Showing papers by "Luc Van Gool published in 1997"

Method and system for acquiring a three-dimensional shape description

Abstract: Method for acquiring a three-dimensional shape or image of a scene, wherein a predetermined pattern of lines is projected onto the scene and the shape is acquired on the basis of relative distances between the lines and/or intersections of the lines of the pattern.

Reading between the lines—a method for extracting dynamic 3D with texture

Abstract: A method is presented that extracts the 3D shape of objects, together with their surface texture. Both shape and texture are obtained from a single image. The paper sketches the complete system but focuses on the problem of texture extraction. The underlying principle is based on an active technique. A high resolution pattern is projected onto the object and the deformations as observed by a single camera yield the 3rd dimension. The surface texture is extracted from the same image by literally reading between the lines that are used for the shape extraction. This is done using a combination of interpolation and non-linear diffusion techniques. Because the whole procedure is based on a single image, a frame-byframe reconstruction of a video taken with the pattern projected throughout, yields 3D shape dynamics. 1 A paradigm shift in 3D Most methods for three-dimensional shape extraction go via the explicit calculation of the distance between the sensor and the object. The 3D shape of the surface then follows from the variation of this distance. This is what 3D acquisition devices for reverse engineering, shape inspection, and mobile robotics wotild typically do. More recent applications, however, focus on visualization, such as virtual and augmented reality, 3D on the Internet, special effects in movies, etc. These kind of applications come with different priorities: l Extracting the absolute scale of objects usually is not crucial. Objects. will be shown at completely different scales anyway. Permission to make digitalhud copies ofall or pat ofthis material for personal or classroom use is granted without fee provided that the copies are not made or distributed for profit or commercial advantage, the copyright notice, the title of the publication and its date appear, and notice is given that copyright is by pemhsion ofthe ACM, Inc. To copy oh&se, to republish, to post on servers or to redistribute to lists, requires specific permission and/or fee / ACM KWT ‘97 Lausanne Switzerland Copyright 1997 ACM 0-89791-953497/9..$3.50 l The extraction of surface texture becomes a prerequisite. Traditional 3D acquisition systems often lack this ability to align the shape and texture data. l In the top range of visualisation applications such as the movies and virtual worlds 3D dynamics becomes more important. Rather than building a still model and animating it via off-line motion tracking, direct extraction of 3D motion would strongly alleviate such tasks. l Also, with Internet at the fingertips of users in small companies and amateurs at home, 3D acquisition devices should become much cheaper, less bulky and easier to use in order for 3D models to fully pen& trate the Net. The paper proposes a novel, 3D acquisition system that is geared towards these new requirements, i.e. the extraction of 3D models for visualisation. The hardware needed is a simple slide projector, a normal camera, and a computer. Setting up tlie system is easy and requires no exotic calibration objects. It uses special illumination to obtain good geometric precision. From a single image, both 3D shape and surface texture are extracted. This turns the system into a say 4D scanner as it becomes possible to extract dynamic 3D by frame by frame reconstruction of video data. A complete description of the system is outside the scope of this short paper. This paper will mainly focus on the aspect of surface texture extraction (section 3). Nevertheless, section 2 will sketch how the 3D shape is extracted. Results of 3D shapes together with the texture and 3D motions are shown in section 4. That section also shows a preliminary example of how such data can be used for special effects or animation. Section 5 concludes the paper. 2 One-shot 3D acquisition 2.1 Comparison with other methods As passive 3D acquisition systems often lack precision, certainly in untextured areas, we opted for an active ap preach, where a simple square pattern is projected on the

Shape equivalence under perspective and projective transformations.

Abstract: When a planar shape is viewed obliquely, it is deformed by a perspective deformation. If the visual system were to pick up geometrical invariants from such projections, these would necessarily be invariant under the wider class of projective transformations. To what extent can the visual system tell the difference between perspective and nonperspective but still projective deformations of shapes? To investigate this, observers were asked to indicate which of two test patterns most resembled a standard pattern. The test patterns were related to the standard pattern by a perspective or projective transformation, or they were completely unrelated. Performance was slightly better in a matching task with perspective and unrelated test patterns (92.6%) than in a projective-random matching task (88.8%). In a direct comparison, participants had a small preference (58.5%) for the perspectively related patterns over the projectively related ones. Preferences were based on the values of the transformation parameters (slant and shear). Hence, perspective and projective transformations yielded perceptual differences, but they were not treated in a categorically different manner by the human visual system.

The Cascaded Hough Transform as Support for Grouping and Finding Vanishing Points and Lines

Abstract: In the companion paper [7] a grouping strategy with a firm geometrical underpinning and without the problem of combinatorias is proposed. It is based on the exploitation of structures that remain fixed under the transformations that relate corresponding contour segments in regular patterns. In this paper we present a solution for the complementary task of extracting these fixed structures in an efficient and non-combinatorial way, based on the iterated application of the Hough transform. Apart from grouping, this ‘Cascaded Hough Transform’ or CHT for short can also be used for the detection of straight lines, vanishing points and vanishing lines.

