A heuristic method has been developed for registering two sets of 3-D curves obtained by using an edge-based stereo system, or two dense 3-D maps obtained by using a correlation-based stereo system. Geometric matching in general is a difficult unsolved problem in computer vision. Fortunately, in many practical applications, some a priori knowledge exists which considerably simplifies the problem. In visual navigation, for example, the motion between successive positions is usually approximately known. From this initial estimate, our algorithm computes observer motion with very good precision, which is required for environment modeling (e.g., building a Digital Elevation Map). Objects are represented by a set of 3-D points, which are considered as the samples of a surface. No constraint is imposed on the form of the objects. The proposed algorithm is based on iteratively matching points in one set to the closest points in the other. A statistical method based on the distance distribution is used to deal with outliers, occlusion, appearance and disappearance, which allows us to do subset-subset matching. A least-squares technique is used to estimate 3-D motion from the point correspondences, which reduces the average distance between points in the two sets. Both synthetic and real data have been used to test the algorithm, and the results show that it is efficient and robust, and yields an accurate motion estimate.

Abstract not found

Experiments are presented showing that visual search for Mueller-Lyer (ML) stimuli is based on complete configurations, rather than component segments. Segments easily detected in isolation were difficult to detect when embedded in a configuration, indicating preemption by low-level groups. This preemption—which caused stimulus components to become inaccessible to rapid search—was an all-ornothing effect, and so could serve as a powerful test of grouping. It is shown that these effects are unlikely to be due to blurring by simple spatial filters at early visual levels. It is proposed instead that they are due to more sophisticated processes that rapidly bind contour fragments into spatially-extended assemblies. These results support the view that rapid visual search cannot access the primitives formed at the earliest stages of visual processing; rather, it can access only higher-level, more ecologically-relevant structures. The processes that underlie human vision are often divided into two fundamentally different classes: operations that are carried out in parallel over space, and operations that are not (e.g., Neisser, 1967; von Helmholtz, 1867/1962). For the most part, parallel processes are rapid (i.e., they occur within a few hundred milliseconds), effortless, and automatic (i.e., they cannot be affected by immediate changes in higher-level goals), whereas nonparallel processes are slower, more effortful, and nonautomatic. In its current embodiment, this dichotomy divides vision into an early preattentive and a subsequent attentive stage (e.g.,

Recognition of Handwritten Numerals Using Elastic Matching Patrice Scattolin Elastic matching has been used for the recognition of handwritten characters for two decades. It is usually only used for writer-dependent systems with on-line data. We attempt to use this method in a multi-writer environment for both on-line and off-line recognition of handwritten numerals. By its nature, elastic matching is best suited to single writer on-line systems. Two challenges present themselves to attain reasonable results under these conditions. First, the algorithm must be modified to better generalize the models, to recognize a wider variety of patterns with a given number of models. Secondly the off-line data is not in a suitable format as the patterns are not represented by a sequence of ordered points. We will apply two modifications to the typical elastic matching system to adapt it to the multi-writer environment and for the off-line data. To process the off-line data, we use a stroke recon...

Abstract. We present a novel approach to parameterised curve detection. The method is based on the generalised Radon transform, which is traditionally applied to a 2D edge/line map. The novelty of our method is the mapping of the original 2D image to a 3D orientation space, which then forms the input for the Radon transform. The orientation space representation can represent multiple intersecting structures and contains local orientation information. We demonstrate our approach on a problem in geology and show that we can detect curves in a heterogeneous and noisy background. 1

The orientation space transform is a concept that can deal with multiple oriented structures at a single location. In this paper we extend the orientation space transform to 3D images producing a 5D orientation space (x, y, z, # , # ). We employ a tunable, orientation selective quadrature filter to detect edges and planes and a separate filter for detecting lines. We propose a multi-resolution sampling grid based on the icosahedron. We also propose a method to visualize the resulting 5D space. The method can be used in many applications like (parametric) curve and plane extraction, texture characterization and curvature estimation. 1

In this thesis, we present two novel methods for medical image volume segmen- tation and surface registration. The volume segmentation is conceptually formulated as a problem of clustering feature vectors representing each voxel. Feature patterns are constructed by extracting texture measures and multiscale parameters for each voxel. These feature vectors are then projected onto their leading principal axes found by using principal components analysis (PCA). The number of principal com- ponents is selected dynamically using genetic algorithms (GAs). This step provides an effective basis for feature extraction. The reduced patterns are then clustered to different, spatially connected regions using a novel adaptive connectivity satisfaction self-organizing feature map (CSSOFM). This network, which is a type of Kohonen feature map, combines clustering and labeling in one network. Topological constraints are imposed on the clustering algorithm so that only voxels that are connected to each other are grouped together in a certain class. The choice of the optimum number of classes is performed automatically by maximizing a segmentation quality measure. The algorithm's performance was tested on both simulated and actual medical data sets. In both simulation studies and practical medical image segmentation, the system shows promising results in comparison with two well-known methods: the competi- tive Hopfield neural network (CHNN) and ISODATA methods.

Line primitive extraction is one of the most important processes at the low level stage of line analysis in document image recognition. This paper proposes a new interpretive model of line continuation and applies it to line primitive extraction. The proposed model includes interpretation of line connectivity and line smoothness on the raster line images. The combination of these two features is the main contribution of this paper. Testing results of synthesized and real line images show that the proposed model performs well to separate X-type and T-type junctions and to group Ltype junctions. The generality of the proposed model is exhibited by its capability of extracting various raster line primitives, such as straight lines, curves and zigzag lines. 1.

The image of a word or a generic hand made drawing on a piece of paper is usually characterized by a series of interfering zones where the cursive trace intersects itself or printed lines already present on the writing surface. In this zone, the odometric information is ambiguous and any trivial inference on the original pen tip movement cannot be done. In this article, starting from some basic cognitive considerations, a general procedure is developed to analize a generic image of a word or a common hand made scribble. This approach allows to detect each ambiguity part of the image and then interpretate them to finally recover a part of the original temporal information. ii 1 Introduction The generation of a graphic shape on a writing surface during handwriting or drawing is the result of a complex motor planning process starting from an input allograph representation of the shape and then producing a sequence (partially overlapped) of primitive movements (called motor strokes) of t...

In this proposal a new method for content-based retrieval of digital video is proposed based on a neurophysiological model. Effective content-based retrieval requires access to video objects which represent syntactic structures in a video sequence such as people, cars, shots and scenes. Existing techniques for extracting video objects are based on inverse optics, where three-dimensional objects are mapped to a two dimensional image. These techniques have proven to be limited in extracting perceptually significant visual objects. Extracting human identifiable objects from scenes is important as these will be the objects that humans will use for searching. Research in the field of psychology has made inroads into determining the process of human vision. Current literature provides computer simulations of low-level processes of vision as well as theories for high-level object detection. Current work in the fields of psychology relating to vision has been motivated by the need for a comple...