Compressive sensing of noisy multispectral images is considered in this letter. Multispectral images in remote sensing applications are multichannel and inherently noisy. An approach using Bregman split method for optimization in both spatial and transform domains is proposed. The performance of the proposed algorithm is evaluated by comparing with other approaches. It is shown that the proposed algorithm performs favorably compared with other approaches with noisy multispectral images in experiments.

A fusion of multiple sensors with different resolution and noise characteristics is considered. The fusion of two sensors is formulated as a robust estimation problem, then the two-sensor fusion algorithm is generalized to the fusion of multiple sensors. In the two-sensor fusion problem, two sensors are assumed having complimentary characteristics, one with poor resolution and the other with poor noise characteristics. The fusion of multiple sensors is done by applying the weighted sum of images obtained by pairwise fusion. The multisensor fusion algorithm is tested with simulated images and real synthetic aperture radar images. In the experiment, the fusion algorithm yielded images where both resolution and signal-to-noise ratio are substantially improved compared to images before applying the fusion.

A robust modeling approach is applied to remove speckle noise in synthetic aperture images. Conventional adaptive filtering approaches in speckle filtering smoothes the image selectively depending on the details of underlying textures, and tend to blur details after speckle removal. In the proposed approach, the image is assumed to be composed of structural and stochastic components, and the stochastic component is modeled by a robust model. The proposed method is applied to real synthetic aperture images to demonstrate the validity and effectiveness of the algorithm.

A class of random field model having long-correlation characteristic is introduced. Unlike earlier approaches in long-correlation models, the correlation structure is isotropic or elliptical in this new class of random field model. A comprehensive three-step algorithm for parameter estimation is developed, and the validity of the model is demonstrated with real textures.

A new class of random field models, called generalized circular autoregressive (GCAR) models, is introduced. The GCAR models have non-causal neighbors which have the same autoregressive parameter values if they are on the same circle or ellipse, and have circular or elliptical correlation structure. The parameter estimation is also considered, and a multi-step estimation algorithm is presented. The efficacy of GCAR models in modeling real textures is demonstrated by synthesizing images resembling real textures.

A class of random field model having long-correlation characteristics is introduced. Unlike earlier approaches in long-correlation models, the correlation structure is isotropic or elliptical in this new class of random field model. The new model has an advantage of modeling diverse real textures with less than five model parameters. Further, the model parameters match with intuitive attributes of textures, such as smoothness, pattern size, elongation or orientation of patterns. The new long-correlation models are based on the fractional differencing of a two-dimensional (2-D) autoregressive polynomial defined by eight symmetric neighbors, and they are either persistent or periodic models depending whether the roots of the polynomial are real or imaginary. A comprehensive three-step algorithm for parameter estimation is developed, and the statistical properties of the estimators are also discussed. The validity of the new model in modeling textures is tested by synthesizing images from manually selected parameters as well as parameters estimated from real textures. It is shown that an image with desired attributes can be synthesized by selecting proper values of the parameters. Further, it is shown that the models introduced can be used in modeling wide range of textures by synthesizing images resembling real textures from estimated parameters.

Supervised and unsupervised texture segmentation using features extracted with statistical modeling approach is considered. A neural network is used for supervised segmentation, and a fuzzy clustering algorithm is used for unsupervised segmentation. The model used in this approach is a two-dimensional moving average model, and parameters estimated by a maximum likelihood method are used as texture features. The performance of the segmentation algorithms using model features are demonstrated in the experiment with both synthetic and natural images.

Contour modeling by a time-varying autoregressive (TVAR) model is considered. A least squares estimator of the TVAR model parameters is presented, and the maximum likelihood approach for determining the model order is also presented. In the experiment, curvature extrema points of synthesized contours are detected from the time frequency distribution estimated with TVAR model. In the classification experiment with contours of various planar shapes, about 97% of samples are correctly classified.

An algorithm for synthesizing color textures from a small set of parameters is presented. The synthesis algorithm is based on the 2-D moving average model, and realistic textures resembling many real textures can be synthesized using this algorithm. A maximum likelihood estimation algorithm to estimate parameters from a sample texture is also presented. Using the estimated parameters, a texture larger than the original image can be synthesized from a small texture sample. In the experiment, various textures suitable for multimedia applications are synthesized from parameters estimated from real textures.

In this correspondence, a random field model based on moving average (MA) time-series model is proposed for modeling stochastic and structured textures. A frequency domain algorithm to synthesize MA textures is developed, and maximum likelihood estimators are derived. The Cramer-Rao lower bound is also derived for measuring the estimator accuracy. The estimation algorithm is applied to real textures, and images resembling natural textures are synthesized using estimated parameters.

Modern image exploitation tasks have evolved from the early single-image, pixel-based and model-less methods to the current multi-image, multisensor, multiplatform, and model-based approaches. In this context, image positioning, which is the process of establishing the precise geometric relationship of an acquired image to the three-dimensional (3-D) world, has become an enabling technique for state-of-the-art multisensor data exploitation. Precise image positioning provides several benefits. Image registration, traditionally formulated as an image-to-image alignment problem, can now be carried out in accordance with interior and exterior sensor geometries. Images from sensors in arbitrary locations and orientations can be positioned with respect to a focal vertical and geocentric coordinate systems. This paper presents techniques for positioning images derived from various sensors such as electro-optical (E-O), synthetic aperture radar (SAR), and interferometric synthetic aperture radar (IFSAR). Applications to model-supported image exploitation are also discussed.

Considers the classification of radar signals by using stochastic models at different scales. The signal at a different scale is modeled by a hierarchical autoregressive moving average (ARMA) model, and the features at coarse scales are extracted from the model without performing expensive filtering operations. The hierarchical modeling can increase the accuracy of radar signal classification by exploiting features at different scales. For radar signal classification, model parameters at five different scales obtained by hierarchical modeling are used as features. A minimum distance classifier is implemented, and is tested on real aperture radar signals.

