Catálogo de publicaciones - libros

This paper presents a similarity-based adaptive neighborhood (SBAN) dense stereovision algorithm which uses color for comparing pixels. In SBAN methods, the neighbor pixels which are not similar to the central one are excluded of the window when computing the correlation index, which corresponds to adapting the equivalent size and shape of the correlation neighborhood. We present a specific type of SBAN algorithms, in which the similarity criterion is based on a pre-classification of pixels, and show that they can be efficiently implemented using recursive computations. As an example, we show that color, more precisely hue, is an efficient similarity criterion for SBAN methods and present the result on a classical stereo pair.

An algorithm of affine image transformations search on the base of morphological operations has been proposed. Images are represented as sets of binary cuts, that gives possibilities to find affinity via solution of overdetermined linear equations systems under formation by means of obtained corresponding affineequivalent points. Presence of several image cuts provides a reliability increasing due to result averaging.

This paper proposes a new algorithm for estimating lens distortion for a pantilt- zoom camera. No calibration pattern is needed: detection of a single feature point at different camera configurations allows the generation of a virtual calibration pattern which is composed of points that, in an undistorted image, are ensured to lie on straight lines. An iterative algorithm estimates lens distortion by minimizing a non-straightness error metric solving two linear subproblems. Experiments on synthetic and real images show the validity of the proposed method.

We present a discrete approach for solving the Shape From Shading problem. This approach allows us to recover directly the depth of the surface and is based on the propagation of equal height regions. The reconstruction can use more than one light source direction and can resolve the concave/convex ambiguity when only one source image is available. Moreover the robustness to noise is demonstrated on both synthetic and real images.

This paper describes a new method of generating real time acoustics with the use of widely available graphic video cards. Algorithm that was mapped to video hardware is ray tracing. Computed echogram was later used during real time auralization. Simplified acoustic model allows to use this method in real time simulations like video games or fast acoustics approximation. Advantage of this method is ability to move source and listener during simulation without the need for long precomputation phase. Test scenes consisted of highly occluded building architectures. Results show that performing acoustic computation on GPU can significantly speed it up. Propagation of 16384 rays (at 10 reflections) for few room flat took about 32ms on AMD 2GHz with ATI Radeon 9800. Such results allow for real time simulations. Computation of above one million rays took around 1s. Described algorithm is optimized for working with ATI Radeon 9800 hardware with the use of DirectX 9 API.

The original shadow mapping algorithm grants that construction of the light's depth buffer and the screen rendering are sequential jobs that cannot be interlaced.

For creating a CG model from an existing 3D object, a laser scanner is commonly used for providing a geometry model and a digital camera for preparing photographs for texture mapping. It is important to have an accurate estimate of camera parameters in order to precisely align photographs on overlapping regions. The alignment based on silhouettes has been most frequently employed. In this paper we examine the combination of silhouettes and texture patterns for aligning textures. Initially an optimal strategy for camera position allocation is described. Then a new method of pattern matching is presented for more precise camera position estimation. Two toy objects were used to investigate the validity of our new method. It was found that our method could yield an alignment accuracy within one to two pixels(that corresponds to .5 to .6mm).

Antialiasing is still a challenge in real-time computer graphics. In this paper we present a real-time selective antialiasing solution that builds upon our experience. We investigated existing approaches to real-time antialiasing and finally found a new simpler solution. Our new idea is to use the z-buffer directly for extracting visible edges information. Method presented here can be used to improve image quality in graphics accelerators but also applications such as real-time ray tracing.

A priority driven based algorithm for surface reconstruction from a set of surface points is presented. It calculates the shape change at the boundary of mesh area and builds a priority queue for the advance front of mesh area according to the changes. Then, the mesh growing process is forced to propagate in reliable direction through the queue. The algorithm reconstructs surface in a fast and reliable way. The efficiency of the proposed algorithm is demonstrated by the experimental results.

Registration is an important step to combine pictures that are taken either from different perspectives (multi-viewpoint), at different points in time (multi-temporal) or even from diverse sensors (multi-modal). The result is a single image that contains the combined information of all original images. We use 2 1 2 -D range data taken by a laser range scanner. During recording occlusions can leave some of the objects scanned incompletely. Thus it is the task of the registration to produce images which give the most complete view of the scene possible. An algorithm often used for the registration of range images is the ICP algorithm (Rusinkiewicz and Levoy, 2001). This point-based algorithm only uses local information, and it has the advantage that it is universally applicable. The drawbacks are that its very time consuming, it may converge in a local minimum, and the result cannot be evaluated absolutely. Therefore, for our domestic scenario we present an approach that first detects certain global features of the room and then uses their semantic and spatial information to register the range images. Thus this approach has the advantage that its tendency to get stuck in a local minimum is reduced and the overall performance is increased.