We propose a novel approach for face verification by encoding 2D and 3D face images as a high order tensor. To perform tensor dimensionality reduction for both the unsupervised and supervised cases, we propose multilinear whitened principal component analysis (MWPCA) and tensor exponential discriminant analysis (TEDA), respectively. MWPCA is utilized to solve the small sample size problem in the high-dimensional space and to improve the discrimination power achieved by classical MPCA. In the supervised case, we extend multilinear discriminant analysis to TEDA in order to emphasize the discriminant data included in the null space of the within-class scatter matrix of each tensor's mode. Additionally, TEDA enlarges the margin between samples belonging to different classes via distance diffusion mappings. Our proposed approach can be seen as a novel data fusion method based on tensor representation. Indeed, the histograms of different local descriptors extracted from both 2D and 3D face modalities are combined through different tensor modes. The extensive experimental evaluation carried out on FRGC v2.0, Bosphorus, and CASIA 2D and 3D face databases indicates that the proposed approach performs significantly better than the state-of-the-art approaches.

There are two main contributions in this paper. Firstly, we estimate the rank for the truncation of the Parafac decomposition in an optimal sense. For this, we propose a least squares criterion and justify the choice of the fast Nelder-Mead method to minimize this criterion. Secondly, we combine the truncation of the Parafac decomposition with multidimensional wavelet packet transform. A single rank value is estimated for each decomposition level, which simplifies the implementation. We exemplify the proposed method with an application to multispectral image denoising: we study the performance of the proposed method based on Parafac decomposition, compared to ForWaRD.

Positron Emission Tomography (PET) is a 3-D functional imaging modality which help physicians to diagnose neurodegenerative diseases like Alzheimer's Disease (AD). Computer-aided detection and diagnosis, based on medical imaging techniques is of importance for a quantitative evaluation. A novel method of ranking the effectiveness of brain regions to separate AD from healthy brains images is presented. Brain images are first mapped into 116 anatomical regions of interest. The first four moments and the entropy of the histograms of these regions are computed. Receiver Operating Characteristics curves are then used to rank the ability of regions to separate PET brain images. Twenty one selected regions are input to both Support Vector Machine and Random Forest classifiers and evaluation is done on 142 brain PET images. Classification results are better than those obtained when using the whole 116 initial regions or when inputting the whole brain voxels. In addition, an important computational time reduction was obtained.

We propose an analytical approach to evaluate the performance of space‐diversity free‐space optical (FSO) systems over correlated Gamma–Gamma (ΓΓ) fading channels. To do this, we firstly derive an infinite series representation for the moment generating function (MGF) of the sum of arbitrarily correlated ΓΓ random variables (RVs). Then, a closed‐form approximate expression for the probability density function (PDF) of the sum of ΓΓ RVs is obtained from the MGF with the aid of Padé approximation. The performance of the FSO system is then evaluated based on this PDF. Through some numerical results we illustrate the accuracy of the proposed method and further discuss its limitations.

In amplify‐and‐forward wireless relay networks, channel state information is required at the destination for signal detection when coherent distributed space‐time coding is performed at the relays. The authors consider in this study signal detection in the case of imperfect partial channel knowledge at the destination. For the case of maximum a posteriori signal detection, they propose a modified detector formulation that takes the variance of channel estimation errors into account, and reduces the impact of channel uncertainty on the receiver performance. They show that the proposed modification permits a significant performance improvement, while increasing only slightly the receiver's complexity. It is of special interest for non‐constant modulus signal constellations of relatively large size, as well as for relatively low pilot powers.

