Home Home
Jorge E. Pérez-Jácome, Georgia Institute of Technology (U.S.A.)
Vijay K. Madisetti, Georgia Institute of Technology (U.S.A.)
A method to automatically detect targets from sets of pixel-registered visual, thermal, and range images is outlined. It uses operations specifically designed to work on the different kinds of images to explote the information given by each of them. Five features are used to distinguish the targets from the clutter: texture, brightness, temperature, surface planarity, and height. The results from individual detectors are then combined to improve the detection rate while reducing the number of false alarms. A morphological operation called ``erosion of strength $n$'' is also introduced and utilized as a powerful tool for removal of spurious information. The excellent results obtained for detection support the suitability of this approach for other ATR (Automatic Target Recognition) problems.
ic972741.pdf ic972741.pdf
TOPCassandra T. Swain, AT&T Labs (U.S.A.)
This paper presents a novel approach for using monocular cues in a single 2D image to improve depth perception. Monocular depth cues--blur, shading, brightness, and occlusion--are applied to 2D images. The contribution of the first three cues to depth perception is additive, each weight being equivalent. Since occlusion cues modify the object geometry, they are applied after the application of other cues. Results show that monocular depth cues can successfully improve depth perception in a single 2D image, creating a pseudo 3D image. The advantage of this approach is that it requires a single image, rather than an image pair used in traditional methods. The main limitation is that only depth perception, not precise depth measurements, is possible. This work looks very promising for low bitrate video coding and other applications where bandwidth is limited.
ic972745.pdf ic972745.pdf TOP
André Redert, Delft University of Technology (The Netherlands)
Emile Hendriks, Delft University of Technology (The Netherlands)
Jan Biemond, Delft University of Technology (The Netherlands)
In this paper we present an algorithm for the synthesis of multi viewpoint images at non-intermediate positions, based on stereoscopic images. We consider the synthesis of images from virtual camera positions and the synthesis of images for scene reconstruction using stereo displays. The algorithm provides scene reconstruction without geometric distortion and without any restriction to the position of the viewer. All synthesized images are based on extrapolation of a single source image and a single disparity field. This provides low use of bandwidth and compatibility with mono video systems. With teleconferencing images, the generated views were subjectively evaluated as good for viewing positions not more than one half camera baseline from the centre position. Objectively, reconstructed left and right images have PSNR values of 41 dB.
ic972749.pdf ic972749.pdf TOP
Vasudev Parameswaran, University of Maryland (U.S.A.)
Philippe Burlina, University of Maryland (U.S.A.)
Rama Chellappa, University of Maryland (U.S.A.)
Robustness as well as the ability to work in an unsupervised mode are two desirable features of algorithms employed on large image databases. This paper describes parameter optimization strategies for such algorithms and motivates these strategies by focussing on aerial image exploitation and studying certain specific aerial image understanding algorithms, namely local vehicle detection and global vehicle configuration detection. The paper first gives a brief introduction to the problem in the context of aerial imagery. Next, a high level description of the algorithms and parameters that need to be optimized is given. Strategies for parameter optimization are illustrated using examples. Finally a discussion on the applicability and scope for improvement of the strategies is given.
ic972753.pdf ic972753.pdf TOP
Arnaldo J. Abrantes, ISEL/INESC (Portugal)
Jorge S. Marques, IST/INESC (Portugal)
Deformable models have been intensively investigated during the last decade. Several well known algorithms, proposed in other contexts can also be included in this class (e.g., Kohonen maps, elastic nets and fuzzy c-means). In all these methods the model parameters are obtained in a deterministic framework by the minimization of an energy function. This paper proposes a novel class of probabilistic shape models related to the unified framework presented in [1]. Shape modelling is addressed as a MAP estimation problem, by assuming that the image features are random variables with Gibbs-Boltzmann distribution, and provides extensions for several well known algorithms. The main difference between the proposed algorithms and the original ones lies on the partition function which depends on the model parameters and influences the shape estimates. For example, it is shown that in snakes the partition function generates short-range repulsive forces between the model units which prevent their collapse when they are attracted by common data.
ic972757.pdf ic972757.pdf TOP
Roger Fjørtoft, ENSEEIHT (France)
Philippe Marthon, ENSEEIHT (France)
Armand Lopès, CESBIO (France)
Eliane Cubero-Castan, CNES (France)
Edge detection is a fundamental issue in image analysis. Due to the presence of speckle, which can be modelled as a strong multiplicative noise, edge detection in Synthetic Aperture Radar (SAR) images is very difficult and methods developed for optical images are inefficient. We here propose a new edge detector for SAR images which is optimum in the MSSE sense for a stochastic multiedge model. It computes a normalized Ratio Of Exponentially Weighted Averages (ROEWA) on opposite sides of the central pixel. This is done in the horizontal and vertical direction, and the magnitude of the two components yields an edge strength map. Thresholding of the edge strenght map and postprocessing to eliminate false edges are briefly discussed. We present results on simulated SAR images and ERS1 data.
