HomeChair: M. Orchard, Princeton University, USA
Home
Sang-Hwa Lee, Seoul National University (Korea)
Jong-Il Park, MIC3, ATR (Japan)
Choong-Woong Lee, Seoul National University (Korea)
In this paper we derive the general formula of Bayesian model for stereo matching algorithm and implement it with simplified probabilistic models. The probabilistic models are independence property and similarity between the neighborhood disparities in the configuration. The formula is the generalization of Bayesian model for stereo matching, and can be implemented into some differnet forms corresponding to to the probabilistic models in the configuration. We propose a new probabilistic model in order to simplify the joint probability distribution of disparities in the configuration. According to the experimental results, we can conclude that the derived formula generalizes the Bayesian model for stereo matching , and the simplified probabilistic modela reasonable and approximate the pure joint probability distribution very well. Compared with the conventional method of Bayesian model and sum of squared difference(SSD) algorithm, the proposed algorithm outperforms the other ones.
ic981430.pdf (From Postscript)
TOP
Lydia L Ng, Macquarie University (Australia)
Victor Solo, Macquarie University (Australia)
Although still in practice, the use of total least squares (TLS) in optical flow estimation is unreliable. TLS implicitly assumes that the error terms affecting the partial derivatives of the image intensities are independent. The usual methods for estimating the partial derivatives ensures that the errors are strongly correlated. Due to this correlation, an alternative method is required to treat the resulting errors-in-variables (EIV) problem. In this paper we propose a new method for estimating optical flow based on Sprent's procedure. This method incorporates a general EIV model and provides a far simpler computational procedure than found in previous solutions.
ic981098.pdf (From Postscript) TOP
Mingqi Kong, Washington University in Saint Louis (U.S.A.)
Jean-Pierre Leduc, Washington University in Saint Louis (U.S.A.)
Bijoy K Ghosh, Washington University in Saint Louis (U.S.A.)
Jon Corbett, Washington University in Saint Louis (U.S.A.)
Victor M Wickerhauser, Washington University in Saint Louis (U.S.A.)
This paper addresses the problem of estimating, analyzing and tracking objects moving with spatio-temporal rotational motion ( spin or orbit). It is assumed that the digital signals of interest are acquired from a camera and structured as digital image sequences. The trajectories in the signal are two-dimensional spatial projection in time of motion taking place in a three-dimensional space. The purpose of this work is to focus on the rotational motion i.e. estimate the angular velocity. In natural scenes, the rotational motion usually composes with translational or accelerated motion on a trajectory. In thispaper, we show that the trajectory can be estimated and tracked either simultaneously or separately from the rotational motion and that the analysis of the trajectory and the rotational motion can be done efficiently. The final goal of this work is to provide selective reconstructions of moving objects of interest. This paper constructs new continuous wavelet transforms that can be tuned to both translational and rotational motion. The link between rotational motion, symmetry and critical sampling is also presented. Applications are presented with tracking and estimation. The parameters of analysis that are taken into account in these rotational wavelet transforms are translation (space and time), velocity, spatial scale, angular position and angular velocity. the continuous wavelet functions are finally discretized for signal processing.
ic981406.pdf (From Postscript) TOP
Jean-Pierre Leduc, Washington University in Saint Louis (U.S.A.)
Jon Corbett, Washington University in Saint Louis (U.S.A.)
Mingqi Kong, Washington University in Saint Louis (U.S.A.)
Victor M Wickerhauser, Washington University in Saint Louis (U.S.A.)
Bijoy K Ghosh, Washington University in Saint Louis (U.S.A.)
This paper addresses the problem of estimating and analyzing accelerated motion in spatio-temporal discrete signals. It is assumed that the digital signals of interest are acquired from imaging sensors and structured as digital image sequences. The motion trajectories in the signal are two-dimensional spatial projections in time of three-dimensional motions. onsequently, they contain all the orders of acceleration. The purpose of this work is to estimate the trajectory and the motion parameters of selected moving objects in the scene. The final goal is to provide selective reconstructions of accelerated objects of interest. This paper presents the construction of new continuous wavelet transforms that can be tuned to any order of accelerations, we demonstrate their existence and provide the related admissibility conditions.The parameters for analysis that are taken into account in these accelerated wavelet transforms are spatial and temporal translations, velocity, acceleration (second or nth order), spatial scale and angular orientation. The continuous wavelet functions are finally discretized for signal processing.
ic981417.pdf (From Postscript) TOP
George A Tsihrintzis, Northeastern University (U.S.A.)
Anthony J Devaney, Northeastern University (U.S.A.)