Tensor-decomposition-based methods for reducing random noise components in hyperspectral images (HSIs), both dependent and independent from signal, are proposed. In this paper, noise is described by a parametric model that accounts for the dependence of noise variance on the signal. This model is thus suitable for the cases where photon noise is dominant compared with the electronic noise contribution. To denoise HSIs distorted by both signal-dependent (SD) and signal-independent (SI) noise, some hybrid methods, which reduce noise by two steps according to the different statistical properties of those two types of noise, are proposed in this paper. The first one, named as the PARAFACSI- PARAFACSD method, uses a multilinear algebra model, i.e., parallel factor analysis (PARAFAC) decomposition, twice to remove SI and SD noise, respectively. The second one is a combination of the well-known multiple-linear-regression-based approach termed as the HYperspectral Noise Estimation (HYNE) method and PARAFAC decomposition, which is named as the HYNE-PARAFAC method. The last one combines the multidimensional Wiener filter (MWF) method and PARAFAC decomposition and is named as the MWF-PARAFAC method. For HSIs distorted by both SD and SI noise, first, most of the SI noise is removed from the original image by PARAFAC decomposition, the HYNE method, or the MWF method based on the statistical property of SI noise; then, the residual SD components can be further reduced by PARAFAC decomposition due to its own statistical property. The performances of the proposed methods are assessed on simulated HSIs. The results on the real-world airborne HSI Hyperspectral Digital Imagery Collection Experiment (HYDICE) are also presented and analyzed. These experiments show that it is worth taking into account noise signal-dependence hypothesis for processing HYDICE data.

In this paper, we consider an issue of hand posture classification. We improve a recently proposed signature, a matrix containing the distance of all contour pixels to an arbitrary reference point. Adequate pre-processings ensure the invariance properties of the signature. Candidate postures are pre-selected with a surface criterion, and Principal Component Analysis (PCA) reduces the dimensionality of the data, which improves the classification process.

A novel mixture model with spatial constraint is proposed for image segmentation. This model assumes that the pixel label prior probabilities are similar if the pixels are geometric close. An energy function is defined on the spatial space for measuring the spatial information. We also derive an energy function on the observed data space from the log-likelihood function of the standard mixture model. We estimate the model parameters and posterior probability by minimizing the combination of the two energy functions, using the gradient descent algorithm. Numerical experiments are presented where the proposed method is tested on synthetic and real world images. These experimental results demonstrate that the proposed method achieves competitive performance compared to spatially variant finite mixture model.

Denoising is an important preprocessing step for several applications in the hyperspectral imaging (HSI) domain, such as classification and target detection, to achieve good performances. Because the signal-dependent photonic noise has become as dominant as the signal-independent noise generated by the electronic circuitry in HSI data collected by new-generation hyperspectral sensors, the reduction of the additive signal-dependent photonic noise becomes the focus of the current research in this field. To reduce the optoelectronic noise from HSIs, a new method is developed in this paper. First, a prewhitening procedure is proposed to whiten noise in HSIs. Second, a multidimensional wavelet packet transform (MWPT) in tensor form is presented to find different component tensors of the HSI. Then, to jointly filter a component tensor in each mode, a multiway Wiener filter is introduced. Moreover, to determine the best transform level and basis of the MWPT, a risk function is proposed. The effectiveness of our method in denoising and classification is experimentally demonstrated on a real-world HSI acquired by an airborne sensor.

In this paper, we propose an automatic tracking method to extract blood vessels in retinal images. Seed points are firstly picked out on a retinal image for initialization. Our algorithm detects vessel edge points iteratively based on a statistical sampling model using a Bayesian method. At a given step, local vessel's sectional intensity profile is approximated by a Gaussian model. New vessel edge points are detected by using local grey level statistics and expected vessel structures. For evaluation purpose, we use the STARE public database. Experiments results show effective detection of blood vessels when using the proposed method.

The noise in the acquired hyperspectral image (HSI) is generally assumed as additive white Gaussian noise (WGN), while the estimation of the noise in each band of the real-world HSI shows that the noise is not white. Reducing the additive non-white noise is an important preprocessing step to further analyze the information in the HSIs by classification. A PMWF method, prewhitening and MWF (multidimensional Wiener filtering), was suggested and the non-white noise could be whitened by a pre-whitening procedure. While this method is time-consuming due to both the pre-whitening step and the estimation of three ranks of the MWF method. In this paper, we introduce a powerful multilinear algebra model, named parallel factor analysis (PARAFAC), which has only one rank and need not the pre-whitening procedure. To improve the classification, the rank of PARAFAC decomposition is estimated according to the maximum of the overall accuracy (OA) of the support vector machine (SVM) classification. The experiment results show that the PARAFAC model has high efficiency in the reduction of non-white noise and is a preferable preprocessing method for the accuracy improvement of the SVM classification.

