPU processing we reached promising speedups of 170x and 334x for single-profile and multi-profile processing, respectively. The significant 334x speedup means that the proposed GPU-based highperformance forward model is able to compute one day's amount of 1,296,000 Infrared Atmospheric Sounding Interferometer (IASI) spectra nearly within 12 minutes, whereas the original CPU-based version will impractically take nearly 3 days.

Line features are very important in photogrammetry, and compression often causes a loss of line features in images. In order to study the effects of compression on feature extraction, a SPOT5 image was selected as test data and the image was compressed with seven compression ratios. Features of the original and compressed images were extracted with the Canny operator. A change detection method was used to compare the features of the original and compressed images. The results show that the features extracted vary with the compression ratio, and the change ratio for those features increases with increasing compression ratio.

Estimating the number of spectral signal sources, denoted by p, in hyperspectral imagery is very challenging due to the fact that many unknown material substances can be uncovered by very high spectral resolution hyperspectral sensors. This paper investigates a recent approach, called maximum orthogonal complement algorithm (MOCA), for this purpose. The MOCA was originally developed by Kuybeda et al. for estimating the rank of a rare vector space in a highdimensional noisy data space. Interestingly, the idea of the MOCA is essentially derived from the automatic target generation process (ATGP) developed by Ren and Chang. By appropriately interpreting the MOCA in context of the ATGP a potentially useful technique, called maximum orthogonal subspace projection (MOSP) can be further developed where determining a stopping rule for the ATGP turns out to be equivalent to estimating the rank of a rare vector space by the MOCA and the number of targets determined by the stopping rule for the ATGP to generate is the desired value of the parameter p. Furthermore, a Neyman-Pearson detector version of MOCA, NPD-MOCA can be also derived by the MOSP as opposed to the MOCA considered as a Bayes detector. Surprisingly, the MOCA-NPD has very similar design rationale to that of a technique referred to as Harsanyi-Farrand-Chang method that was developed to estimate the virtual dimensionality (VD) which is defined as the p.

The Global Positioning System (GPS) has long been used as a scientific tool, and it has turned into a very powerful technique in domains like geophysics, where it is commonly used to study the dynamics of a large variety of systems, like glaciers, tectonic plates and others. In these cases, the large distances between receivers as well as their remote locations usually pose a challenge for data transmission. The standard format for scientific applications is a compressed RINEX file - a raw data format which allows post-processing. Its associated compression algorithm is based on a pre-processing stage followed by a commercial data compressor. In this paper we present a new compression method which can achieve better compression ratios with a faster operation. We have improved the pre-compression stage, split the resulting file into two, and applied the most appropriate compressor to each file. FAPEC, a highly resilient entropy coder, is applied to the observables file. The results obtained so far demonstrate that it is possible to obtain average compression gains of about 35% with respect to the original compressor.

An efficient compression algorithm for hyperspectral imagery based on compressive sensing and interband linear prediction is proposed which has the advantages of high compression performance and low computational complexity by exploiting the strong spectral correlation. At the encoder, the random measurements of each frame are made, quantized and transmitted to the decoder independently. The prediction parameters between adjacent bands are also estimated using the linear prediction algorithm and transmitted to the decoder. At the decoder, a new reconstruction algorithm with the proposed initialization and stopping criterion is employed to reconstruct the current frames with the assistance of the prediction frame, which is derived from the previous reconstructed neighboring frames and the received prediction parameters using the same prediction algorithm. Experimental results show that the proposed algorithm not only obtains about 1.1 dB gains but greatly decreases decoding complexity. Furthermore, our algorithm has the characteristics of low-complexity encoding and facility in hardware implementation.

Recently there is a boom on the use of graphic processor units (GPU) for speedup of scientific computations. By identifying the time dominant portions of the code that can be executed in parallel, significant speedup can be achieved by a GPU-based implementation. For the voluminous ultraspectral sounder data, lossless compression is desirable to save storage space and transmission time without losing precision in retrieval of geophysical parameters. Predictive partitioned vector quantization (PPVQ) has been proven to be an effective lossless compression scheme for ultraspectral sounder data. It consists of linear prediction, bit partition, vector quantization, and entropy coding. Two most time consuming stages of linear prediction and vector quantization are chosen for GPU-based implementation. By exploiting the data parallel characteristics of these two stages, a speedup of 42x has been achieved in our GPU-based implementation of the PPVQ compression scheme.

Hyperspectral images are of very large data size and highly correlated in neighboring bands, therefore, it is necessary to realize the efficient compression performance on the condition of low encoding complexity. In this paper, we propose a method based on both partitioning embedded block and lossless adaptive-distributed arithmetic coding (LADAC). Combined with three-dimensional wavelet transform and SW-SPECK algorithm, LADAC is adopted according to the correlation between the adjacent bit-plane. Experimental results show that our proposed algorithm outperforms 3D-SPECK, furthermore, our method need not take the inter-band prediction or transform into account, so the complexity is small relatively.