Robert J. Schalkoff

Lattice-Boltzmann lighting

A new technique for lighting participating media is suggested. The technique is based on the lattice-Boltzmann method, which is gaining popularity as alternative to finite-element methods for flow computations, due to its ease of implementation and ability to handle complex boundary conditions. A relatively simple, grid-based photon transport model is postulated and then shown to describe, in the limit, a diffusion process. An application to lighting clouds is provided, where cloud densities are generated by combining two well-established techniques. Performance of the new lighting technique is not real-time, but the technique is highly parallel and does offer an ability to easily represent complex scattering events. Sample renderings are included.

Standardized database development for EEG epileptiform transient detection: EEGnet scoring system and machine learning analysis

The routine scalp electroencephalogram (rsEEG) is the most common clinical neurophysiology procedure. The most important role of rsEEG is to detect evidence of epilepsy, in the form of epileptiform transients (ETs), also known as spike or sharp wave discharges. Due to the wide variety of morphologies of ETs and their similarity to artifacts and waves that are part of the normal background activity, the task of ET detection is difficult and mistakes are frequently made. The development of reliable computerized detection of ETs in the EEG could assist physicians in interpreting rsEEGs. We report progress in developing a standardized database for testing and training ET detection algorithms. We describe a new version of our EEGnet software system for collecting expert opinion on EEG datasets, a completely web-browser based system. We report results of EEG scoring from a group of 11 board-certified academic clinical neurophysiologists who annotated 30-s excepts from rsEEG recordings from 100 different patients. The scorers had moderate inter-scorer reliability and low to moderate intra-scorer reliability. In order to measure the optimal size of this standardized rsEEG database, we used machine learning models to classify paroxysmal EEG activity in our database into ET and non-ET classes. Based on our results, it appears that our database will need to be larger than its current size. Also, our non-parametric classifier, an artificial neural network, performed better than our parametric Bayesian classifier. Of our feature sets, the wavelet feature set proved most useful for classification.

Image labelling: a neural network approach

Abstract The solution of the image labelling problem using the emerging computational paradigm of neural networks is shown. A brief introduction to neural network technology is provided. The labelling problem is formulated as a problem in symbolic constraint satisfaction. Alternative solution methods are cited. A Hopfield neural network structure which embodies the labelling constraints is developed in detail. The procedure to determine the energy function and interconnection weight is described. Experimental results and network convergence properties are analysed. Future research diections are outlined.

ANN implementation of stereo vision using a multi-layer feedback architecture

An artificial neural network (ANN), consisting of three interacting neural modules, is developed for stereo vision. The first module locates sharp intensity changes in each of the images. The edge detection process is basically a bottom-up, one-to-one input-output mapping process with a network structure which is time-invariant. In the second module, a multilayered connectionist network is used to extract the features or primitives For disparity analysis (matching). A similarity measure is defined and computed for each pair of primitive matches and is passed to the third module. The third module solves the difficult correspondence problem by mapping it into a constraint satisfaction problem. Intra- and inter-scanline constraints are used in order to restrict possible feature matches. The inter-scanline constraints are implemented via interconnections of a three-dimensional neural network. The overall process is iterative. At the end of each network iteration, the output of the third constraint satisfaction module feeds back updated information on matching pairs as well as their corresponding location in the left and right images to the input of the second module. This iterative process continues until the output of the third module converges to a stable state. Once the matching process is completed, the disparity can be calculated, and camera calibration parameters can be used to find the three-dimensional location of object points. Results using this computational architecture are shown. >

Control Implications in Tracking Moving Objects Using Time-Varying Perspective-Projective Imagery

Control implications which arise when tracking moving objects contained in time-varying perspective-projective imagery are studied. First, a transformation is derived to relate pan/tilt camera mount movement to image plane perturbations. Ramifications of this model, particularly with respect to magnification ratio, noncentered targets and camera mount angular magnitudes are illustrated. A pan/tilt control algorithm is then developed using this transformation and a large magnification ratio assumption. Experimental results with real imagery are shown to confirm the approach validity.

Dynamic Imagery Modeling and Motion Estimation Using Weak Formulations

Basic research into the modeling and analysis of image sequence dynamics is presented. A distributed parameter system (DPS) framework and the concept of weak solutions are used to develop image motion estimation algorithms. These algorithms represent a region-oriented (as opposed to point-by-point) approach to image motion analysis which is theoretically justifiable, computationally advantageous, and leads to interesting extensions. Particularly noteworthy is the use of weak solution-based motion features to obtain static image structural information and multiple object motion estimates. Experimental results confirm the validity and accuracy of the approach. Future research topics are described.

Morphology-based wavelet features and multiple mother wavelet strategy for spike classification in EEG signals

New wavelet-derived features and strategies that can improve autonomous EEG classifier performance are presented. Various feature sets based on the morphological structure of wavelet subband coefficients are derived and evaluated. The performance of these new feature sets is superior to Gulers classic features in both sensitivity and specificity. In addition, the use of (scalp electrode) spatial information is also shown to improve EEG classification. Finally, a new strategy based upon concurrent use of several mother wavelets is shown to result in increased sensitivity and specificity. Various attempts at reducing feature vector dimension are shown. A non-parametric method, k-NNR, is implemented for classification and 10-fold cross-validation is used for assessment.

Direct surface parameter estimation using structured light: a predictor-corrector based approach

The projection of structured light is a technique frequently used in computer vision to determine the surface structure of scene objects. In this work, higher level features are extracted from the images and used for a direct estimation of second-order object surface model parameters. In particular, a class of cylinders is emphasized, due to an underlying application involving the inspection of stored drums containing mixed waste. The strategy is based upon a predictor-corrector approach which utilizes an initial estimate for the surface parameters, followed by iterative parameter refinement. A predicted passive image is generated using the current surface parameter estimates and significant features are extracted and compared with those in the true passive image. The estimated surface parameters are corrected based upon feature disparities. In computer simulations and laboratory experiments using real image data, the algorithm was found to converge quickly and to yield accurate results.

Stereo vision: a neural network application to constraint satisfaction problem

In this paper, a stereo vision matching algorithm, implemented via a neural network architecture, is described. The stereo matching problem, that is, finding the correspondence of features between two images, can be cast as a constraint satisfaction problem. The algorithm uses image edge features and assumes a parallel-axis camera geometry such that the corresponding image points must lie in the same scanline. Intra-scanline constraints are used to to perform multipleconstraint satisfaction searches for the correct match. Further, inter-scanline constraints are used to enforce consistent matches by eliminating those that are not getting enough support from the neighboring scanlines. The inter-scanline constraints are implemented in a 3-D neural network which is formed by a stack of 2-D neuron layers. First, a mulilayered network is designed to extract the features points for matching using a static neural network. A similarity measure is defined for each pair of feature point matches which are then passed on to the second stage of the algorithm. The purpose of the second stage is to turn the difficult correspondence problem into a constraint satisfaction problem by imposing relational constraints. The result of computer simulations are presented to demonstrate the effectiveness of the approach.

Edge detection and thinning in time-varying image sequences using spatio-temporal templates

Abstract A new computational paradigm for the extraction of spatio-temporal edge information in a sequence of images is presented. Extracted edge information consists of an edge type, orientation, and motion class. A set of three-dimensional (two-dimensional spatial and one-dimensional temporal) templates is developed. The extracted edge features facilitate efficient thinning of edge data. Sample experimental results are shown. The proposed approach forms the preprocessing stage of an image motion understanding system.