Volumetric synthetic aperture radar (VolSAR) analysis techniques and image formation algorithms suitable for short ranges and large scenes are presented. From a diffractive wave field inversion perspective, ultrawide beamwidth and near range SAR imaging scenarios can be viewed as a form ofin situSAR. Novel Huygens–Fresnel processing methods are introduced that empowerin situvolumetric imaging with 2-D and 3-D aperture synthesis. These methods support coherent fusion across multiple separated frequency bands and also support spatial subaperture fusion of data from sparse sensor swarms. A novel signal analysis tool we call achirp coupletis developed and shown to be useful in the expression and exploitation of temporal and spatial SAR chirp signals. Chirp couplets provide a physically motivated and unifying tool for both temporal and spatial elements in SAR signal analysis. In the temporal domain, chirp couplets provide a generalized formulation of the coupling of time and frequency that exists, for example, in linear frequency modulated waveforms. In the spatial domain, chirp couplets describe the coupling of sensor position and sensed wavenumber. Formulations of chirp couplets suitable for ray tomographic SAR algorithms (i.e., polar format algorithms) are contrasted with chirp couplets capable of supporting diffractive tomographic SAR algorithms (i.e., Stolt format algorithms). Scenarios in which diffraction limited resolution can be achieved within situVolSAR and finite length synthetic apertures are explored. VolSAR imaging methods are shown to escape the approximations that constrain the application and performance of range-Doppler and polar format methods.

In this work, for a given a set of code vector assignments to an input by a multistage residual vector quantizer RVQ [1], Bayesian framework is formulated to find the most probable class membership of the input. Furthermore, Markov structure is also used to improve the memory cost of the classification.

Multistage residual vector quantizers (RVQ) with optimal direct sum decoder codebooks have been successfully designed and implemented for data compression. Due to its multistage structure, RVQ has the ability to densely populate the input space with voronoi cell partitions. The same design concept has yielded good results in the application of image-content classification. Furthermore, the multistage RVQ, with stage-wise codebooks, provides an opportunity to perform fine-grained feature attribution for image understanding, in general, and feature foundation data generation for natural and man-made structure recognition, in specific. In, the information at the stages of RVQ is heuristically integrated to perform class conditional pattern recognition; hence the process is not robust. Markov random field (MRF) provides a suitable Bayesian framework to integrate the information available at the various stages of RVQ to achieve optimized classification in the maximum a-posteriori sense (MAP).

Summary form only given. An analytical expression for the uniform midrise quantizer function can be written as a sum of the input and a residual sawtooth wave. We propose an equivalent but alternate approach. Although it is well known that no clever mathematical manipulation can yield extra information, nevertheless, an alternate form of expression can and has on several occassions proven useful.

In this paper, we propose a real time, robust contour tracker for vehicles in IR images captured from an aerial moving camera. Lack of a background model, lack of a motion model and lack of any training data make the job challenging, but realistic. We compare a standard approach of contour evolution using energy minimization in the snakes algorithm with our Harris corner based Elliptical Fourier contour representation. We show that our approach produces better results and is robust to different initialization strategies.

In this paper, we discuss methods to enable robust surveillance on compressed video. We show that if the particular surveillance algorithm that is likely to be run on the compressed video is known a priori, then steps can be taken during the encoding process to facilitate the performance of the algorithm. We show that by performing signal processing on the input video signal before it is encoded, or by adaptively changing the parameters of the encoding process, we can make the resulting signal more robust to degradations in the encoding process. The result is better and more consistent tracking on the compressed video from high to low bitrates, but with some loss in PSNR. We demonstrate the validity of this approach for mean shift tracking running on MPEG-4 coded video.

In this paper, our goal is to understand what needs to be done to enable computer vision algorithms running on uncompressed image sequences to run as well on image sequences that have undergone compression and then decompression. The central conflict of context based computer vision algorithms versus the structured block based approach of today's codecs means that more has to be done than to simply create a divide between coding foreground preferentially and giving less importance to background. We take as example, a single computer vision algorithm, the mean shift tracker and see that its performance can be improved substantially in low bit rate scenarios, albeit some tradeoffs.

Fourier analysis-based focusing of synthetic aperture radar (SAR) data collected during circular flight path is a recent advancement in SAR signal processing. This paper uses the Householder transform to obtain a ground plane circular SAR (CSAR) signal phase history from the slant plane CSAR phase history by inverting the linear shift-varying system model, thereby circumventing the need for explicitly computing a pseudo-inverse. The Householder transform has recently been shown to have improved error bounds and stability as an underdetermined and ill-conditioned system solver, and the Householder transform is computationally efficient.