Automatic 3D model building from video sequences

Abstract: We describe progress on automating the construction of 3D geometric models from video sequences. The system we are developing allows a person to walk around a room with a video camera, and from the video sequence alone extract a 3D model of the room. The room can then be graphically rendered together with possible arrangements of previously modelled objects, e.g. CAD models of furniture. The system is designed to ease model acquisition for virtual reality and telepresence applications, however the approach also has application in model based image coding. The approach is based on recent work in computer vision where image primitives are extracted and matched through an image sequence. For example, the simultaneous estimation of epipolar geometry and image corner correspondences. The matching techniques are both robust (detecting and discarding mismatches) and fully automatic. No knowledge of the camera (e.g. its focal length) or motion is required. 3D structure is computed from the matched image primitives (corners, lines and texture). The structure is then triangulated and texture is mapped onto the triangles from intensities in the images. Experimental results are provided for a variety of scenes, including isolated objects acquired with a standard CCD camera, and outdoor scenes acquired with a hand-held camcorder.

One-shot 3D-shape and Texture Acquisition of Facial Data

Abstract: In this paper we present new methods to simultaneously extract and exploit the three-dimensional shape of a face and its surface texture It is based on an active technique, ie special illumination, but in contrast to traditional active sensing does not require scanning or sequential projection of multiple patterns This one-shot nature of the devise allows to capture moving objects, eg for making a 3D reconstruction of a face even when the person is talking The use of the system is illustrated using simple methods to extract both textural and geometrical features from faces, that can be used for authentication purposes The advantage of using 3D data is that both types of features can be made more invariant under changes in head pose or illumination conditions

One-shot active 3D image capture

Abstract: A system is proposed that simultaneously captures the three- dimensional shape of an object and its surface texture. The 3D acquisition system is based on an active technique, but in contrast to traditional active sensing does not require scanning or sequential projection of multiple patterns. The system projects a simple pattern of squares on a scene and views it from a different angle. The underlying software automatically detects projected pattern in the image and determines the shape. At the same time, the algorithm allows us to extract the texture from the same image, thereby avoiding shape/texture alignment problems. Furthermore, its one-shot operation principle enables the system to retrieve the shape of moving objects, such as talking heads. Experiments show that the algorithm is robust and provides accurate three-dimensional reconstructions. The experiments have been carried out on various industrial or other objects, faces and other parts of the human body. The recovered shape also allows us to extract both textural and geometrical features, that can be used for identification or authentication (faces) purposes.© (1997) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

A Systematic Approach to Geometric-Based Grouping and Non-accidentalness

Abstract: Geometric regularities have often been used for grouping. Nonetheless, their foundations have typically been rather ad hoc - with “regular” or “non-accidental” features being listed according to intuition or based on application-specific considerations. This paper describes a more systematic line of thought towards such visual grouping. Based on an earlier observation that fixed structures in images are directly related to object regularities and grouping specific invariants, fixed structures are propounded as a theoretical glue. Moreover, grouping strategies with less than combinatorial complexity are difficult to develop. The propounded approach is also intended to keep grouping complexity under control. To that end, it combines the use of invariants with a Cascaded Hough Transform to efficiently extract candidate fixed structures.

Recognition of suspects through 3D photographs

Abstract: A system is proposed that simultaneously captures the three-dimensional shape of a face and its surface texture.Such a three-dimensional model allows to compare surveillance images of an offender irrespective of the pose ofthe offender's head. Also a single model for face albedo has been elaborated and its use will be demonstrated forviewing under different lighting conditions. The 3D acquisition system is based on an active technique, i.e. special illumination, but in contrast to traditional active sensing does not require scanning or sequential projection of multiple patterns. As a consequence, 3D photographs can be taken from a single image, and thus also when suspects do not collaborate. 1. INTRODUCTION One of the problems in forensics is to compare suspect photographs with the surveillance images of an offender. For one thing, the pose of the head will in general be different from that of the photos (usually only a front and a3/4 view or profile view from only one side) .

3D reconstruction of generic house roofs from aerial images of urban areas

Abstract: A method is presented to automatically generate 3D models of house roofs from aerial images of residential areas in urban sites. First, homogeneus regions with consistent photometric and chromatic properties corresponding to roof structures are delineated in the images by navigating through a constraint triangulation network. Stereo matching of straight line segments is performed between corresponding regions only. Line segments that are matched across at least three views are reconstructed by a bundle adjustment procedure. The reconstructed line segments are then grouped into coplanar configurations and polygonal patch hypotheses are formed. Subsequently, each polygon hypothesis is subjected to a consistency verification with respect to the 3D reconstruction and the original image data, and, if necessary, corrected accordingly. Observe that the combinatorics is kept under control by processing one region at the time. In a next stage, the polygons are glued together into a roof model. The emphasis here is on extracting the correct topology of the roof structure. Metric accuracy of the reconstruction is obtained in an additional step by backprojecting the recovered (wireframe) model of the roof structure onto the images and minimizing the total reprojection error. The viability of this approach has been tested on a state-of-the-art dataset of aerial images of residential areas in Brussels.