Noise reduction is an important preprocessing step to analyze the information in the hyperspectral image (HSI). Because the common filtering methods for HSIs are based on the data vectorization or matricization while ignoring the related information between image planes, there are new approaches considering multidimensional data as whole entities, for example, multidimensional Wiener filtering (MWF) based on Tucker3 tensor decomposition. However, if HSIs are not disturbed by white noise, MWF cannot effectively remove the nonwhite noise and obtain the expected signal. To reduce nonwhite noise from HSIs, a new method is proposed in this letter. The first step of this method is to whiten the noise in HSIs through a prewhitening procedure. Then, MWF can help to denoise the prewhitened data. At last, an inverse prewhitening process can rebuild the estimated signal. Comparative studies with existing denoising methods show that the proposed approach has promising prospects in this field.

Denoising is an important preprocessing step to further analyze a hyperspectral image (HSI). The common denoising methods, such as 2-D (two dimensions) filter, actually rearrange data from 3-D to 2-D and ignore the spectral relationships among different bands of the image. Tensor decomposition method has been adapted to denoise HSIs, for instance Tucker3 (three-mode factor analysis) model. However, this model has problems in uniqueness of decomposition and in estimation of multiple ranks. In this paper, to overcome these problems, we exploit a powerful multilinear algebra model, named parallel factor analysis (PARAFAC), and the number of estimated rank is reduced to one. Assumed that HSI is disturbed by white Gaussian noise, the optimal rank of PARAFAC is estimated according to that the covariance matrix of the$n$-mode unfolding matrix of the removed noise should be approach to a scalar matrix. Then, the denoising results by PARAFAC decomposition are presented and compared with those obtained by Tucker3 model and 2-D filters. To further verify the denoising performance of PARAFAC decomposition, Cramer–Rao lower bound (CRLB) of denoising is deduced theoretically for the first time, and the experiment results show that the PARAFAC model is a preferable denoising method since the variance of the HSI denoised by it is closer to the CRLB than by other considered methods.

Target detection is a key issue in processing hyperspectral images (HSIs). Spectral-identification-based algorithms are sensitive to spectral variability and noise in acquisition. In most cases, both the target spatial distributions and the spectral signatures are unknown, so each pixel is separately tested and appears as a target when it significantly differs from the background. In this paper, we propose two algorithms to improve the signal-to-noise ratio (SNR) of hyperspectral data, leading to detectors that are robust to noise. These algorithms consist in integrating adaptive spatial/spectral filtering into the adaptive matched filter and adaptive coherence estimator. Considering the HSIs as tensor data, our approach introduces a data representation involving multidimensional processing. It combines the advantages of spatial and spectral information using an alternating least square algorithm. To estimate the signal subspace dimension in each spatial mode, we extend the Akaike information criterion, and we develop an iterative algorithm for spectral-mode rank estimation. We demonstrate the interest of integrating the quadtree decomposition to perform an adaptive 3-D filtering and thereby preserve the local image characteristics. This leads to a significant improvement in terms of denoised tensor SNR and, consequently, in terms of detection probability. The performance of our method is exemplified using simulated and real-world HYperspectral Digital Imagery Collection Experiment images.

This paper deals with the development of a new texture analysis method based on both spatial and spectral information for texture classification purposes. The idea of the Spatial and Spectral Gray Level Dependence Method (SSGLDM) is to extend the concept of spatial gray level dependence method by assuming texture joint information between spectral bands. In addition, new texture features measurement related to (SSGLDM) which define the image properties have been also proposed. Extensive experiments have been carried out on many multi-spectral images for use in prostate cancer diagnosis and quantitative results showed the efficiency of this method compared to the Gray Level Co-occurrence Matrix (GLCM). The results indicate a significant improvement in classification accuracy.

A nondestructive control system is proposed. We consider an application of defect detection in metal workpieces. To solve this defect detection problem, we propose an image processing system including denoising, segmentation, and mapping. A contour preserving denoising method based on wavelet decomposition is adapted; various tools of contour based image segmentation are combined adequately, to segment successively high contrast and low contrast contours; generalized Hough transform is adapted to map processed and reference workpieces without defects. Comparative results are proposed which involve the proposed image processing system, and two region based segmentation methods.