Image-driven data mining methods are described for image content segmentation, classification, and attribution, where each pixel location of an image-under-analysis is the center point of a pixel-block query that returns an estimated class label. Feature attribute estimates may also be mined when sufficient attribute strata exist in the data warehouse. Novel methods are presented for pixel-block mining, pattern similarity scoring, class label assignments, and attribute mining. These methods are based on a direct sum tree structure called a sigma-tree that is utilized with near-neighbor similarity scoring. The sigma-tree structure provides a solution to the challenge of high computation/memory costs of pixel-block similarity searching. The sigma-trees are integrated into warehouse subsystems that provide referential capability into feature attribute data, resulting in a foundation for data mining called Source Optimized, Labeled, DIgital Expanded Representations (SOLDIER). The variable depth "bit-plane" data representations produced by sigma-tree path selections provide an approach to image content segmentation, and provide a structure for formulation of Bayesian classification with data-adaptive Parzen classifiers with variably sized windows. Preliminary methods and results for postprocessing of mined feature-thematic layers for higher level scene understanding are also presented. Sample results are shown with synthetic aperture radar images and with high-resolution pan-sharpened satellite images of the Payagala, Sri Lanka area before the site was devastated by the 2004 Asian Tsunami.

Detection, classification, and attribution of high-resolution satellite image features in nearshore areas in the aftermath of Hurricane Katrina in Gulfport, MS, are investigated for damage assessments and emergency response planning. A system-level approach based on image-driven data mining with sigma-tree structures is demonstrated and evaluated. Results show a capability to detect hurricane debris fields and storm-impacted nearshore features (such as wind-damaged buildings, sand deposits, standing water, etc.) and an ability to detect and classify nonimpacted features (such as buildings, vegetation, roadways, railways, etc.). The sigma-tree-based image information mining capability is demonstrated to be useful in disaster response planning by detecting blocked access routes and autonomously discovering candidate rescue/recovery staging areas

Synthetic aperture radar (SAR) image classification for late-season rural land-cover estimation is investigated. A novel tree-structured nearest neighbor-like classifier is applied to polarimetric SAR intensity image pixel blocks. The novel tree structure, called a sigma-tree, is generated by an ordered summation of unweighted template refinements. Computation and memory costs of a sigma-tree classifier grow linearly. The reduced costs of sigma-tree classifiers are obtained with the tradeoff of a guarantee of nearest neighbor mappings. Causal-anticausal refinement-template design methods, combined with causal multiple-stage search engine structures, are shown to yield sequential search decisions that are acceptably near-neighbor mappings. The performance of a sigma-tree classifier is demonstrated for rural land-cover estimation with detected polarimetric C-band AirSAR pixel data. Experiments are conducted on various polarization/pixel block size combinations to evaluate the relative utility of spatial-only, polarimetric-only, and combined spatial/polarimetric classifier inputs

This paper deals with successive approximation source coding and image enabled data mining. Successive approximation source codes provide query returns consisting of sequences of aggregate data-tuples with image sets. A data mining statistical analysis or pattern search over a sequence of aggregates provides a sequence of data mining answers that is desirably fuzzy. Residual vector quantization (RVQ) provides a successive approximation source code with utility in image-enabled queries in image data mining tasks.

Sigma-trees (/spl sigma/-trees) are a class of geometric structures that include lattices as a constrained special case. These structures allow for signal sets, in spaces of arbitrary dimension, that are more spherical in shape than signal sets based on lattices. In this correspondence, it is shown that /spl sigma/-trees can be used in the construction of non-lattice trellis-coded modulation schemes (TCM) schemes. A low-complexity /spl sigma/-tree encoder is presented for multidimensional TCM codes. An optimal TCM decoder that uses the efficient sequential search property of /spl sigma/-trees is then described. Simulation results based on one- and two-dimensional examples are used to show that the performance of /spl sigma/-tree-based codes is comparable to that provided by conventional lattice-based codes.

Advances in residual vector quantization (RVQ) are surveyed. Definitions of joint encoder optimality and joint decoder optimality are discussed. Design techniques for RVQs with large numbers of stages and generally different encoder and decoder codebooks are elaborated and extended. Fixed-rate RVQs, and variable-rate RVQs that employ entropy coding are examined. Predictive and finite state RVQs designed and integrated into neural-network based source coding structures are revisited. Successive approximation RVQs that achieve embedded and refinable coding are reviewed. A new type of successive approximation RVQ that varies the instantaneous block rate by using different numbers of stages on different blocks is introduced and applied to image waveforms, and a scalar version of the new residual quantizer is applied to image subbands in an embedded wavelet transform coding system.

One of the more effective data compression systems that has been recently proposed is the relatively simple embedded wavelet image coder developed by J.M. Shapiro (1994). Two key components of Shapiro's system are the use of zerotrees to keep track of insignificant subband coefficients and progressive transmission of successive bit planes of significant coefficients. Shapiro's quantization mechanism is the use of scaled successive approximation uniform scalar quantizers. This paper investigates ways of improving the performance of embedded wavelet coders with the use of optimized successive approximation direct sum quantization structures.

