Showing papers on "Kernel (image processing) published in 2004"

3D human pose from silhouettes by relevance vector regression

Abstract: We describe a learning based method for recovering 3D human body pose from single images and monocular image sequences. Our approach requires neither an explicit body model nor prior labelling of body pans in the image. Instead, it recovers pose by direct nonlinear regression against shape descriptor vectors extracted automatically from image silhouettes. For robustness against local silhouette segmentation errors, silhouette shape is encoded by histogram-of-shape-contexts descriptors. For the main regression, we evaluate both regularized least squares and relevance vector machine (RVM) regressors over both linear and kernel bases. The RVM's provide much sparser regressors without compromising performance, and kernel bases give a small but worthwhile improvement in performance. For realism and good generalization with respect to viewpoints, we train the regressors on images resynthesized from real human motion capture data, and test it both quantitatively on similar independent test data, and qualitatively on a real image sequence. Mean angular errors of 6-7 degrees are obtained - a factor of 3 better than the current state of the art for the much simpler upper body problem.

Nonstationary Multivariate Process Modeling through Spatially Varying Coregionalization

Abstract: Models for the analysis of multivariate spatial data are receiving increased attention these days. In many applications it will be preferable to work with multivariate spatial processes to specify such models. A critical specification in providing these models is the cross covariance function. Constructive approaches for developing valid cross-covariance functions offer the most practical strategy for doing this. These approaches include separability, kernel convolution or moving average methods, and convolution of covariance functions. We review these approaches but take as our main focus the computationally manageable class referred to as the linear model of coregionalization (LMC). We introduce a fully Bayesian development of the LMC. We offer clarification of the connection between joint and conditional approaches to fitting such models including prior specifications. However, to substantially enhance the usefulness of such modelling we propose the notion of a spatially varying LMC (SVLMC) providing a very rich class of multivariate nonstationary processes with simple interpretation. We illustrate the use of our proposed SVLMC with application to more than 600 commercial property transactions in three quite different real estate markets, Chicago, Dallas and San Diego. Bivariate nonstationary process models are developed for income from and selling price of the property.

Cancelable biometric filters for face recognition

Abstract: In this paper, we address the issue of producing cancelable biometric templates; a necessary feature in the deployment of any biometric authentication system. We propose a novel scheme that encrypts the training images used to synthesize the single minimum average correlation energy filter for biometric authentication. We show theoretically that convolving the training images with any random convolution kernel prior to building the biometric filter does not change the resulting correlation output peak-to-sidelobe ratios, thus preserving the authentication performance. However, different templates can be obtained from the same biometric by varying the convolution kernels thus enabling the cancelability of the templates. We evaluate the proposed method using the illumination subset of the CMU pose, illumination, and expressions (PIE) face dataset. Our proposed method is very interesting from a pattern recognition theory point of view, as we are able to 'encrypt' the data and perform recognition in the encrypted domain that performs as well as the unencrypted case, regardless of the encryption kernel used; we show analytically that the recognition performance remains invariant to the proposed encryption scheme, while retaining the desired shift-invariance property of correlation filters.

Face recognition by applying wavelet subband representation and kernel associative memory

Abstract: In this paper, we propose an efficient face recognition scheme which has two features: 1) representation of face images by two-dimensional (2D) wavelet subband coefficients and 2) recognition by a modular, personalised classification method based on kernel associative memory models. Compared to PCA projections and low resolution "thumb-nail" image representations, wavelet subband coefficients can efficiently capture substantial facial features while keeping computational complexity low. As there are usually very limited samples, we constructed an associative memory (AM) model for each person and proposed to improve the performance of AM models by kernel methods. Specifically, we first applied kernel transforms to each possible training pair of faces sample and then mapped the high-dimensional feature space back to input space. Our scheme using modular autoassociative memory for face recognition is inspired by the same motivation as using autoencoders for optical character recognition (OCR), for which the advantages has been proven. By associative memory, all the prototypical faces of one particular person are used to reconstruct themselves and the reconstruction error for a probe face image is used to decide if the probe face is from the corresponding person. We carried out extensive experiments on three standard face recognition datasets, the FERET data, the XM2VTS data, and the ORL data. Detailed comparisons with earlier published results are provided and our proposed scheme offers better recognition accuracy on all of the face datasets.

