By Topic

Vision, Image and Signal Processing, IEE Proceedings -

Issue 4 • Date Aug. 2001

Filter Results

Displaying Results 1 - 13 of 13
  • New fast algorithm for multidimensional type-IV DCT

    Page(s): 263 - 268
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (510 KB)  

    The authors first propose an index mapping such that the type-IV m-dimensional discrete cosine transform (m-D DCT-IV) is turned into a sum involving a number of (m-1)-dimensional discrete cosine transforms ((m-1)-D DCTs). Then a polynomial transform is used for implementing the sum. Based on symmetrical properties, a refined fast polynomial transform algorithm is proposed for computing the polynomial transform. Compared to the row-column m-D DCT-IV algorithm, the proposed algorithm achieves remarkable savings in arithmetic operations. More precisely, the numbers of multiplications and additions for m-dimensional DCT-IV are nearly 1/m and (2m+1)/3m times those of the row-column method, respectively. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Bayesian blind and semi-blind equalisation of channels with Markov inputs

    Page(s): 269 - 274
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (440 KB)  

    An original full Bayesian approach is developed for blind and semi-blind equalisation of fading channels with Markov inputs. The sequence of discrete symbols is estimated according to a marginal maximum a posteriori criterion; the other unknown parameters are regarded as random nuisance parameters and are integrated out analytically. A batch algorithm is proposed to maximise the marginal posterior distribution. Simulation results are presented to demonstrate the effectiveness of the method View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Recursive blind identification of non-Gaussian time-varying AR model and application to blind equalisation of time-varying channel

    Page(s): 275 - 282
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (592 KB)  

    A novel method for the blind identification of a non-Gaussian time-varying autoregressive model is presented. By approximating the non-Gaussian probability density function of the model driving noise sequence with a Gaussian-mixture density, a pseudo maximum-likelihood estimation algorithm is proposed for model parameter estimation. The real model identification is then converted to a recursive least squares estimation of the model time-varying parameters and an inference of the Gaussian-mixture parameters, so that the entire identification algorithm can be recursively performed. As an important application, the proposed algorithm is applied to the problem of blind equalisation of a time-varying AR communication channel online. Simulation results show that the new blind equalisation algorithm can achieve accurate channel estimation and input symbol recovery View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Efficient and consistent method for superellipse detection

    Page(s): 227 - 233
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (648 KB)  

    Superellipses are a flexible representation for a variety of objects and the detection of such a primitive is an interesting issue in machine vision research. Least-mean-square fitting using an algebraic distance has been suggested to determine the parameters of a superellipse, but the computational cost is high and a high curvature bias problem is involved. An efficient, consistent and threshold-free scheme is derived for the estimation of superellipse parameters. The closed solutions for the centre, orientation and squareness parameters are obtained by using the zeroth harmonic of its Fourier description, the consistent symmetric axis method and the theorem of diagonal segment, respectively. Only the lengths of the major and the minor axes are repeatedly estimated by Powell's conjugate direction technique to reduce the sensitivity of noise. The proposed method is suitable for use on relatively complete, closed superellipse curves. Both convex and concave superellipses have been considered, and a compensation technique is suggested for concave superellipses. Experiments with complete and disjoint superellipses, defective superellipses and superellipses extracted from a photograph of a real object indicate the efficiency, accuracy and reliability of the proposed method, both theoretically and practically View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Adaptive cancellation of selected harmonics from a signal

    Page(s): 295 - 303
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (728 KB)  

    The problem of extracting a cosine of unknown frequency in the presence of cosines with known frequencies is presented. Two methods are compared: the constant coefficient digital notch filter and the adaptive subtraction method where the known frequencies are input along with a guess for the unknown frequency. In the latter method, the amplitudes and phases of the known components are estimated using a Kalman filter and this information is used to subtract out these components from the signal to leave the cosine of interest. Two implementations of the Kalman filter are considered: an `optimal' method, where all the elements of the estimated error covariance matrix are kept and a `suboptimal' method, where the off-diagonal elements of this matrix are put to zero. Simulated and experimental data are analysed. The high-order Yule-Walker method is applied to determine the frequency content of the filtered signal. It is shown that, if the assumed frequency of the harmonic of interest is in error, then the suboptimal method can have a better performance than the optimal method. The reasons for this are explained by using a theoretical analysis of the estimation equations. The notch filter has by far the worst performance, as the frequencies to be subtracted out do not, in general, correspond to the zeros of the notch filter View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Fast face identification under varying pose from a single 2-D model view

    Page(s): 248 - 253
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (736 KB)  

    One of the key remaining problems in face recognition is that of handling the variability in appearance due to changes in pose. The authors present a simple and computationally efficient 3-D pose recovery methodology. It addresses the computationally expensive problem of current generic 3-D model pose recovery methods and thus is able to be used in real-time applications. Compared with the virtual view methods, the face identification system with the proposed pose recovery method demands much less storage space as it transforms the 2-D rotated face to the 2-D fronto-parallel view for subsequent identification rather than generating multiple virtual views for a single input face. Experiments evaluating the effectiveness of the technique are reported. The systems are compared with human performances and existing techniques View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Blind separation of spectral signatures in hyperspectral imagery

    Page(s): 217 - 226
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (1424 KB)  

