Local Kernel Research Articles

Locally adaptive bilateral clustering for universal image denoising

AbstractThis paper presents a novel and efficient locally adaptive denoising method based on clustering of pixels into regions of similar geometric and radiometric structures. Clustering is performed by adaptively segmenting pixels in the local kernel based on their augmented variational series. Then, noise pixels are restored by selectively considering the radiometric and spatial properties of every pixel in the formed clusters. The proposed method is exceedingly robust in conveying reliable local structural information even in the presence of noise. As a result, the proposed method substantially outperforms other state-of-the-art methods in terms of image restoration and computational cost. We support our claims with ample simulated and real data experiments. The relatively fast runtime from extensive simulations also suggests that the proposed method is suitable for a variety of image-based products — either embedded in image capturing devices or applied as image enhancement software.

CrustalVSstructure in northwestern Canada: Imaging the Cordillera-craton transition with ambient noise tomography

[1] Crustal seismic velocity structure in northwest Canada is imaged using group velocity measurements derived from ambient noise cross correlations. The focus area surrounds the Canadian Northwest Experiment (CANOE) array, a 16 month deployment of 59 broadband seismic stations. The CANOE array extended from the Northern Cordillera on the west into the Archean Slave province to the east, crossing crustal terranes that span ∼4 Gyr of Earth history. Forty broadband stations from the Canadian National Seismograph Network and the POLARIS network are also included. The Green's function for each pair of stations is estimated by cross-correlating 24 h long time series of ambient noise for each day in the time period July 2004 to June 2005. Fundamental mode Rayleigh waves are observed on cross-correlated vertical component records and Love waves on the transverse components. The group velocity measurements are inverted for 3-D shear wave velocity within the crust. Local sensitivity kernels are used to account for the effects of a laterally variable sedimentary layer at shallow depths, and receiver function P-s differential times at the CANOE stations constrain total crustal thickness. Known sedimentary basins dominate lateral variations in wave speed in the shallow crust. In the middle crust, the model contains a strong low-velocity region that correlates spatially with the location of Proterozoic metasedimentary rocks that have been inferred from active source reflection profiles to occupy much of the crust beneath the Cordilleran terranes. Velocity variations in the lower crust, which are weakly resolved due to the limited period range of our data set (5–20 s), suggest a west-to-east transition from lower to higher wave speeds that aligns with the Cordilleran deformation front in parts of the study area. The results presented here are consistent with but do not confirm the notion of Proterozoic strata throughout much of the Cordilleran crust. These metasedimentary rocks, if present, must have experienced some deformation and metamorphism in order to also be consistent with upper mantle seismic models that contain a sharp transition from low to high velocity across the deformation front.

A New Approach to Estimation of the Length—Weight Relationship ofPollicipes pollicipes(Gmelin, 1789) on the Atlantic Coast of Galicia (Northwest Spain): Some Aspects of Its Biology and Management

This study was undertaken using data drawn from 5 sites along the Atlantic shoreline of Galicia (Northwest Spain) for a period of 2 y. The length—weight relationship of Pollicipes pollicipes (Gmelin, 1789) was estimated to observe the way in which individuals of this species gain weight as they increase in size. A classic allometric model was used for the purpose. As an alternative, a more general nonparametric model was also estimated, using local linear kernel smoothers. Comparison of these two models showed that use of the nonparametric model resulted in a better fit of the data. In addition, derivatives were used for estimating a size of capture for this species. For the same purpose, we also estimated this crustacean's mean size at sexual maturation (L 50) and the number of broods that it spawns per annum. Individuals' weight gain, a female maturity size of 15.7 mm, and P. pollicipes' estimated 1.73 broods per annum tend to suggest a size of capture based on a rostrocarinal length of 21.50 mm.

