Medical imaging is shifting from film to electronic images. The STARE (structured analysis of the retina) system is a sophisticated image management system that will automatically diagnose images, compare images, measure key features in images, annotate image contents, and search for images similar in content. We concentrate on automated diagnosis. The images are annotated by segmentation of objects of interest, classification of the extracted objects, and reasoning about the image contents. The inferencing is accomplished with Bayesian networks that learn from image examples of each disease. This effort at image understanding in fundus images anticipates the future use of medical images. As these capabilities mature, we expect that ophthalmologists and physicians in other fields that rely in images will use a system like STARE to reduce repetitive work, to provide assistance to physicians in difficult diagnoses or with unfamiliar diseases, and to manage images in large image databases. 1.

This paper describes steps towards an information system for the storage and content-based retrieval of ocular fundus images. Based on the Virage Incorporated framework for defining similarity metrics, the authors have developed a number of primitives for the representation of ocular fundus images. A prototype Query By Pictorial Example (QBPE) system yields similarity rankings in approximate agreement with those of a human expert.

No abstract available.

No abstract available.

In this paper, a statistical error analysis is developed to show how the performance of a stereo ranging system relies on system parameters. Formulas for expected absolute depth error, expected relative depth error, and lower and upper bounds of depth error are derived. We show that for a given convergence angle, a simple, but important, parameter called generalized depth resolvability completely determines the performance of the system. These results are useful for the design of fixed ranging systems in constrained environments, or for the active selection of system parameters in "purposive" stereo systems.>

In this work, the correspondence problem in stereo vision is handled by matching two sets of dense feature vectors. Inspired by biological evidence, these feature vectors are generated by a correlation between a bank of Gabor sensors and the intensity image. The sensors consist of two-dimensional Gabor filters at various scales (spatial frequencies) and orientations, which bear close resemblance to the receptive field profiles of simple V1 cells in visual cortex. A hierarchical, stochastic relaxation method is then used to obtain the dense stereo disparities. Unlike traditional hierarchical methods for stereo, feature based hierarchical processing yields consistent disparities. To avoid false matchings due to static occlusion, a dual matching, based on the imaging geometry, is used.

The authors consider the problem of occlusion in computing stereo disparity from a pair of images. Usual approaches to stereo matching, e.g. area-based, feature-based, etc., can perform poorly in the neighborhood of occluding boundaries if no surface models are assumed. Qualitative improvements are possible based on conditions derived from the geometry of occlusion. A pair of correspondence processes, using information near an occlusion boundary to localize the boundary, can determine the sign of associated depth discontinuity unambiguously. The proposed method is able to identify the presence and extent of occlusion regions and assign disparities in a consistent way near the occlusion regions.>

An algorithm is presented to obtain the total least squares (TLS) estimates of the motion parameters of an object from range/stereo data or perspective views in a closed form. TLS estimates are suitable when data in both time frames are corrupted by noise, which is an appropriate model for motion analysis in practice. The robustness of different linear least squares methods is analyzed for the estimation of motion parameters against the sensor noise and possible mismatches in establishing object feature point correspondence. As the errors in point correspondence increase, the performance of an ordinary least squares (LS) estimator was found to deteriorate much faster than that of the TLS estimator. The Cramer-Rao lower bound (CRLB) of the error covariance matrix was derived for the TLS model under the assumption of uncorrelated additive Gaussian noise. The CRLB for the TLS model is shown to be always higher than that for the LS model.>

A fast algorithm is presented for a globally smooth interpolation for a visual surface from scattered range data. This method is based on matching lower order spatial moments, where the reconstructed surface is given by a linear combination of Legendre polynomials. This method can handle data that do not lie on a regular grid. The value of the function at points where the grid is not defined, or where the original data is missing, can be approximated from the constituent polynomials. Some simulation experiments evaluate the performance of the proposed scheme. An arbitrarily shaped surface is shown on a 35*35 grid, and the depth information is available only at a few selected grid points. The result of the implementation of Grimson's (1981) algorithm where the surface is constrained to pass through the given data set is given. The proposed method runs much faster than Grimson's algorithm. The coefficients of the Legendre polynomials can be precomputed.>

A Bayesian approach is proposed for stereo matching to derive the maximum a posteriori estimation of depth. How a pyramid data structure can be combined with simulated annealing to speed up convergence in stereo matching is described. Using the invariant property of image intensity and modeling the disparity as a Markov random field (MRF), the pyramid structure is followed from high (coarse) level to low (fine) level to derive the maximum a posteriori estimates. Simulation results on both random dot diagrams and synthesized images show the promise of this multiresolution stereo approach.>

Maximum a posteriori (MAP) estimation of depth based on the Bayesian approach is described. The authors investigate simulated annealing in a hierarchical pyramid to accelerate the convergence. They also utilize an unobservable line process to recover the piecewise continuous depth map over the image plane. Simulation results are provided using random dot diagrams as well as synthesized images

If the correspondence between two sets of points representing the coordinates of different points of an object undergoing rotational motion and deformation is known, the parameters can be estimated using different least-squares estimators. The total-least-squares (TLS) method is very appropriate when the observation and the data matrices are both perturbed by random noise. For Gaussian-distributed noise, the TLS solution is equivalent to maximum-likelihood estimation. The mean-square error in TLS is always smaller than in an ordinary least-squares (LS) estimator. The scope is analyzed of TLS in estimating the generalized motion parameters, as is the feasibility of decomposing the generalized motion parameters in terms of rotation and deformation parameters. The performance of TLS is compared to that of the LS estimator.>

The interpretation is presented of image flow (optical flow) and its relationship with object structure. Depth extraction is described from stereo image flow when the stereo cameras have parallel optical axes or nonparallel optical axes. Using epipolar geometry, the relationship between the optical flow fields on the image planes (stereo) is derived and applied for matching the images to extract the depth information. Three methods are considered, of which the first two assume parallel optical axes: (1) spatial approach which is the same as the traditional stereo method except that binocular difference flow is used as primitives for matching; (2) spatial-temporal approach which extends previous formulation to incorporate temporal variations of image flow; and (3) under nonparallel optical axes assumption from which the model of the stereo cameras is generalized so that they are arbitrarily separated in space. Then two measures are used to evaluate how well the epipolar constraint is satisfied.>