Abstract Previous works have shown that for characters with obvious articulation, animations comprising the blending of skeletal bone transformations, significantly reduce the computation and data requirements, permitting much greater visual detail. However increasingly, animators and engine designers must produce animations that can run on platforms with quite different runtime capabilities. They may simply author multiple skeletons and sets of animations, or reduce the temporal or spatial detail to achieve acceptable frame rates. In this paper we show how these skeletal skinning methods can be extended to support a continuous level of animation detail. We show that by manipulating the skeletal hierarchy and skinning parameters we can throttle the computational load of a character model in real‐time according to its position in a scene and any hardware constraints. Our method is compatible with additional geometric level of detail (LoD) methods. Copyright © 2007 John Wiley & Sons, Ltd.

The majority of camera calibration methods, including the Gold Standard algorithm, use point-based information and simultaneously estimate all calibration parameters. In contrast, we propose a novel calibration method that exploits line orientation information and decouples the problem into two simpler stages. We formulate the problem as minimization of the lateral displacement between single projected image lines and their vanishing points. Unlike previous vanishing point methods, parallel line pairs are not required. Additionally, the invariance properties of vanishing points mean that multiple images related by pure translation can be used to increase the calibration data set size without increasing the number of estimated parameters. We compare this method with vanishing point methods and the Gold Standard algorithm and demonstrate that it has comparable performance.

We consider the problem of decoupling translation and rotation for a collec-tion of 3D data related to 2D images by a projection. The main contribution is to show that equations describing image formation can be decoupled to form a component independent of translation. The decomposition is based on the invariance properties of the projection of points at infinity; if image formation is expressed in terms of points at infinity, general motions can be reduced to pure rotations. Contrary to other methods based on vanishing points, our approach does not require parallel directions to be present in the scene. We use the invariance property to simplify camera calibration equa-tions. We consider three cases for a known calibration object: full calibration, pose estimation and internal calibration by pure translation. Experiments on synthetic and real data show that the decomposition can obtain similar re-sults to a full parameter search. For pure translation, the decomposition can be effectively used to obtain more accurate parameters, without increasing the dimensionality of the problem. 1

This paper centres on the problem of estimating the location of features in the environment within a map building framework for a mobile robot with a single camera. Most of the existing approaches to feature location from vision data have been developed by a combination of matching and geometric triangulation. Triangulation is used to obtain equations of reconstruction or structure-from-motion. In contrast, our approach avoids the matching problem by looking for evidence of the location of features on an occupancy grid. Similar to map building via sonar methods, we gather evidence of the trace of potential position of features. Thus, maxima in the occupancy grid define robust feature positions. Preliminary results show that our approach can obtain accurate results in real time. The approach can handle noise in the feature detection and in the robot position. The integration of vision data over time reduces the covariance of the estimates.

Generalized cylinders provide a compact representation for modeling many components of natural objects as well as a great variety of human-made industrial parts. This paper presents a new approach to modeling generalized cylinders based on cross-sectional curves defined using Fourier descriptors. This modeling is based on contour interpolation and is implemented using a subdivision technique. The definition of generalized cylinders uses a three-dimensional trajectory which provides an adequate control for the smoothness of bend with a small number of parameters and includes the orientation of each cross-section (i.e, the local coordinate system) in the interpolation framework. Fourier representations of cross-sectional curves are obtained from contours in digital images, and corresponding points are identified by considering angular and arc-length parametrizations. Changes in cross-section shape through the trajectory are performed using Fourier morphing. The technique proposed provides a comprehensive definition that allows the modeling of a wide variety of shapes, while maintaining a compact characterization to facilitate the description of shapes and displays.

The generalised Hough transform extracts arbitrary objects by using a non-analytic model shape representation obtained from gradient direction information. The main drawback of this technique is the excessive computational burden because of the four-dimensional parameter space required when orientation and scale are unknown. We present a novel representation of a model shape defined by the geometric relationship given by the position of a collection of edge points. This representation avoids errors due to unreliable gradient direction information and is used to reduce the computational requirements by decomposing the four-dimensional parameter space into two two-dimensional sub-spaces. Experimental results show the efficacy of the new technique for extracting shapes from synthetic and real images.

In this paper we present a novel formulation for the extraction of arbitrary shapes in model-based recognition. The formulation is based on the mapping defined in the Hough transform. We develop this mapping for the analytic representation of a shape characterised by a Fourier parameterisation. Edge direction information is included in the formulation as a way of reducing the computational requirements in the extraction process. The proposed approach extends the analytic formulation of the Hough transform to arbitrary shapes. An analytic representation provides a compact and extensive coverage of a shape which leads to an accurate and efficient evidence accumulation process. Experimental results show that the new approach can handle noise and occlusion in synthetic and real images.

This paper presents how morphological transformations can be related to representations of a set on different lattices. A hierarchical definition of structuring element conveys to a class of multigrid transformations /spl Psi//sub k/ that handle changes on discrete representations of regions. The transformations correspond to upward and downward processes in a hierarchical structure. Based on multigrid transformations, a method to delineate not-perfectly-isolated objects in an nxn image requiring O(log n) time is presented. The approach considers grey level regions as sets and processes through a pyramid to carry out geometric manipulations. Extending the concept of boundary to cope with hierarchical representations of a set, a second method which identifies the boundaries in an image is discussed.>