Vessel detection is an important process in many medical imaging applications. In this paper, an edge tracking scheme is proposed for the detection of blood vessels in retinal images. This method detects edge points iteratively based on a Bayesian approach using local grey levels statistics and continuity properties of blood vessels. Combining the grey level profile and vessel geometric properties improves the accuracy and robustness of the tracking process. Experiments on both synthetic and real retinal images show promising results.

Hyperspectral image (HSI) classification requires spectral dimensionality reduction (DR) and spatial filtering. While common DR and denoising methods use linear algebra, we propose a tensorial method to jointly achieve denoising and DR. Tensorial processing models HSI data as a whole entity to treat jointly spatial and spectral modes. A multidimensional Wiener filter (MWF) was successfully applied to denoise multiway data such as color images. First, we adapt the quadtree decomposition to tensor data in order to take into account the local image characteristics in the MWF. We demonstrate that this novel version of the filter called adaptive multidimensional Wiener filtering (AMWF)-(K1, K2, K3) performs well as a denoising preprocessing to improve classification results. Then, we propose a novel method, referred to as AMWFdr-(K1, K2, D3) which performs both spatial filtering and spectral DR. Support vector machine is applied to the output of four-dimensionality and noise-reduction methods to compare their efficiency: the proposed AMWFdr-(K1, K2, D3), PCAdr, ICAdr, MNFdr, and DWTdrassociated with Wiener filtering.

Image capturing and image content description can be regarded as the two major steps of a computer vision process. This paper focuses on both within the field of specular surface inspection, by generalizing a previously defined stripebased inspection method to free-form surfaces on the basis of a specific stripe illumination technique and by outlining a general feature-based stripe image characterization approach by means of new theoretical concepts. One major purpose of this paper is to propose a general stripe image interpretation approach on the basis of a three-step procedure: 1) comparison of different image content description techniques, 2) fusion of the most appropriate ones, and 3) selection of the optimal features. It is shown that this approach leads to an increase in the classification rates of more than 2 percent between the initial fused set and the selected one. The new contributions encompass 1) the generalization of a cylindrical specular surface enhancement technique to more complex specular geometries, 2) the generalization of the previously defined stripe image description by using the same number of features for the bright and the dark stripes, and 3) the definition of an optimal, in terms of classification rates and computational costs, stripe feature set.

Linear structures such as blood vessels in medical images are important features for computer-aided diagnosis and follow-up of many diseases. In this letter a new tracking-based segmentation method is proposed to detect blood vessels in retinal images. Bayesian segmentation with the Maximuma posteriori(MAP) Probability criterion is used for that purpose. Tests on simulated and retinal images are presented and compared with a vessel detection technique. Our method performs better results.

Linear structures such as blood vessels in medical images are important features for computer-aided diagnosis and follow-up of many diseases. In this letter a new tracking-based segmentation method is proposed to detect blood vessels in retinal angiorams. Bayesian segmentation with the maximum a posteriori (MAP) probability criterion is used for that purpose. First promising results are presented and discussed.

Subspace-based methods rely on dominant element selection from second order statistics. They have been extended to tensor processing, in particular to tensor data filtering. For this, the processed tensor is flattened along each mode successively, and singular value decomposition of the flattened matrix is classically performed. Data projection on the dominant singular vectors results in noise reduction. The numerical cost of SVD is elevated. Now, tensor processing methods include an ALS (Alternating Least Squares) loop, which implies that a large number of SVDs are performed. Fixed point algorithm estimates an a priori fixed number of singular vectors from a matrix. In this paper, we generalize fixed point algorithm as a higher-order fixed point algorithm to the estimation of only the required dominant singular vectors in a tensor processing framework. We compare the proposed method in terms of denoising quality and speed through an application to color image and hyperspectral image denoising.