ic972761.pdf ic972761.pdf TOP
Alexandre Winter, ENST (France)
Henri Maître, ENST (France)
Nicole Cambou, Service Vision, E/ETS/V, Aérospatiale (France)
Eric Legrand, Service Vision, E/ETS/V, Aérospatiale (France)
This paper presents a new, fully automatic and robust feature extraction algorithm based on the selection of a given range of scales. It compares consecutive band-pass images of a Gaussian multiscale decomposition to extract the objects that appear between given scales. The comparison is performed using original distributed entropic measures. The application to building detection in aerial images shows that scale is a robust and precise criterion for the detection of man-made objects. They also show that distributed entropic tools are relevant for the comparison of band-pass images. From a more theoretical point of view, this method stands between Scale-space and wavelet approaches. It tries to infer the geometrical conception of scale found in the Scale-space theory into the algebraic scale of the wavelet theory.
ic972765.pdf ic972765.pdf TOP
Abbie L. Warrick, Lawrence Livermore National Laboratory (U.S.A.)
Pamela A. Delaney, Wichita State University (U.S.A.)
One problem of interest to the oceanic engineering community is the detection and enhancement of internal wakes in open water synthetic aperture radar (SAR) images. Internal wakes, which occur when a ship travels in a stratified medium, have a ``V'' shape extending from the ship, and a chirp-like feature across each arm. The Radon transform has been applied to the detection and the enhancement problems in internal wake images to account for the linear features while the wavelet transform has been applied to the enhancement problem in internal wake images to account for the chirp-like features. In this paper, a new transform, a localized Radon transform with a wavelet filter (LRTWF), is developed which accounts for both the linear and the chirp-like features of the internal wake. This transform is then incorporated into optimal and sub-optimal detection schemes for images (with these features) which are contaminated by additive Gaussian noise.
ic972769.pdf ic972769.pdf TOP
Aparecido Nilceu Marana, IGCE - UNESP (Brazil)
Luciano da Fontoura Costa, IFSC - USP (Brazil)
Sergio A. Velastin, EEE - KCL (U.K.)
Roberto A. Lotufo, FEE - UNICAMP (Brazil)
This paper presents a technique for oriented texture classification which is based on the Hough transform and Kohonen's neural network model. In this technique, oriented texture features are extracted from the Hough space by means of two distinct strategies. While the first operates on a non-uniformly sampled Hough space, the second concentrates on the peaks produced in the Hough space. The described technique gives good results for the classification of oriented textures, a common phenomenon in nature underlying an important class of images. Experimental results are presented to demonstrate the performance of the new technique in comparison with an implemented technique based on Gabor filters.
ic972773.pdf ic972773.pdf TOP
Yves Goussard, École Polytechnique, Montréal (Canada)
Jérôme Idier, Lab. Signaux et Systèmes (France)
Alain DeCesare, Lab. Signaux et Systèmes (France)
This communication presents a non-supervised segmentation method based upon a discrete-level unilateral Markov field model of the image. Such models have been shown to yield numerically efficient algorithms, for segmentation and for hyperparameter estimation as well. Our contribution lies in the derivation of a parsimonious {telegraphic} parameterization of the unilateral Markov field. On a theoretical level, this parameterization ensures that some important properties of the field (e.g., stationarity) do hold. On a practical level, it reduces the computational complexity of the algorithm used in the segmentation and parameter estimation stages of the precedure. In addition, it decreases the number of hyperparameters that must be estimated, thereby improving convergence speed and accuracy of the corresponding estimation method.
ic972777.pdf ic972777.pdf TOP
Max Mignotte, Ecole Navale (France)
Christophe Collet, Ecole Navale (France)
Patrick Perez, IRISA/INRIA (France)
Patrick Bouthemy, IRISA/INRIA (France)
This work deals with unsupervised sonar image segmentation. We present a new estimation segmentation procedure using the recent iterative method of estimation called Iterative Conditional Estimation (ICE). This method takes into account the variety of the laws in the distribution mixture of a sonar image and the estimation of the parameters of the label field (modeled by a Markov Random Field (MRF)). For the estimation step we use a maximum likelihood estimation for the noise model parameters and the least square method proposed by Derin et al. to estimate the MRF prior model. Then, in order to obtain a good segmentation and to speed up the convergence rate, we use a multigrid strategy with the previously estimated parameters. This technique has been sucessfully applied to real sonar images and is compatible with an automatic treatment of massive amounts of data.
ic972781.pdf ic972781.pdf TOP