Semiparametric likelihood estimation in survival models with informative censoring

Semiparametric proportional hazard regression models are the cornerstone in modern survival analysis. Most estimation methodologies developed in the literature, such as the famous partial likelihood based estimation, are built on the ground that the censoring is noninformative. However, in many applications, the censoring is indeed informative. In this paper, we study the survival regression models with an informative censoring that is easy to detect and apply. A very important problem in practice is how to estimate the survival models more efficiently with the information from the informative censoring. We propose a semiparametric maximum likelihood approach that is easily implementable to estimate both the nonparametric baseline hazard and the parametric coefficients in the survival models with informative censoring. Different from the methods in the literature, we do not apply least informative approach to the baseline, which does not work well in our simulation. We solve the difficulty in semiparametric estimation by suggesting an indirect application of local kernel smoothing to the baseline. Asymptotic theory of the proposed estimators is established under informative and noninformative likelihoods, respectively. We suggest a cross-validation method to detect the informative censoring in application. The performances of the estimators in finite samples are investigated by Monte Carlo simulation. Both asymptotic theory and simulation show that the suggested semiparametric approach provides more efficient estimators of the parameters for informative censoring, and estimates the baseline function accurately. The proposed method is applied to analyse the data about the infants hospitalised for pneumonia, which leads to interesting findings.

Efficient Learning of Sample-Specific Discriminative Features for Scene Classification

Learning the sample-specific discriminative features based on numerous local learning may not scale well to real world scene classification tasks and suffer from the risk of overfitting. Hence we cast it in SVM based localized multiple kernel learning framework, and design a new strategy to alternately optimize the standard SVM solver and the sample-specific kernel weights, by either a linear programming (for <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">l</i> <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -norm) or with closed-form solutions (for <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">l</i> <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">p</sub> -norm). Experiments on both natural scene dataset and cluttered indoor scene dataset demonstrate the effectiveness and efficiency of our approach.

Facial Expression Recognition Based on Local Binary Patterns and Kernel Discriminant Isomap

Facial expression recognition is an interesting and challenging subject. Considering the nonlinear manifold structure of facial images, a new kernel-based manifold learning method, called kernel discriminant isometric mapping (KDIsomap), is proposed. KDIsomap aims to nonlinearly extract the discriminant information by maximizing the interclass scatter while minimizing the intraclass scatter in a reproducing kernel Hilbert space. KDIsomap is used to perform nonlinear dimensionality reduction on the extracted local binary patterns (LBP) facial features, and produce low-dimensional discrimimant embedded data representations with striking performance improvement on facial expression recognition tasks. The nearest neighbor classifier with the Euclidean metric is used for facial expression classification. Facial expression recognition experiments are performed on two popular facial expression databases, i.e., the JAFFE database and the Cohn-Kanade database. Experimental results indicate that KDIsomap obtains the best accuracy of 81.59% on the JAFFE database, and 94.88% on the Cohn-Kanade database. KDIsomap outperforms the other used methods such as principal component analysis (PCA), linear discriminant analysis (LDA), kernel principal component analysis (KPCA), kernel linear discriminant analysis (KLDA) as well as kernel isometric mapping (KIsomap).

Classification of schizophrenia using feature-based morphometry

The objective of this study was to use a combined local descriptor, namely scale invariance feature transform (SIFT), and a non linear support vector machine (SVM) technique to automatically classify patients with schizophrenia. The dorsolateral prefrontal cortex (DLPFC), considered a reliable neuroanatomical marker of the disease, was chosen as region of interest (ROI). Fifty-four schizophrenia patients and 54 age- and gender-matched normal controls were studied with a 1.5T MRI (slice thickness 1.25mm). Three steps were conducted: (1) landmark detection and description of the DLPFC, (2) feature vocabulary construction and Bag-of-Words (BoW) computation for brain representation, (3) SVM classification which adopted the local kernel to implicitly implement the feature matching. Moreover, a new weighting approach was proposed to take into account the discriminant relevance of the detected groups of features. Substantial results were obtained for the classification of the whole dataset (left side 75%, right side 66.38%). The performances were higher when females (left side 84.09%, right side 77.27%) and seniors (left side 81.25%, right side 70.83%) were considered separately. In general, the supervised weighed functions increased the efficacy in all the analyses. No effects of age, gender, antipsychotic treatment and chronicity were shown on DLPFC volumes. This integrated innovative ROI-SVM approach allows to reliably detect subjects with schizophrenia, based on a structural brain marker for the disease such as the DLPFC. Such classification should be performed in first-episode patients in future studies, by considering males and females separately.

