By Topic

Vision, Image and Signal Processing, IEE Proceedings -

Issue 5 • Date Oct 2001

Filter Results

Displaying Results 1 - 8 of 8
  • Practical self-calibration of pan-tilt cameras

    Page(s): 349 - 355
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (1019 KB)  

    The authors propose a practical self-calibration method of rotating and zooming cameras. The problem with previous methods occurs when the camera motion is almost fully zoomed with very little rotation, which is called the 'near-degenerate' configuration. In that case, the solutions become unstable and rotation angles cannot be calculated. When a pan-tilt camera (without z-axis rotation) is adopted and the intrinsic camera parameters are simplified, the near-degenerate configuration can be overcome and a closed-form solution obtained. Because pan-tilt cameras can be assumed for most stationary cameras (i.e. without translation) and the assumptions about the intrinsic camera parameters do not seem to effect the self-calibration, the method provides a simple, practical solution to the self-calibration problem. In addition, the authors introduce a nonlinear algorithm that adjusts not only the camera parameters but also the inter-image homography so that more accurate image registration is made possible. Simulations and experiments with real images are presented View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Laplace spectrum for exponential decomposition and pole-zero estimation

    Page(s): 305 - 314
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (658 KB)  

    A novel Laplace transform power spectrum is proposed as a basic tool for complex-exponential decomposition of finite-duration continuous functions. The weighted spectrum leads to a direct decomposition in terms of exponential damping function content in addition to the usual decomposition in terms of sinusoidal circular content produced by Fourier and Laplace transforms. The weighting results in a two-dimensional spectral peak on the Laplace s-plane identifying the exponential and frequency content of finite-duration functions. An algorithm for two-dimensional generalised spectral analysis of finite duration functions is described. It is shown that the proposed spectral weighting eliminates the exponential divergence of spectra of usual transforms which masks poles along one dimension of the complex plane. As an application, an algorithm for mathematical modelling pole-zero estimation through a direct two-dimensional localisation of pole-zero peaks of weighted spectra is described View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • New nonparametric dominant point detection algorithm

    Page(s): 363 - 374
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (1058 KB)  

    The authors propose a new nonparametric dominant point detection algorithm which is divided into two phases: an initial detection phase that locates possible dominant points and a suppression phase that removes redundant dominant points. In the initial detection phase, not only the points with high local curvatures, but also the end points and interception points are detected. In the suppression phase, a novel measurement to act as a suppressing criterion for the removal of the redundant dominant points is proposed. The curvature of a contour segment is modelled by the average cosine angle. If the contour is slightly curved, more points will be suppressed. On the other hand, fewer points will be suppressed if the curved contour is highly curved. The experimental results show that the proposed algorithm can obtain a set of dominant points to represent contours efficiently View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Analytical model for the first and second moments of an adaptive interpolated FIR filter using the constrained filtered-X LMS algorithm

    Page(s): 337 - 347
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (736 KB)  

    The authors present an analytical model for the mean weight behaviour and weight covariance matrix of an adaptive interpolated FIR filter using the LMS algorithm to adapt the filter weights. The particular structure of this adaptive filter determines that special analytical considerations must be used. First, the introduction of an interpolating block cascaded with the adaptive sparse filter requires that the input signal correlations must be considered. It is well known that such correlations are disregarded by the independence theory, which is the basis for the analysis of the LMS algorithm adapting FIR structures. Secondly a constrained analysis is used to deal mathematically with the sparse nature of the adaptive section. Experimental results demonstrate the effectiveness of the proposed analytical models as compared with the results obtained by classical analysis View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Online unsupervised learning of hidden Markov models for adaptive speech recognition

    Page(s): 315 - 324
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (1067 KB)  

    A novel framework of an online unsupervised learning algorithm is presented to flexibly adapt the existing speaker-independent hidden Markov models (HMMs) to nonstationary environments induced by varying speakers, transmission channels, ambient noises, etc. The quasi-Bayes (QB) estimate is applied to incrementally obtain word sequence and adaptation parameters for adjusting HMMs when a block of unlabelled data is enrolled. The underlying statistics of a nonstationary environment can be successively traced according to the newest enrolment data. To improve the QB estimate, the adaptive initial hyperparameters are employed in the beginning session of online learning. These hyperparameters are estimated from a cluster of training speakers closest to the test environment. Additionally, a selection process is developed to select reliable parameters from a list of candidates for unsupervised learning. A set of reliability assessment criteria is explored for selection. In a series of speaker adaptation experiments, the effectiveness of the proposed method is confirmed and it is found that using the adaptive initial hyperparameters in online learning and the multiple assessments in parameter selection can improve the recognition performance View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Design of two-dimensional FIR digital filters by McClellan transform and quadratic programming

    Page(s): 325 - 331
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (529 KB)  

    A quadratic programming (QP) approach for determining the coefficients of the McClellan transform is presented for the design of 2-D FIR digital filters. Three features of the proposed method are as follows. First, the transform parameters are determined by minimising the integration of the squared errors along the desired contour. Second, a set of linear constraints are incorporated into the QP formulation such that the conventional scaling problem of the transform can be avoided. Third, the optimal cutoff frequencies of a 1-D prototype filter are obtained directly from the QP solution. Several design examples, including fan filters, elliptic filters, diamond filters and bandpass filters, are illustrated to demonstrate the effectiveness of the QP method View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Estimation of depth from defocus as polynomial system identification

    Page(s): 356 - 362
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (1310 KB)  

    The two-image depth from defocus problem is considered as a system identification problem. The underlying phenomenon of defocusing is modelled as a linear system with an appropriate transfer function. Using a general measure for the spread, a relation between the spread and the system transfer function is obtained. A parametric transfer function is utilised in deriving an analytical expression for the spread. The two images taken with different camera parameters then become the input-output of such a system. An algorithm for estimating the coefficients of the transfer function is derived. The method is tested on synthetic images as well as images obtained from a camera View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Robust least mean square adaptive FIR filter algorithm

    Page(s): 332 - 336
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (416 KB)  

    The authors propose a new robust adaptive FIR filter algorithm for system identification applications based on a statistical approach named the M estimation. The proposed robust least mean square algorithm differs from the conventional one by the insertion of a suitably chosen nonlinear transformation of the prediction residuals. The effect of nonlinearity is to assign less weight to a small portion of large residuals so that the impulsive noise in the desired filter response will not greatly influence the final parameter estimates. The convergence of the parameter estimates is established theoretically using the ordinary differential equation approach. The feasibility of the approach is demonstrated with simulations View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.