Dimensionality reduction (DR) is the key issue to improve the classifiers' efficiency for hyperspectral images (HSIs). In this paper, principal component analysis (PCA), independent component analysis, and projection pursuit (PP) approaches to DR have been investigated. These matrix-algebra methods are applied on vectorized images. Thereof, the spatial rearrangement is lost. To jointly take advantage of the spatial and spectral information, HSI has been recently represented as tensor. Offering multiple ways to decompose data orthogonally, we introduced DR methods based on multilinear-algebra tools. The DR is performed on spectral way using PCA, or PP, joint to an orthogonal projection onto a lower subspace dimension of the spatial ways. We show the classification improvement using the introduced methods in function to existing methods. This experiment is exemplified using real-world HYDICE data.

In this paper we propose a new multidimensional filtering method based on fourth order cumulants to denoise of data tensor impaired by correlated Gaussian noise. We overview the multidimensional Wiener filtering that overcomes the well known lower rank-(K1,..., KN) tensor approximation. But this method only exploits second order statistics. In some applications, it may be interesting to consider a correlated Gaussian noise. Then, we propose to introduce the fourth order statistics in the denoising algorithm. Indeed, the use of fourth order cumulants enables to remove the Gaussian components of an additive noise. Qualitative results of the improved multidimensional Wiener filtering are shown for the case of noise reduction in hyperspectral imagery.

Subspace-based methods rely on the selection of leading eigenvectors, associated with dominant eigenvalues. They have been extended to tensor data processing, such as denoising. Usually EVD (eigenvalue decomposition) is performed and data projection on leading eigenvectors results in noise reduction. Tensor processing methods, in particular multiway Wiener filtering algorithm, include an ALS (alternating least squares) loop, which involves several EVDs. Fixed point algorithm is a faster method than EVD to estimate a fixed number of eigenvectors. In this paper, we adapt fixed point algorithm to the estimation of only the required leading eigenvectors in a tensor processing framework. We adapt inverse power method to estimate the required noise variance. We provide a comparative study in terms of speed through an application to hyperspectral image denoising.

Dimensionality reduction (DR) is a major issue to improve the efficiency of the classifiers in Hyperspectral images (HSI). Recently, the independent component analysis (ICA) approach to DR has been investigated. But, this signal processing is applied on vectorized images, losing spatial rearrangement. To jointly take advantage of the spatial and spectral information, HSI has been recently represented as tensor. Offering multiple ways to decompose data orthogonally, we develop a new DR method based on multilinear algebra tools and on ICA. The DR is performed on spectral way using ICA jointly to an orthogonal projection onto a lower subspace dimension of the spatial ways. We show the Maximum Likelihood classifier improvement using the proposed method.

A generalized multidimensional Wiener filter for denoising is adapted to hyperspectral images (HSIs). multidimensional wiener filtering (MWF) uses the signal subspace of eachn-mode flattening matrix of the HSI, which is a third-order tensor. However, in the HSI case, then-mode ranks are close to then-mode dimensions. Thus, the signal subspace dimension can be underestimated. This leads to a loss of spatial resolution-edge blurring-and artifacts in the restored HSI. To cope with the underestimation while preserving edges, a new method is proposed. It estimates the relevant directions of flattening that may not be parallel to HSI dimensions. We adapt the bidimensional straight line detection algorithm that estimates the HSI main directions, which are used to flatten the HSI tensor. We also generalize the quadtree decomposition to tensors in order to adapt the filtering to the local image characteristics. Comparative studies with MWF, principal component analysis-stationary wavelet transform, and channel-by-channel Wiener filtering show that our algorithm provides better performance while restoring impaired HYDICE HSIs.

In hyperspectral image (HSI) analysis, classification requires spectral dimensionality reduction (DR). While common DR methods use linear algebra, we propose a multilinear algebra method to jointly achieve denoising reduction and DR. Multilinear tools consider HSI data as a whole by processing jointly spatial and spectral ways. The lower rank-(K1, K2, K3) tensor approximation [LRTA-(K1, K2, K3)] was successfully applied to denoise multiway data such as color images. First, we demonstrate that the LRTA-(K1, K2, K3) performs well as a denoising preprocessing to improve classification results. Then, we propose a novel method, referred to as LRTAdr-(K1, K2, D3), which performs both spatial lower rank approximation and spectral DR. The classification algorithm Spectral Angle Mapper is applied to the output of the following three DR and noise reduction methods to compare their efficiency: the proposed LRTAdr-(K1, K2, D3), PCAdr, and PCAdrassociated with Wiener filtering or soft shrinkage of wavelet transform coefficients.