Localized Multiple Kernel Learning for Realistic Human Action Recognition in Videos

Realistic human action recognition in videos has been a useful yet challenging task. Video shots of same actions may present huge intra-class variations in terms of visual appearance, kinetic patterns, video shooting, and editing styles. Heterogeneous feature representations of videos pose another challenge on how to effectively handle the redundancy, complementariness and disagreement in these features. This paper proposes a localized multiple kernel learning (L-MKL) algorithm to tackle the issues above. L-MKL integrates the localized classifier ensemble learning and multiple kernel learning in a unified framework to leverage the strengths of both. The basis of L-MKL is to build multiple kernel classifiers on diverse features at subspace localities of heterogeneous representations. L-MKL integrates the discriminability of complementary features locally and enables localized MKL classifiers to deliver better performance in its own region of expertise. Specifically, L-MKL develops a locality gating model to partition the input space of heterogeneous representations to a set of localities of simpler data structure. Each locality then learns its localized optimal combination of Mercer kernels of heterogeneous features. Finally, the gating model coordinates the localized multiple kernel classifiers globally to perform action recognition. Experiments on two datasets show that the proposed approach delivers promising performance.

Local kernel canonical correlation analysis with application to virtual drug screening

Drug discovery is the process of identifying compounds which have potentially meaningful biological activity. A major challenge that arises is that the number of compounds to search over can be quite large, sometimes numbering in the millions, making experimental testing intractable. For this reason computational methods are employed to filter out those compounds which do not exhibit strong biological activity. This filtering step, also called virtual screening reduces the search space, allowing for the remaining compounds to be experimentally tested.In this paper we propose several novel approaches to the problem of virtual screening based on Canonical Correlation Analysis (CCA) and on a kernel-based extension. Spectral learning ideas motivate our proposed new method called Indefinite Kernel CCA (IKCCA). We show the strong performance of this approach both for a toy problem as well as using real world data with dramatic improvements in predictive accuracy of virtual screening over an existing methodology.

A new flexible direct ROC regression model: Application to the detection of cardiovascular risk factors by anthropometric measures

The receiver operating characteristic (ROC) curve is the most widely used measure for evaluating the accuracy of diagnostic tests in terms of differentiating between two conditions. It is known that, in certain circumstances, the characteristics of the patient or the place where the diagnostic test is performed can modify the test’s accuracy. A new estimator for the conditional ROC curve, based on direct modelling, is proposed. In this approach, the effect of covariates and false positive fraction on the ROC curve is modelled non-parametrically using generalised additive models (GAM) combined with local polynomial kernel smoothers. The method allows for incorporation of more than one covariate in the regression model for the ROC curve and the possible interaction between them. The proposed model’s performance is examined in an in-depth simulation study. Finally, endocrine data are analysed with the aim of assessing the performance of several anthropometric measures in predicting clusters of cardiovascular risk factors in an adult population in Galicia (NW Spain), with adjustment for age and gender.

Edge detection for highly distorted images suffering Gaussian noise based on improve Canny algorithm

PurposeThe purpose of this paper is to detect edge of image in high noise level, suffering Gaussian noise.Design/methodology/approachCanny edge detection algorithm performs poorly when applied to highly distorted images suffering from Gaussian noise. In Canny algorithm, 2D‐gaussian function is used to remove noise and preserve edge. In high noise level, 2D‐gaussian function cannot meet the needs. In this paper, an improving Canny edge detection algorithm is presented. The algorithm presented is based on local linear kernel smoothing, in which local neighborhoods are adapted to the local smoothness of the surface measured by the observed data. The procedure can therefore remove noise correctly in continuity regions of the surface, and preserve discontinuities at the same time.FindingsThe statistical model of removing noise and preserving edge can meet the need of edge detection in images highly corrupted by Gaussian noise.Research limitations/implicationsIt was found that when the noise ratio is higher than 40 percent, the edge detection algorithm performs poorly.Practical implicationsA very useful method for detecting highly distorted images suffering Gaussian noise.Originality/valueSince an image can be regarded as a surface of the image intensity function and such a surface has discontinuities at the outlines of objects, this algorithm can be applied directly to detect edge of image in high noise level.

Structure of Mixture Kernel Function and its Application in Soft Measurement about Beating Degree

Kernel function is dominant in regression process of Support Vector Machine (SVM), it influences the prediction performance of SVM directly. Single local or global kernel function has limitation in generalization or learning capacity. The mixture kernel function with two good capacities had been structured, and it had been used in soft-sensor modeling of refining process. The contrast of simulation results show: the generalization ability of SVM based on mixture kernel function is better than the ones based on the single local or global kernel function, the algorithm had been applied in the refining process of HETAO wood industry factory, and the prediction result meets the requirement of craft production.