    For the purpose of material identification, methods for exploring hyperspectral images with minimal human intervention have been investigated. Without any prior knowledge, it is extremely difficult to identify or determine how many endmembers in a scene. To tackle this problem, a new spectral unmixing technique, the spectral data explorer (SDE), is presented. SDE is a hybrid approach combining the optimal parts of fast independent component analysis (FastICA) and noise-adjusted principal components analysis (NAPCA). Experimental results show that SDE is highly efficient for separating significant signatures of hyperspectral images in a blind environment View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Robust M-estimate adaptive filtering

    Page(s): 289 - 294
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (528 KB)  

    An M-estimate adaptive filter for robust adaptive filtering in impulse noise is proposed. Instead of using the conventional least-square cost function, a new cost function based on an M-estimator is used to suppress the effect of impulse noise on the filter weights. The resulting optimal weight vector is governed by an M-estimate normal equation. A recursive least M-estimate (RLM) adaptive algorithm and a robust threshold estimation method are derived for solving this equation. The mean convergence performance of the proposed algorithm is also analysed using the modified Huber (1981) function (a simple but good approximation to the Hampel's three-parts-redescending M-estimate function) and the contaminated Gaussian noise model. Simulation results show that the proposed RLM algorithm has better performance than other recursive least squares (RLS) like algorithms under either a contaminated Gaussian or alpha-stable noise environment. The initial convergence, steady-state error, robustness to system change and computational complexity are also found to be comparable to the conventional RLS algorithm under Gaussian noise alone View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Coding efficiency of multi-ring and single-ring differential chain coding for telewriting application

    Page(s): 241 - 247
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (584 KB)  

    The availability of small personal digital assistants (PDAs) with a touchscreen and communication capabilities has been an influential factor in the renewed interest in telewriting, a technique for the exchange of handwritten information through telecommunications means. In this context, differential chain coding algorithms for compression of the handwritten ink are revisited. In particular, it is shown that the coding efficiency of multi-ring differential chain coding (MRDCC) is not always better when compared to single ring differential chain coding (DCC), as previously suggested. These algorithms were tested on over 300 handwritten messages using a relative compactness criterion and a per-length distortion measure. The probabilities of relative vectors in MRDCC and DCC are related, an expression for relative compactness in the MRDCC case is introduced, and the application of Freeman's (1974) criteria for the selection of the appropriate code for a family of curves is illustrated View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Automatic two-stage IR and MMW image registration algorithm for concealed weapons detection

    Page(s): 209 - 216
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (752 KB)  

    A two-stage registration scheme for the concealed weapons detection (CWD) problem is developed. The goal is to automatically register images taken simultaneously from two different (infrared (IR) and millimetre wave (MMW)) but parallel sensors whose lines of sight (LOS) are close to each other. The purpose of the first stage is to register the images coarsely. A feature-based image registration algorithm based on human body silhouettes is developed at this stage. The pose parameters found at this stage are used as the starting search point for the second stage of the registration algorithm. At the second stage, maximisation of the mutual information measure between IR and MMW images is performed to improve the pose parameters obtained at the first stage. Two-dimensional partial volume interpolation is employed to estimate the joint histogram that is needed to calculate mutual information (MI). The simplex search algorithm is utilised to maximise the MI measure. In both stages, the distortion between the two images is assumed to be a rigid body transformation. Experimental results indicate that the automated two-stage registration algorithm performs fairly well View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Bayesian self-organising map for Gaussian mixtures

    Page(s): 234 - 240
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (624 KB)  

    A Bayesian self-organising map (BSOM) is proposed for learning mixtures of Gaussian distributions. It is derived naturally from minimising the Kullback-Leibler (1951) divergence between the data density and the neural model. The inferred posterior probabilities of the neurons replace the common Euclidean distance winning rule and define explicitly the neighbourhood function. Learning can be retained in a small but fixed neighbourhood of the winner. The BSOM in turn provides an insight into the role of neighbourhood functions used in the common SOM. A formal comparison between the BSOM and the expectation-maximisation (EM) algorithm is also presented, together with experimental results View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Multirate signal processing on finite fields

    Page(s): 254 - 262
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (704 KB)  

    The paper addresses the problem of multirate signal processing over arbitrary fields. Studies of multirate systems and filter banks have proceeded in parallel, and a wealth of results are available in literature. The authors concentrate their attention on cyclic systems. These structures are ideally suited to generalising the concepts to finite fields. The perfect reconstruction property for quadrature mirror filter banks is obtained. It is shown how the cyclic wavelet transform (CWT) can be derived from such systems; the relationships between cyclic filter banks and CWTs are explored in detail. The results obtained are potentially very well suited for speech and image encoding, as well as for fast algorithms in signal processing View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Subband adaptive filtering with real-valued subband signals for acoustic echo cancellation

    Page(s): 283 - 288
    Save to Project icon | Click to expandQuick Abstract | PDF file iconPDF (480 KB)  

    Adaptive filtering in subbands was originally proposed to overcome the limitations of conventional least-mean-square (LMS) algorithms. In general, subband adaptive filters offer computational savings, as well as faster convergence over the conventional LMS algorithm. However, improvements to current subband adaptive filters could be further enhanced by a more elegant choice of their design/structure. Classical subband adaptive filters employ DFT-based analysis and synthesis filter banks which results in subband signals that are complex-valued. The authors modify the structure of subband adaptive filters by using single-sideband (SSB) modulated analysis and synthesis filter banks, which result in subband signals that are real-valued. This simplifies the realisation of subband adaptive filters View full abstract»

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