An entropy-constrained residual vector quantization design algorithm is used to design codebooks for image coding. Entropy-constrained residual vector quantization has several important advantages. It can outperform entropy-constrained vector quantization in terms of rate-distortion performance, memory, and computation requirements. It can also be used to design vector quantizers with relatively large vector sizes and high output rates. Experimental results indicate that good image reproduction quality can be achieved at relatively low bit rates. For example, a peak signal-to-noise ratio of 30.09 dB is obtained for the 512x512 LENA image at a bit rate of 0.145 b/p.

Necessary conditions for the optimality of variable-rate residual vector quantizers are derived, and an iterative descent algorithm based on a Lagrangian formulation is introduced for designing residual vector quantizers having minimum average distortion subject to an entropy constraint. Simulation results for entropy-constrained residual vector quantizers are presented for memoryless Gaussian, Laplacian, and uniform sources. A Gauss-Markov source is also considered. The rate-distortion performance is shown to be competitive with that of entropy-constrained vector quantization and entropy-constrained trellis-coded quantization.

Multiple path searching can provide varying degrees of joint search optimization of residual vector quantizer encoder stages. A short coming of the conventional multiple path M-search algorithm, however, is that joint search optimization of encoder stages is limited to consecutive stages. A new iterated multipath (IM)-search algorithm is introduced that is not subject to any particular ordering of the residual quantizer stages. The IM-search algorithm may be combined with the sequential M-search algorithm to provide additional enhancement of residual vector quantization encoder performance. Furthermore, additional details of design methods for residual quantizers with separate, and, in general, different encoder and decoder cookbooks are given. Separate encoder and decoder cookbooks facilitate the use and design of various suboptimal, but computationally efficient encoder structures, while maintaining the use of decoder stage codebooks which satisfy necessary conditions for the joint optimality of direct sum codebooks.>

A design method for adaptive successive approximation residual vector quantizers with suboptimal but computationally efficient sequential search encoders is given. Image-specific codebook adaptation and the sequential search encoder structural constraint are shown to be more easily accommodated with the use of separate, and in general, different encoder and decoder code books. The design method generates encoder codebooks that provide computationally efficient successive approximation tree structures, and generates decoder codebooks that satisfy conditions necessary for joint optimality. Design monotonicity is maintained even with the use of small training sets sizes, as required by image-specific codebook adaptation. The memory advantage of the direct sum structure permits image adapted codebooks to be efficiently stored or transmitted as overhead information.>

The linear growth with the dimension-rate product of RVQ computation and memory requirements permits practical implementations of RVQs with large dimensions or high rates. This feature is exploited by quantizing low-resolution subbands with small-dimension high-rate RVQs, and high-resolution subbands with large-dimension low-rate RVQs. The RVQ vector sizes vary by a factor of four in parallel with the decimation and up-sampling processes from one resolution level to the next. Two forms of rate allocation are achieved with the RVQ subband system. A type of concentric shell partitioned vector classifier with side information is used to separate noise-like subband vectors from structured subband vectors. For the large-dimension low-rate RVQs, variable rate RVQ with side information permits different numbers of RVQ stages to be used on different vectors within a concentric shell partition class.>

The use of direct sum codebooks to minimize the memory requirements of vector quantizers is investigated. Assuming arbitrary fixed partitions, necessary conditions for minimum distortion codebooks are derived, first for scalar codebooks, assuming mean-squared error distortion, and then for vector codebooks and a broader class of distortion measures. An iterative procedure is described for designing locally optimal direct sum codebooks. Both optimal and computationally efficient suboptimal encoding schemes are considered. It is shown that although an optimal encoding can be implemented by a sequential encoder, the complexity of implementing optimal stagewise partitions generally exceeds the complexity of an exhaustive search of the direct sum codebook. It is also shown that sequential nearest-neighbor encoders can be extremely inefficient. The M-search method is explored as one method of improving the effectiveness of suboptimal sequential encoders. Representative results for simulated direct sum quantizers are presented.>

A new residual vector quantizer (RVQ) design algorithm is modified so that the multistage structure can be exploited to produce variable-rate RVQ (VR-RVQ) systems. VR-RVQ systems are shown to have very useful properties: (1) the codebook storage requirement and the search complexity are both reduced; (2) the VR-RVQ system is able to exploit the spatial variance of perceptually important information; and (3) the VR-RVQ codebook can operate over a wide range of rates, without having to store several codebooks. Experiments were performed using VR-RVQ systems with vectors of many sizes, and results show significant improvement over fixed-rate RVQ systems with the same block size.>

This paper shows that tree search encoders are ineffective when used to determine the closest code vector in residual quantizer (RQ) alphabets. In particular, the equivalent cell boundaries are poorly chosen and the labelling of equivalent code vectors produced by the decoder and by a tree-structured encoder are inconsistent. This problem does not arise when two-level RQ alphabets are used in trellis coded vector quantizers. Trellis-coded RQs are designed for the memoryless Gaussian, Laplacian, and Gauss-Markov sources at a rate of R=1 bit per sample with encouraging results; a SQNR of 5.92 dB has been achieved on the Gaussian source.>