Whereas most previous works treating color or hyper- spectral image restoration use hybrid filters or data splitting, some new approaches consider multidimensional or tensor signal processing techniques. Tensor processing methods are based on multilinear algebra and are more efficient than 2-D filtering. However, they rely on orthogonal tensor flattening. The aim of this letter is to show that this orthogonal flattening may not be optimal for multidimensional images. We propose a method to adapt the flattening depending on the data set. Our proposed method is based on the estimation of main directions in multidimensional data. For this purpose, we extend the straight line detection algorithm. Multidimensional filtering method HOSVD - (K1,..., KN) is applied along the estimated directions. We also adapt a quadtree partitioning in order to split tensors into homogeneous sub-tensors to keep local characteristics. Considering some examples of color and hyperspectral images, we present some promising results.

A generalized multidimensional Wiener filter for denoising is adapted to hyperspectral images (HSIs). Commonly, multidimensional data filtering is based on data vectorization or matricization. Few new approaches have been proposed to deal with multidimensional data. Multidimensional Wiener filtering (MWF) is one of these techniques. It considers a multidimensional data set as a third-order tensor. It also relies on the separability between a signal subspace and a noise subspace. Using multilinear algebra, MWF needs to flatten the tensor. However, flattening is always orthogonally performed, which may not be adapted to data. In fact, as a Tucker-based filtering, MWF only considers the useful signal subspace. When the signal subspace and the noise subspace are very close, it is difficult to extract all the useful information. This may lead to artifacts and loss of spatial resolution in the restored HSI. Our proposed method estimates the relevant directions of tensor flattening that may not be parallel either to rows or columns. When rearranging data so that flattening can be performed in the estimated directions, the signal subspace dimension is reduced, and the signal-to-noise ratio is improved. We adapt the bidimensional straight-line detection algorithm that estimates the HSI main directions, which are used to flatten the HSI tensor. We also generalize the quadtree partitioning to tensors in order to adapt the filtering to the image discontinuities. Comparative studies with MWF, wavelet thresholding, and channel-by-channel Wiener filtering show that our algorithm provides better performance while restoring impaired HYDICE HSIs.

Detection and classification are key issues in processing hyperspectral images (HSIs). Spectral-identification-based algorithms are sensitive to spectral variability and noise in acquisition. In this paper, we propose two detection algorithms that are robust to noise. These algorithms consist in integrating spatial/spectral filtering into the adaptive matched filter and adaptive coherence/cosine estimator. Considering the HSI as tensor data, our approach introduces a data representation involving multilinear algebra. It combines the advantages of spatial and spectral information using an alternating least squares algorithm. To estimate the signal subspace dimension in each mode, we extended the Akaike information criterion and the minimum description length criterion. We demonstrate that integrating a multiway restoration leads to significant improvement of the detection probability. The performance of our method is exemplified using simulated and real-world Hyperspectral Digital Imagery Collection Experiment images.

Thanks to a specific formalism for signal generation, it is possible to transpose an image processing problem to an array processing problem. For straight line characterization, the existing method subspace-based line detection (SLIDE) works on virtual signals generated on a linear antenna. In this paper we propose to retrieve circular and nearly circular contours in images. We propose a novel method for radius estimation, and we extend the estimation of circles to the retrieval of circular-like distorted contours. For this purpose we develop a new model for virtual signal generation: we simulate a circular antenna, so that a high resolution method can be employed for radius estimation. An application to biomedical imaging is proposed.

Target detection in hyperspectral images (HSI) in one of the most common applications. But the classical detection algorithms are sensitive to noise. It is crucial to well restore the spectral signature in order to decrease the noise dependence of the detection algorithm. In this paper, we propose a restoration method which takes advantage of spatial and spectral information in order to estimate the spectral signature without impair the discriminate power. Our method is based on tensor decomposition where all ways are processed simultaneously. By considering the cross-dependency of spatial and spectral information for the filtering, we improve the probability of detection. Our optimization criterion is the minimization of the mean square error between the estimated and the desired tensors. This minimization leads to estimate the n-mode filter for each way and are jointly estimated by using an Alternating Least Squares (ALS) algorithm. Comparative studies with the classical bidimensional restoration methods show that our algorithm exhibits better detection probability in noisy situation. Indeed, the detection probability obtained after our algorithm is higher than 0.7 until a signal to noise ratio equal to -3 dB.