A Study on Convolution Kernels for Shallow Statistic Parsing

Abstract: In this paper we have designed and experimented novel convolution kernels for automatic classification of predicate arguments. Their main property is the ability to process structured representations. Support Vector Machines (SVMs), using a combination of such kernels and the flat feature kernel, classify Prop-Bank predicate arguments with accuracy higher than the current argument classification state-of-the-art.Additionally, experiments on FrameNet data have shown that SVMs are appealing for the classification of semantic roles even if the proposed kernels do not produce any improvement.

An EM-like algorithm for color-histogram-based object tracking

Abstract: The iterative procedure called 'mean-shift' is a simple robust method for finding the position of a local mode (local maximum) of a kernel-based estimate of a density function. A new robust algorithm is given here that presents a natural extension of the 'mean-shift' procedure. The new algorithm simultaneously estimates the position of the local mode and the covariance matrix that describes the approximate shape of the local mode. We apply the new method to develop a new 5-degrees of freedom (DOF) color histogram based non-rigid object tracking algorithm.

Image and Video Segmentation by Anisotropic Kernel Mean Shift

Abstract: Mean shift is a nonparametric estimator of density which has been applied to image and video segmentation. Traditional mean shift based segmentation uses a radially symmetric kernel to estimate local density, which is not optimal in view of the often structured nature of image and more particularly video data. In this paper we present an anisotropic kernel mean shift in which the shape, scale, and orientation of the kernels adapt to the local structure of the image or video. We decompose the anisotropic kernel to provide handles for modifying the segmentation based on simple heuristics. Experimental results show that the anisotropic kernel mean shift outperforms the original mean shift on image and video segmentation in the following aspects: 1) it gets better results on general images and video in a smoothness sense; 2) the segmented results are more consistent with human visual saliency; 3) the algorithm is robust to initial parameters.

Convolution Quadrature Revisited

Abstract: This article reviews convolution quadrature and its uses, extends the known approximation results for the case of sectorial Laplace transforms to finite-part convolutions with non-integrable kernel, and gives new, unified proofs of the optimal error bounds for both locally integrable and non-integrable convolution kernels.

Anomaly intrusion detection using one class SVM

Abstract: Kernel methods are widely used in statistical learning for many fields, such as protein classification and image processing. We recently extend kernel methods to intrusion detection domain by introducing a new family of kernels suitable for intrusion detection. These kernels, combined with an unsupervised learning method - one-class support vector machine, are used for anomaly detection. Our experiments show that the new anomaly detection methods are able to achieve better accuracy rates than the conventional anomaly detectors.

Fan-beam and cone-beam image reconstruction via filtering the backprojection image of differentiated projection data.

Abstract: In this paper, a new image reconstruction scheme is presented based on Tuy's cone-beam inversion scheme and its fan-beam counterpart. It is demonstrated that Tuy's inversion scheme may be used to derive a new framework for fan-beam and cone-beam image reconstruction. In this new framework, images are reconstructed via filtering the backprojection image of differentiated projection data. The new framework is mathematically exact and is applicable to a general source trajectory provided the Tuy data sufficiency condition is satisfied. By choosing a piece-wise constant function for one of the components in the factorized weighting function, the filtering kernel is one dimensional, viz. the filtering process is along a straight line. Thus, the derived image reconstruction algorithm is mathematically exact and efficient. In the cone-beam case, the derived reconstruction algorithm is applicable to a large class of source trajectories where the pi-lines or the generalized pi-lines exist. In addition, the new reconstruction scheme survives the super-short scan mode in both the fan-beam and cone-beam cases provided the data are not transversely truncated. Numerical simulations were conducted to validate the new reconstruction scheme for the fan-beam case.

Hardware efficient fast parallel FIR filter structures based on iterated short convolution

Abstract: This paper presents an iterated short convolution (ISC) algorithm, based on the mixed radix algorithm and fast convolution algorithm. This ISC-based linear convolution structure is transposed to obtain a new hardware efficient fast parallel finite-impulse response (FIR) filter structure, which saves a large amount of hardware cost, especially when the length of the FIR filter is large. For example, for a 576-tap filter, the proposed structure saves 17% to 42% of the multiplications, 17% to 44% of the delay elements, and 3% to 27% of the additions, of those of prior fast parallel structures, when the level of parallelism varies from 6 to 72. Their regular structures also facilitate automatic hardware implementation of parallel FIR filters.

