Computer vision

The field of computer vision is devoted to discovering algorithms, data representations, and computer architectures that embody the principles underlying visual capabilities. This article describes how the field of computer (and robot) vision has evolved, particularly over the past 20 years, and introduces its central methodological paradigms.

Mechanisms of attentional selection: temporally modulated priority tags

When a single abrupt onset appears in a multielement display, it captures attention. When multiple onset elements occur, they have conditional priority over no-onset elements such that a limited number of onsets can be serviced with high priority in visual search (Yantis & Johnson, 1990). We report three experiments in which we assess two possible mechanisms for attentional prioritization: a priority queue into which a fixed number of high-priority elements are placed for early servicing during search, and a mechanism that temporarily tags all high-priority elements for early servicing or more frequent sampling. We manipulated the visual quality or inter-letter confusability of the stimuli to prolong encoding and/or comparison operations; this manipulation led to a decrease in the estimated number of elements serviced with high priority. We conclude that a mechanism incorporating temporally decaying priority tags is implicated in servicing multiple abrupt onsets in visual search.

Geometrically deformed models

We propose a new approach to the problem of generating a simple topologically-closed geometric model from a point-sampled volume data set. We call such a model a Geometrically Deformed Model or GDM. A GDM is created by placing a 'seed' model in the volume data set. The model is then deformed by a relaxation process that minimizes a set of constraints that provides a measure of how well the model fits the features in the data. Constraints are associated with each vertex in the model that control local deformation, interaction between the model and the data set, and the shape and topology of the model. Once generated, a GDM can be used for visualization, shape recognition, geometric measurements, or subjected to a series of geometric operations. This technique is of special importance because of the advent of nondestructive sensing equipment (CT, MRI) that generates point samples of true three-dimensional objects.

Deformable curve and surface finite-elements for free-form shape design

The finite element method is applied to generate primitives that build continuous deformable shapes designed to support a new free-form modeling paradigm. The primitives autonomously deform to minimize an energy functional subject to user controlled geometric constraints and loads. The approach requires less user input than conventional free-form modeling approaches because the shape can be parameterized independently of the number of degrees of freedom needed to describe the shape.Both a curve and a surface finite element are developed. The properties of these geometric primitives have been engineered to support an interactive three phase approach for defining very fair free-form shapes as found in automobiles, ship hulls and car bodies. The shape's character lines or folds and edges are defined with deformable curve segments. These character lines are then "skinned" with a deformable surface. The final shape is sculpted interactively by applying loads to the surface to control the surface shape between character lines. Shapes created with this technique enjoy the advantage that they are already meshed for further finite element analysis.

Priming contour-deleted images: evidence for intermediate representations in visual object recognition

The speed and accuracy of perceptual recognition of a briefly presented picture of an object is facilitated by its prior presentation. Picture priming tasks were used to assess whether the facilitation is a function of the repetition of: (a) the object's image features (viz., vertices and edges), (b) the object model (e.g., that it is a grand piano), or (c) a representation intermediate between (a) and (b) consisting of convex or singly concave components of the object, roughly corresponding to the object's parts. Subjects viewed pictures with half their contour removed by deleting either (a) every other image feature from each part, or (b) half the components. On a second (primed) block of trials, subjects saw: (a) the identical image that they viewed on the first block, (b) the complement which had the missing contours, or (c) a same name-different exemplar of the object class (e.g., a grand piano when an upright piano had been shown on the first block). With deletion of features, speed and accuracy of naming identical and complementary images were equivalent, indicating that none of the priming could be attributed to the features actually present in the image. Performance with both types of image enjoyed an advantage over that with the different exemplars, establishing that the priming was visual rather than verbal or conceptual. With deletion of the components, performance with identical images was much better than that with their complements. The latter were equivalent to the different exemplars, indicating that all the visual priming of an image of an object is through the activation of a representation of its components in specified relations. In terms of a recent neural net implementation of object recognition (Hummel & Biederman, in press), the results suggest that the locus of object priming may be at changes in the weight matrix for a geon assembly layer, where units have self-organized to represent combinations of convex or singly concave components (or geons) and their attributes (e.g., aspect ratio, orientation, and relations with other geons such as TOP-OF). The results of these experiments provide evidence for the psychological reality of intermediate representations in real-time visual object recognition.

Statistical analysis of collagen alignment in ligaments by scale-space analysis

Injuries to ligaments in the knee are common in sports and other physical activities. Some clinical methods are available for qualitatively evaluating the degree of ligament injury and healing. It is, however, desirable to objectively assess the healing of ligaments and to predicate optimal treatment on quantitative measurements of their structure. Information such as areas of coverage and spatial orientations of collagen fibrils, for example, may provide important information about the internal structure of ligament tissues. Since normal ligament tissues are made up of collagen fibrils which are highly organized, they can be considered as oriented piecewise linear patterns. In this paper, we propose a computational technique for statistical analysis of collagen alignment in ligament images using the scale-space approach. In this method, a ligament image is preprocessed by a sequence of filters which are second derivatives of two-dimensional Gaussian functions with different scales. This gives a set of zero-crossing maps (the scale space) from which a stability map is generated. Significant linear patterns are captured by analyzing the stability map. The directional information in terms of orientation distributions of the collagen fibrils in the image and the area covered by the fibrils in specific directions are extracted for statistical analysis. Examples illustrating the performance of this method with scanning electron microscope images of the collagen fibrils in healing rabbit medial collateral ligaments are presented in this paper.

Scale space approach to directional analysis of images

In this paper a new technique for directional analysis of linear patterns in images is proposed based on the notion of scale space. A given image is preprocessed by a sequence of filters which are second derivatives of 2-D Gaussian functions with different scales. This gives a set of zero crossing maps (the scale space) from which a stability map is generated. Significant linear patterns are detected from measurements on the stability map. Information regarding orientation of the linear patterns in the image and the area covered by the patterns in specific directions is then computed. The performance of the method is illustrated through applications to synthetic patters and to scanning electron microscope images of collagen fibrils in rabbit ligaments.