We propose a semi-blind channel estimator based on superimposed pilots for the case of single-carrier single-user MIMO systems. The proposed estimator is based on the expectation maximization (EM) algorithm. Numerical results show that even with this semi-blind estimator, the classical approach of time-multiplexed pilots still outperforms the superimposed approach, except for a large number of receive antennas and not too short frames.

A noise reduction scheme on digitized mammographic phantom images is presented. This algorithm is based on a direct contrast modification method with an optimal function, obtained by using the mean squared error as a criterion. Computer simulated images containing objects similar to those observed in the phantom are built to evaluate the performance of the algorithm. Noise reduction results obtained on both simulated and real phantom images show that the developed method may be considered as a good preprocessing step from the point of view of automating phantom film evaluation by means of image processing.

An important cause of performance loss in source localization in underwater acoustics is that towed flexible antennas deviate from the assumed rectilinear shape. In this work, the localization of sources in presence of phase errors is studied. Cancellation of phase errors is necessary to solve the source-localization problem. A previous work led to interesting results for antennas composed of a few sensors. We propose here a novel algorithm which is adapted to the antennas composed of a large number of sensors, keeping a small computational load. Our method is based on an orthogonality property between signal and noise subspaces and a novel optimization method: the robust DIviding RECTangles algorithm accelerated by spline interpolation. The performances of the proposed method are illustrated by applying it to the characterization of three sources.

Circular features are commonly sought in digital image processing. The subspace-based line detection (SLIDE) method proposed to estimate the center and the radius of a single circle. In this paper, we introduce a novel method for estimating several radii while extending the circle estimation to retrieve circular-like distorted contours. Particularly, we develop and validate a new model for virtual signal generation by simulating a circular antenna. The circle center is estimated by the SLIDE method. A variable speed propagation scheme toward the circular antenna yields a linear phase signal. Therefore, a high-resolution method provides the radius. Either the gradient method or the more robust combination of dividing rectangles and spline interpolation can extend this method extend this method for free form object segmentation. The retrieval of multiple non concentric circles and rotated ellipses is also considered. To evaluate the performance of the proposed methods, we compare them with a least-squares method, Hough transform, and gradient vector flow. We apply the proposed method to hand-made images while considering some real-world images.

Distorted curves retrieval is faced for robotic way screening, particle trajectory characterization, aerial and satellite image analysis. This image processing problem has been transposed to an array processing problem by adopting specific conventions. Some solutions for wavefront distortions canceling have already been proposed. In this paper we aim at improving an existing method for distorted curves retrieval, making use of a global optimization algorithm. We show that it is possible to combine an optimization method to an interpolation method in order to obtain a reliable and fast algorithm

This study deals with detection and localization of buried objects. We developed a novel method, that extends the subspace method, to estimate the bearing and the range of buried objects with known shapes such as spheres or cylinders. The idea is to use the spatial complexities of the scattered field to form an acoustic model that we used instead of the plane wave model used in the subspace method. The received signals are wideband and correlated, thus we apply a frequential smoothing in order to decorrelate them. We applied this method first, on cylindrical shells and we obtained a satisfying result (Z. Saidi and S. Bourennane, 2005) and here we extend it to spherical shells. Performances of the proposed method are investigated on experimental data recorded during underwater acoustic experiments

This work deals with unsupervised change detection in temporal sets of synthetic aperture radar (SAR) images. We focus on one of the most widely used change detector in the SAR context, the so-called log-ratio. In order to deal with the classification issue, we propose to use a new fuzzy version of hidden Markov chains (HMCs), and thus to address fuzzy change detection with a statistical approach. The main characteristic of the proposed model is to simultaneously use Dirac and Lebesgue measures at the class chain level. This allows the coexistence of hard pixels (obtained with the classical HMC segmentation) and fuzzy pixels (obtained with the fuzzy measure) in the same image. The quality assessment of the proposed method is achieved with several bidate sets of simulated images, and comparisons with classical HMC are also provided. Experimental results on real European Remote Sensing 2 Precision Image (ERS-2 PRI) images confirm the effectiveness of the proposed approach.