In this paper, we consider the hierarchical modeling of signals in a scale space using autoregressive and moving average (ARMA) models with applications to speech compression. We show that the AR polynomial can be uniquely determined when the scale is changed. When the scale changes from fine to coarse (aggregation) and from coarse to fine (disaggregation), the model parameters can be obtained from the parameters of the model at different scales. Data disaggregation is the estimation of data at a finer scale from data at a coarse scale. We present a data disaggregation algorithm based on the minimum mean square error (MMSE) criterion. The MMSE data disaggregation algorithm is computationally more efficient than the weighted least squares (WLS) approach. The data disaggregation algorithm is then applied to speech compression.>

In this paper, we consider the classification of speech signals by using stochastic models at different scales. The signal at different scales is modeled by a hierarchical autoregressive moving average (ARMA) model, and the features at coarse scales are extracted from the model without performing expensive filtering operation. The hierarchical modeling can increase the accuracy of speech classification by exploiting features at different scales. For speech classification, model parameters at five different scales obtained by hierarchical modeling are used as features. A minimum distance classifier is implemented, and tested on TIMIT speech data.

A statistical contour model is developed. A digital contour is modeled by a noisy observation which is represented by polynomial functions of coordinate variables. To estimate curvature functions of digital contours, the authors develop maximum likelihood estimators by fitting the model over a small neighborhood. The neighborhood size is determined by a maximum likelihood decision rule. Statistical properties of the estimators are also investigated. The contour is decomposed at curvature extrema points by finding zero-crossings of the first derivative of estimated curvature function. Experimental results show that the model based approach performs better in estimating curvature functions and detecting extrema points than other conventional approaches based on low-pass filtered curvature functions.>

The recognition of Korean characters by a syntactic method is considered. Korean characters are composed of phonetic symbols in two dimensions and contain very little redundancy. In addition, the phonetic symbols in each character are different in shape and number depending on how they are composed. Thus, attribute information is important. A Korean character recognition algorithm based on an attribute-dependent programmed grammar is presented. The preprocessing and primitive extraction algorithm is also described. The algorithm was implemented and tested with more than 9600 Korean characters in pages randomly selected from children's story books. The algorithm based on the attribute-dependent programmed grammar recognized characters reasonably quickly, with more than 95.1% accuracy.>

The recognition of unknown shapes by using maximum-likelihood methods is considered. The contour of a shape is represented by its centroidal profile, and it is fitted by a circular autoregressive model. The authors develop a decision rule to test whether two unknown shapes are identical. Then they develop a decision rule to classify an unknown object as a known object. Maximum-likelihood decision rules for these cases are derived. The decision rules are invariant to translation, rotation, and size change after the normalization of estimates. The developed algorithms are used to classify different machine parts and aircraft shapes.>

A pixel is regarded as an edge pixel if there is either a sharp change in the intensity values on either side of the pixel like a step function, yielding a so-called step edge or the texture on either side of the pixel is different, yielding a so-called texture edge. A two stage generate-and-confirm paradigm for detecting all the edge pixels in the scene is used. In the first stage, a directional derivatives approach for determining all potential edge pixels and the direction of the edge is employed. At this stage some of the edge pixels could be spurious, typically caused by either the noise in the image or the microedges inside a texture. In the second stage, each candidate pixel is subjected to two separate tests to confirm whether the edge pixel is a step edge or texture edge. The texture edge is confirmed by a likelihood-ratio test. The likelihood function is computed by fitting a nonsymmetric half-plane random-field model to the texture in a rectangular strip where the dominant direction is perpendicular to the estimated edge direction. Only the edge pixels that pass at least one of the two tests is accepted. The validity of the method is demonstrated by testing four different images.>

The problem of detecting texture boundaries without assuming any knowledge on the number of regions or the types of textures is considered. Texture boundaries are often regarded as better features than intensity edges, because a large class of images can be considered a composite of several different texture regions. An algorithm is developed that detects texture boundaries at reasonably high resolution without assuming any prior knowledge on the texture composition of the image. The algorithm utilizes the long correlation texture model with a small number of parameters to characterize textures. The parameters of the model are estimated by a least-squares method in the frequency domain. The existence and the location of texture boundary is estimated by the maximum-likelihood method. The algorithm is applied to several different images, and its performance is shown by examples. Experimental results show that the algorithm successfully detects texture boundaries without knowing the number of types of textures in the image.>

The authors describe a character recognition algorithm for Korean characters that uses attribute-dependent regular programmed grammar (ADRPG). Each Korean character has more strokes than an English character. and the recognition of Korean characters is heavily dependent on the attributes of strokes. They implemented a syntactic algorithm based on the ADPRG, which easily incorporates attribute information such as position, angle, length, and branch points. The ADPRG-algorithm is implemented on a PC-AT compatible, and 377 production rules and 78 attribute test functions are derived from a printed document containing 4206 Korean characters and applied another document containing 3092 Korean characters having some defects. More than 95.1% of 3092 Korean characters are correctly recognized with an average speed of 1.15 characters/s.>

An algorithm is developed to recognize shapes by a maximum-likelihood decision method after representing the contour of a shape by an autoregressive model. A decision rule is developed to test the similarity of objects pairwise. The rule is given in terms of the parameter estimates. The recognition of an arbitrary number of objects is accomplished by applying the decision rules to all possible pairwise combinations. The contour recognition algorithm developed is applied to contours of seven different machine parts and five different aircraft shapes. Without additive noise, all seven machine parts are classified 100% correctly. Contours of images contaminated by additive white Gaussian noise are tested. The proposed method has performed better than conventional methods on noisy images.>