Nonparametric regression function estimation for errors-in-variables models with validation data

This paper develops estimation approach for nonparametric regression analysis with measurement error in covariables assuming the availability of inde- pendent validation data on covariables in addition to primary data on the response variable and surrogate covariables. Without specifying any error model structure between the surrogate and true covariables, we propose an estimator which inte- grates local linear regression and Fourier transformation method. An estimator combined by two local linear kernel estimators, is flrstly used to calibrate the con- ditional expectation of unknown objective regression function given the surrogate covariates, and then the flnal estimator can be derived by passing a trigonometric series approach suggested by Delaigle et al. (2006). Under mild conditions, the consistency of the proposed estimator is established and the convergence rate is also obtained. Numerical examples show that it performs well in applications.

Transverse domain wall propagation in modulated cylindrical nanostructures and possible geometric control

In this paper we present a model that studies domain wall propagation in geometrically modulated cylindrical nanostructures of small radii. Thus, we use the thin wire approximation, i.e., the magnetization is assumed constant on a cross section perpendicular to the longitudinal direction ($z$) of the nanostructure; we deal with transverse domain walls. We obtain general expressions for the average fields in these cross sections: Zeeman, anisotropy, exchange, and magnetostatic, and thus solve the Landau-Lifshitz-Gilbert equation for the dynamics of the magnetization. This effective one-dimensional model involves a dependence of the average exchange and dipolar fields on the changing geometry through local (similar to a nonadiabatic spin transfer term) and nonlocal kernels (written in terms of special functions), respectively. We do examine the effect of ``generic'' obstacles on the domain wall: localized changes of radius, bumps, and necks. The speed of propagation depends on these geometric obstacles and interestingly enough one may stop the propagation of the wall by adjusting the parameters of the geometry. Thus, one sees the potential of using geometry as a way to control domain wall motion.

Accurate singlet and triplet excitation energies using the Localized Hartree–Fock Kohn–Sham potential

We assess the accuracy of the LHFX Time-Dependent Density-Functional Theory (TD-DFT) approach, which uses Kohn–Sham orbitals and eigenvalues from the Localized Hartree–Fock (LHF) method and the exchange-only adiabatic local density approximation kernel. We compute 172 singlet and triplet excitation energies of π → π ∗, n → π ∗, σ → π ∗ and Rydberg character, for organic molecules of different size. We find that the LHFX method, which is free from the Self-Interaction-Error (SIE) and from empirical parameters, outperforms the state-of-the-art hybrid TD-DFT approaches, and provides the same accuracy for all different classes of excitations. The SIE-free Kohn–Sham orbitals can be thus considered as starting point for TD-DFT developments.

Iterative Filtering Decomposition Based on Local Spectral Evolution Kernel

The synthesizing information, achieving understanding, and deriving insight from increasingly massive, time-varying, noisy and possibly conflicting data sets are some of most challenging tasks in the present information age. Traditional technologies, such as Fourier transform and wavelet multi-resolution analysis, are inadequate to handle all of the above-mentioned tasks. The empirical model decomposition (EMD) has emerged as a new powerful tool for resolving many challenging problems in data processing and analysis. Recently, an iterative filtering decomposition (IFD) has been introduced to address the stability and efficiency problems of the EMD. Another data analysis technique is the local spectral evolution kernel (LSEK), which provides a near prefect low pass filter with desirable time-frequency localizations. The present work utilizes the LSEK to further stabilize the IFD, and offers an efficient, flexible and robust scheme for information extraction, complexity reduction, and signal and image understanding. The performance of the present LSEK based IFD is intensively validated over a wide range of data processing tasks, including mode decomposition, analysis of time-varying data, information extraction from nonlinear dynamic systems, etc. The utility, robustness and usefulness of the proposed LESK based IFD are demonstrated via a large number of applications, such as the analysis of stock market data, the decomposition of ocean wave magnitudes, the understanding of physiologic signals and information recovery from noisy images. The performance of the proposed method is compared with that of existing methods in the literature. Our results indicate that the LSEK based IFD improves both the efficiency and the stability of conventional EMD algorithms.