This paper deals with unsupervised segmentation of multicomponent images. In order to address the classification issue, we propose to use a new rectorial fuzzy version of hidden Markov chains (HMC). The main characteristic of the proposed model is to simultaneously use Dirac and Lebesgue measures at the class chain level. It then allows the coexistence of crisp pixels (obtained with the uncertainty measure of the model) and fuzzy pixels (obtained with the fuzzy measure of the model) in the same image. Crisp and fuzzy multidimensional densities can then be estimated in the segmentation process, according to the assumption considered to model the statistical links between the layers of the multiband image. The efficiency of the proposed method is illustrated with a multiscale decomposition of a synthetic aperture radar (SAR) image and comparisons with one-dimensional fuzzy HMC are also provided. The segmentation results show the interest of the new method

The high resolution methods of array processing lead to an improvement of the results obtained for source localization. By adopting specific conventions, it is possible to employ high resolution methods to characterize straight lines in an image. In this paper, we propose an original method that leads to the estimation of the parameter "offset" of the straight lines. The proposed method is fast and effective compared to an existing method. An extension to non rectilinear contours is developed.

Matched and adaptive subspace detectors apply to a wide range of problems in radar, sonar, and data communication, where the signal is constrained to lie in a multidimensional linear subspace. These detectors generalize known results in matched and adaptive detection theory. In this paper we propose an original approach to anomaly detection based on whitening and spatial correlation filtering (WSCF). The performance is investigated in terms of the detection probability, and the false alarm ratio. A comparison permits us to show how this new method can outperform the well-known Reed and Xiaoli Yu (RX) algorithm.

The main idea of the paper is to replace the covariance matrix involved in TUCKALS3 algorithm with the fourth order cumulant slice matrix of the data tensor n-mode vectors in order to eliminate the Gaussian components of the additive noise. The good qualitative results of this new multi-mode filtering method are shown in the case of correlated noise reduction in a color image and a polarized seismic wave, modelled by third order tensors.

First order hidden Markov models have been used for a long time in image processing, especially in image segmentation. In this paper, we propose a technique for the unsupervised segmentation of images, based on high-order hidden Markov chains. We also show that it is possible to relax the classical hypothesis regarding the state observation probability density, which allows to take into account some particular correlated noise. Model parameter estimation is performed from an extension of the general iterative conditional estimation (ICE) method that takes into account the order of the chain. A comparative study conducted on a simulated image is carried out according to the order of the chain. Experimental results on synthetic aperture radar (SAR) images show that the new approach can provide a more homogeneous segmentation than the classical one, implying higher algorithm complexity and computation time.

This article shows that image filtering based on SVD favors the denoising in the line (horizontal) and column (vertical) direction since the matrix SVD is equivalent to a simultaneous line and column vector principal component analysis (PCA). It also proposes a simple algorithm based on PCA-filtering processed on a rotated image such that the principal directions becomes vertical or horizontal. It brings an important improvement in the final image quality since the filtering in every image direction is improved.

The paper presents a new fast optimal lower rank tensor approximation (FOLRTA) method for lower rank-(R/sub 1/,..., R/sub N/) tensor approximation applied to multidimensional signal processing. It is founded on a new approach which consists of considering multidimensional data as global tensors instead of splitting them into matrices or vectors for later classical second order array processing. Its basic principle is to project the initial data tensor into the signal subspace, in each consecutive mode. The developed method is the first analytical solution to the Tucker3 tensor decomposition. We show in a simple example of noise reduction of a color image the efficiency of this method. It can also be applied in seismic, acoustics or multimedia signal processing.

In this work, the signal processing methods are used for determining the direction of arrival of the wideband acoustic signals. The performance of several frequency methods using the incoherent or coherent signal subspace are tested on the experimental data. The temporal method is used with the same data.>