Relevance feedback in region-based image retrieval

Abstract: Relevance feedback and region-based representations are two effective ways to improve the accuracy of content-based image retrieval systems. Although these two techniques have been successfully investigated and developed in the last few years, little attention has been paid to combining them together. We argue that integrating these two approaches and allowing them to benefit from each other will yield better performance than using either of them alone. To do that, on the one hand, two relevance feedback algorithms are proposed based on region representations. One is inspired from the query point movement method. By assembling all of the segmented regions of positive examples together and reweighting the regions to emphasize the latest ones, a pseudo image is formed as the new query. An incremental clustering technique is also considered to improve the retrieval efficiency. The other is the introduction of existing support vector machine-based algorithms. A new kernel is proposed so as to enable the algorithms to be applicable to region-based representations. On the other hand, a rational region weighting scheme based on users' feedback information is proposed. The region weights that somewhat coincide with human perception not only can be used in a query session, but can also be memorized and accumulated for future queries. Experimental results on a database of 10 000 general-purpose images demonstrate the effectiveness of the proposed framework.

MetaMorphs: Deformable shape and texture models

Abstract: We present a new class of deformable models, MetaMorphs, that consist of both shape and interior texture. The model deformations are derived from both boundary and region information in a common variational framework. This framework represents a generalization of previous model-based segmentation approaches. The shapes of the new models are represented implicitly as "images" in the higher dimensional space of distance transforms. The interior textures are captured using a nonparametric kernel-based approximation of the intensity probability density functions (p.d.f.s) inside the models. The deformations that MetaMorph models can undergo are defined using a space warping technique - the cubic B-spline based Free Form Deformations (FFD). When using the models for boundary finding in images, we derive the model dynamics from an energy functional consisting of both edge energy terms and intensity/texture energy terms. This way, the models deform wider the influence of forces derived from both boundary and regional information. The proposed MetaMorph deformable models are efficient in convergence, have large attraction range, and are robust to image noise and inhomogeities. Various examples on finding object boundaries in noisy images with complex textures demonstrate the potential of the proposed technique.

Prototype Modeling from Sketched Silhouettes based on Convolution Surfaces

Abstract: This paper presents a hybrid method for creating three-dimensional shapes by sketching silhouette curves. Given a silhouette curve, we approximate its medial axis as a set of line segments, and convolve a linearly weighted kernel along each segment. By summing the fields of all segments, an analytical convolution surface is obtained. The resulting generic shape has circular cross-section, but can be conveniently modified via sketched profile or shape parameters of a spatial transform. New components can be similarly designed by sketching on different projection planes. The convolution surface model lends itself to smooth merging between the overlapping components. Our method overcomes several limitations of previous sketched-based systems, including designing objects of arbitrary genus, objects with semi-sharp features, and the ability to easily generate variants of shapes.

Non-Mercer Kernels for SVM Object Recognition.

Abstract: On the one hand, Support Vector Machines have met with significant success in solving difficult pattern recognition problems with global features representation. On the other hand, local features in images have shown to be suitable representations for efficient object recognition. Therefore, it is natural to try to combine SVM approach with local features representation to gain advantages on both sides. We study in this paper the Mercer property of matching kernels which mimic classical matching algorithms used in techniques based on points of interest. We introduce a new statistical approach of kernel positiveness. We show that despite the absence of an analytical proof of the Mercer property, we can provide bounds on the probability that the Gram matrix is actually positive definite for kernels in large class of functions, under reasonable assumptions. A few experiments validate those on object recognition tasks.

Systems and methods for image pattern recognition

Abstract: Systems and methods for image pattern recognition comprise digital image capture and encoding using vector quantization ('VQ') of the image. A vocabulary of vectors is built by segmenting images into kernels and creating vectors corresponding to each kernel. Images are encoded by creating a vector index file having indices that point to the vectors stored in the vocabulary. The vector index file can be used to reconstruct an image by looking up vectors stored in the vocabulary. Pattern recognition of candidate regions of images can be accomplished by correlating image vectors to a pre-trained vocabulary of vector sets comprising vectors that correlate with particular image characteristics. In virtual microscopy, the systems and methods are suitable for rare-event finding, such as detection of micrometastasis clusters, tissue identification, such as locating regions of analysis for immunohistochemical assays, and rapid screening of tissue samples, such as histology sections arranged as tissue microarrays (TMAs).