E. Heyman, Tel-Aviv University (Israel)
We derive an efficient algorithm for the computation of the maximum likelihood estimate of the location of a known target from short pulse wave scatter data. The algorithm constitutes a three step procedure: (i) convolutional data filtering, (ii) time-domain backpropagation, and (iii) summation and consists of a number of projection and backprojection operations integrated in a tomographic scheme. A computer simulation is included for illustration purposes and relevant applications in radar target identification and buried object detection are discussed.
ic982297.pdf (Scanned) TOP
Carlos H Morimoto, IBM Almaden Research Center (U.S.A.)
Rama Chellappa, University of Maryland (U.S.A.)
Several techniques for electronic image stabilization have recently been proposed, but very little research has been done tocompare and evaluate such techniques.In this paper we propose a set of measures to evaluate image stabilization algorithms based of their fidelity, displacement range, and performance. These measures do not require calibration or ground truth, making the evaluation procedure very simple and flexible, i.e., it provides the means to compare techniques based on different motion models. We have used this procedure to compare several image stabilization algorithms and also evaluate the sensitivity of these algorithms to some of its parameters. These same procedures could also be used for the comparison and evaluation of motion estimation and image registration techniques.
ic981849.pdf (From Postscript) TOP
Marius C Vasiliu, Paris-Sud University (France)
Bertrand Zavidovique, Paris-Sud University (France)
We propose a global method to match pair of images using the similarity information. Using a generic similarity distance between pixel pairs, this method can match any kind of images (gray levels, RGB, IR) or more generally any pair of 2D matrix (like spectrogram or wavelet transformations). Our algorithms search the best matching map in a 4-dimensional space defined by the Cartesian product of two input images. Several parameters like topological cost functions or global minimum search method can be adapted, function of specific applications. One of the proposed search method is an original extension of dynamic programming in 4D space. Other methods like iterated global searching or simulated annealing are proposed and their performances are compared. Typical applications are 3D stereo reconstruction, optical flow and velocity field computing or (sub-)pixel texture stretching measurement.
ic982056.pdf (From Postscript) TOP
Sachin G Deshpande, University of Washington (U.S.A.)
Jenq-Neng Hwang, University of Washington (U.S.A.)
We present a new fast motion estimation method useful for high speed video encoding. Most of the motion estimation methods for video coding can be classified as Block Matching (BM) methods or Pel Recursive (PR) methods. Majority of the current fast motion estimation methods belong to block matching category. These methods try to reduce the number of search locations. Our proposed method is based on the pel recursive formulation. However, in order to achieve fast estimation, we operate on a block of pixels using a Total Least Squares (TLS) based estimation scheme which tries to estimate the true motion vector for each block. The major advantages of the proposed method include very fast estimation,almost constant time for motion estimation for all the video sequences, fractional pel accuracy, and better performance for noisy sequences. We present extensive simulation results to illustrate the performance of the proposed method.
ic982206.pdf (From Postscript) TOP
Geert Van der Auwera, VUB, ETRO/IRIS Research Group (Belgium)
Adrian Munteanu, VUB, ETRO/IRIS Research Group (Belgium)
Gauthier Lafruit, IMEC - Leuven (Belgium)
Jan Cornelis, VUB, ETRO/IRIS Research Group (Belgium)
This work proposes a new block based motion estimation and compensation technique applied on the detail images of the wavelet pyramidal decomposition. The algorithm uses two matching criteria, namely the absolute difference and the absolute sum. For a wavelet decomposed one-dimensional step function, it is shown that for odd translations of the step, the absolute sum reaches a smaller minimum than the absolute difference. We also derive in this case a constraint on the highpass filter coefficients so that a zero prediction error can be reached by using the absolute sum. Although this cannot be easily generalized for an arbitrary signal profile, experimental results obtained with photorealistic image sequences indicate that the prediction error can be reduced with respect to techniques that only use the absolute difference as matching criterion.
ic982423.pdf (From Postscript) TOP
Robert J. Fergusson, University of Strathclyde, Scotland (U.K.)
John J. Soraghan, University of Strathclyde, Scotland (U.K.)
In MPEG-4 that uses Video Object Planes (VOP's), systems to efficiently extract objects are required.Based on research into the Human Visual System (HVS), this paper presents an Object Extraction System that uses Motion Segmentation and Image Segmentation in parallel, and then combines them to efficiently extract objects using Fuzzy Reasoning. The most computationally demanding part of this system, the Motion Segmentation (MS), is implemented on the TMS320C80 DSP MVP. The Motion Estimation part of the MS algorithm and its fast implementation on C80 architecture are also described.
ic981592.pdf (From Postscript) TOP