A Inter-Subsystem Private Data Transfer Protocol Based on a New High Secure Computer Architecture

Focusing on the characteristics of the new high secure architecture of network computer, an operating system with internal network structure is designed. The operating system contains two subkernels: local kernel and network kernel, the two subkernels run individually in two subsystems. In order to communicate between two subsystems securely, an inter-subsystem private data transfer protocol is proposed and implemented in this paper. The private protocol is a connection-oriented protocol, it can provide reliable end-to-end connectivity Protocol format and protocol connection management based on signature verification are elaborated.

Nonparametric estimation of varying coefficient error-in-variable models with validation sampling

In this paper, varying coefficient models are investigated with the response and covariate prone to measurement error. Without specifying any error structure equation, four estimators of the coefficient function vector are proposed by using the local linear kernel smoothing technique and also proved to be asymptotically normal. The data-driven bandwidth selection method is discussed. Simulation examples are conducted to evaluate the proposed estimation methods.

Understanding ignition processes in spray-guided gasoline engines using high-speed imaging and the extended spark-ignition model SparkCIMM: Part B: Importance of molecular fuel properties in early flame front propagation

Recent high-speed imaging of ignition processes in spray-guided gasoline engines has motivated the development of the physically-based spark channel ignition monitoring model SparkCIMM, which bridges the gap between a detailed spray and vaporization model and a model for fully developed turbulent combustion. Previously, both SparkCIMM and high-speed optical imaging data have shown that, in spray-guided engines, large variations in turbulence intensity, equivalence ratio, and enthalpy along the stretched and wrinkled spark plasma channel favor localized ignition spot formations in rich-mixture regions. In combination with strong local flow velocity, multiple successful ignition events along the re-striking spark lead to early non-spherical turbulent flame fronts. In this paper, SparkCIMM is enhanced by: (1) criteria to capture localized flame extinction phenomena, (2) a formulation of early flame kernel propagation based on the G-equation theory that includes effects of non-unity Lewis numbers, and (3) an extended equation to compute turbulent burning velocities of stretched flames in stratified mixtures. Localized rich ignition along the spark leads to early flames, whose propagation is, due to initially small turbulent Damköhler numbers, significantly influenced by molecular fuel properties. The analysis reveals that non-unity Lewis number curvature effects, intensified by heavy dilution by exhaust gas recirculation, strongly affect the early flame-kernel development in spray-guided gasoline engines. In particular, these effects significantly bias the flammability limit of flame kernels towards rich-mixtures while inhibiting their propagation in lean regions. Favorable initial conditions for combustion are found in rich-mixture regions, albeit in the presence of substantial equivalence ratio fluctuations and scalar dissipation rates. This paper demonstrates that the full complexity of the model equations developed here is required to reproduce the characteristic experimental features (spark channel stretching, multiple re-strikes, localized flame kernel formation, and early turbulent flame front corrugation) of spray-guided ignition phenomena.

Signal estimation and change detection in tank data for nuclear safeguards

Process monitoring (PM) is increasingly important in nuclear safeguards as a complement to mass-balance based nuclear materials accounting (NMA). Typically, PM involves more frequent but lower quality measurements than NMA. While NMA estimates special nuclear material (SNM) mass balances and uncertainties, PM often tracks SNM attributes qualitatively or in the case of solution monitoring (SM) tracks bulk mass and volume. Automatic event marking is used in several nuclear safeguards PM systems. The aims are to locate the start and stop times and signal changes associated with key events. This paper describes results using both real and simulated SM data to quantify the errors associated with imperfect marking of start and stop times of tank events such as receipts and shipments. In the context of safeguards, one can look both forward and backward in modest time intervals to recognize events. Event marking methods evaluated include differencing, multi-scale principal component analysis using wavelets, and piecewise linear regression (PLR). All methods are evaluated on both raw and smoothed data, and several smoothing options are compared, including standard filters, hybrid filters, and local kernel smoothing. The main finding for real and simulated examples considered is that a two-step strategy is most effective. First, any reasonably effective initial smoother is used to provide a good initial guess at change point locations. Second, PLR is applied, looking for one change point at a time. In contrast, PLR that allows for multiple change points simultaneously has worse performance.