Fast evaluation of nonreflecting boundary conditions for the Schrödinger equation in one dimension

Abstract: We present a fast algorithm for the evaluation of the exact nonreflecting boundary conditions for the Schrodinger equation in one dimension. The exact nonreflecting boundary condition contains a nonlocal term which is a convolution integral in time, with a kernel proportional to 1 √t. The key observation is that this integral can be split into two parts: a local part and a history part, each of which allows for separate treatment. The local part is computed by a quadrature suited for square-root singularities. For the history part, we approximate the convolution kernel uniformly by a sum of exponentials. The integral can then be evaluated recursively. As a result, the computation of the nonreflecting boundary conditions is both accurate and efficient.

Blind deconvolution of bar code signals

Abstract: Bar code reconstruction involves recovering a clean signal from an observed one that is corrupted by convolution with a kernel and additive noise. The precise form of the convolution kernel is also unknown, making reconstruction harder than in the case of standard deblurring. On the other hand, bar codes are functions that have a very special form—this makes reconstruction feasible. We develop and analyse a total variation based variational model for the solution of this problem. This new technique models systematically the interaction of neighbouring bars in the bar code under convolution with a kernel, as well as the estimation of the unknown parameters of the kernel from global information contained in the observed signal.

Fast model-based optical proximity correction

Abstract: An efficient method and system is provided for computing lithographic images that takes into account vector effects such as lens birefringence, resist stack effects and tailored source polarizations, and may also include blur effects of the mask and the resist. These effects are included by forming a generalized bilinear kernel, which is independent of the mask transmission function, which can then be treated using a decomposition to allow rapid computation of an image that includes such non-scalar effects. Dominant eigenfunctions of the generalized bilinear kernel can be used to pre-compute convolutions with possible polygon sectors. A mask transmission function can then be decomposed into polygon sectors, and weighted pre-images may be formed from a coherent sum of the pre-computed convolutions for the appropriate mask polygon sectors. The image at a point may be formed from the incoherent sum of the weighted pre-images over all of the dominant eigenfunctions of the generalized bilinear kernel. The resulting image can then be used to perform model-based optical proximity correction (MBOPC).

On the Use of Kernel PCA for Feature Extraction in Speech Recognition

Abstract: This paper describes an approachfor feature extraction in speech recognition systems using kernel principal componentanalysis (KPCA). This approachconsists in representing speech features as the projection of the extracted speech features mapped into a feature space via a nonlinear mapping onto the principal components. The nonlinear mapping is implicitly performed using the kerneltrick, which is an useful way of not mapping the input space into a featurespace explicitly,makingthis mapping computationally feasible. Better results were obtained by using this approach when compared to the standard technique.

The Kullback-Leibler kernel as a framework for discriminant and localized representations for visual recognition

Abstract: The recognition accuracy of current discriminant architectures for visual recognition is hampered by the dependence on holistic image representations, where images are represented as vectors in a high-dimensional space Such representations lead to complex classification problems due to the need to 1) restrict image resolution and 2) model complex manifolds due to variations in pose, lighting, and other imaging variables Localized representations, where images are represented as bags of low-dimensional vectors, are significantly less affected by these problems but have traditionally been difficult to combine with discriminant classifiers such as the support vector machine (SVM) This limitation has recently been lifted by the introduction of probabilistic SVM kernels, such as the Kullback-Leibler (KL) kernel In this work we investigate the advantages of using this kernel as a means to combine discriminant recognition with localized representations We derive a taxonomy of kernels based on the combination of the KL-kernel with various probabilistic representation previously proposed in the recognition literature Experimental evaluation shows that these kernels can significantly outperform traditional SVM solutions for recognition

Object categorization with SVM: kernels for local features

Abstract: In this paper, we propose to combine an efficient image representation based on local descriptors with a Support Vector Machine classifier in order to perform object categorization. For this purpose, we apply kernels defined on sets of vectors. After testing different combinations of kernel / local descriptors, we have been able to identify a very performant one.

Directional varying scale approximations for anisotropic signal processing

Abstract: A spatially adaptive restoration of a multivariable anisotropic function given by uniformly sampled noisy data is considered. The presentation is given in terms of image processing as it allows a convenient and transparent motivation of basic ideas as well as a good illustration of results. To deal with the anisotropy discrete directional kernel estimates equipped with varying scale parameters are exploited. The local polynomial approximation (LPA) technique is modified for a design of these kernels with a desirable polynomial smoothness. The nonlinearity of the method is incorporated by an intersection of confidence intervals (ICI ) rule exploited in order to obtain adaptive varying scales of the kernel estimates for each direction. In this way we obtain the pointwise varying scale algorithm which is spatially adaptive to unknown smoothness and anisotropy of the function in question. Simulation experiments confirm the advanced performance of the new algorithms.