Shape-From-Silhouette Across Time Part I: Theory and Algorithms

Shape-From-Silhouette (SFS) is a shape reconstruction method which constructs a 3D shape estimate of an object using silhouette images of the object. The output of a SFS algorithm is known as the Visual Hull (VH). Traditionally SFS is either performed on static objects, or separately at each time instant in the case of videos of moving objects. In this paper we develop a theory of performing SFS across time: estimating the shape of a dynamic object (with unknown motion) by combining all of the silhouette images of the object over time. We first introduce a one dimensional element called a Bounding Edge to represent the Visual Hull. We then show that aligning two Visual Hulls using just their silhouettes is in general ambiguous and derive the geometric constraints (in terms of Bounding Edges) that govern the alignment. To break the alignment ambiguity, we combine stereo information with silhouette information and derive a Temporal SFS algorithm which consists of two steps: (1) estimate the motion of the objects over time (Visual Hull Alignment) and (2) combine the silhouette information using the estimated motion (Visual Hull Refinement). The algorithm is first developed for rigid objects and then extended to articulated objects. In the Part II of this paper we apply our temporal SFS algorithm to two human-related applications: (1) the acquisition of detailed human kinematic models and (2) marker-less motion tracking.     

Computational design of iris folding patterns

Computational Visual Media - Iris folding is an art-form consisting of layered strips of paper, forming a spiral pattern behind an aperture, which can be used to make cards and gift tags. This...     

A wide view auto‐stereoscopic 3D display with an eye‐tracking system for enhanced horizontal viewing position and viewing distance

This paper describes the development of auto‐stereoscopic three‐dimensional (3D) display with an eye‐tracking system for not only the X‐axis (right–left) and Y‐axis (up–down) plane directions but also the Z‐axis (forward–backward) direction. In the past, the eye‐tracking 3D system for the XY‐axes plane directions that we had developed had a narrow 3D viewing space in the Z‐axis direction because of occurrence of 3D crosstalk variation on screen. The 3D crosstalk variation on screen was occurred when the viewer's eye position moved back and forth along the Z‐axis direction. The reason was that the liquid crystal (LC) barrier pitch was fixed and the LC barrier was able to control the only barrier aperture position. To solve this problem, we developed the LC barrier that is able to control the barrier pitch as well as the barrier aperture position in real time, corresponding to the viewer's eye position. As a result, the 3D viewing space has achieved to expand up to 320–1016 mm from the display surface in the Z‐axis direction and within a range of ±267 mm in the X‐axis direction. In terms of the Y‐axis direction, the viewing space is not necessary to be considered, because of a stripe‐shaped parallax barrier.     

3D interactive topology optimization on hand-held devices

This educational paper describes the implementation aspects, user interface design considerations and workflow potential of the recently published TopOpt 3D App. The app solves the standard minimum compliance problem in 3D and allows the user to change design settings interactively at any point in time during the optimization. Apart from its educational nature, the app may point towards future ways of performing industrial design. Instead of the usual geometrize, then model and optimize approach, the geometry now automatically adapts to the varying boundary and loading conditions. The app is freely available for iOS at Apple’s App Store and at http://www.topopt.dtu.dk/TopOpt3D for Windows and OSX.     

Effect of surface modification on NO2 sensing properties of SnO2 varistor-type sensors

NO₂ sensing properties of SnO₂-based varistor-type sensors have been investigated in the temperature range of 400-650°C and in the NO₂ concentration range of 15–30 ppm. Pure SnO₂ exhibited a weak nonlinear I–V characteristic in air, but clear nonlinearity in NO₂ at 450°C. The breakdown voltage of SnO₂ shifted to a high electric field upon exposure to NO₂ and the magnitude of the shift was well correlated with NO₂ concentration. Thus, SnO₂ exhibited some sensitivity to NO₂ as a varistor-type sensor. When SnO₂ particles coated with a SiO₂ thin film were used as a raw material for fabricating a varistor, the breakdown voltage in air was approximately the double that of pure SnO₂ and the sensitivity to 15 ppm NO₂ was enhanced slightly. However, the sensitivity to 30 ppm NO₂ decreased. The Cr₂O₃-loading on SnO₂ also led to an increase in the breakdown voltage in air, but the Cr₂O₃ addition was not effective for promoting the NO₂ sensitivity under the present experimental conditions.     

Degeneracy of the Linear Seventeen-Point Algorithm for Generalized Essential Matrix

In estimating motions of multi-centered optical systems using the generalized camera model, one can use the linear seventeen-point algorithm for obtaining a generalized essential matrix, the counterpart of the eight-point algorithm for the essential matrix of a pair of cameras. Like the eight-point algorithm, the seventeen-point algorithm has degenerate cases. However, mechanisms of the degeneracy of this algorithm have not been investigated. We propose a method to find degenerate cases of the algorithm by decomposing a measurement matrix that is used in the algorithm into two matrices about ray directions and centers of projections. This decomposition method allows us not only to prove degeneracy of the previously known degenerate cases, but also to find a new degenerate configuration.     

A two-step surface-based 3D deep learning pipeline for segmentation of intracranial aneurysms

Computational Visual Media - The exact shape of intracranial aneurysms is critical in medical diagnosis and surgical planning. While voxel-based deep learning frameworks have been proposed for this...     

High efficiency mixing by the use of cross-linked micro capillary fluid filter

A novel type passive mixing device that causes the three-dimensional flow was proposed. This mixer consists of the integrated capillary bundle structure. And the capillaries ware cross-linked each other. So it is called cross-linked micro capillary filter. The mixing effect of the cross-linked micro capillary filter was calculated by the computational fluid dynamics (CFD). The fluid behaviour in the fine three-dimensional structure could be analysed by the use of CFD. In the result of CFD calculation, the cross-linked micro capillary filter estimated high mixing effect. Moreover, the mixing efficiency was become higher by change the cross-linked form. The calculation result was decided the form of the cross-linked micro capillary filter. And this filter was fabricated by application of the Deep X-ray lithography. To form the cross-linked capillary, it was operated twice exposures. In these exposures, respectively the exposure stage was tilted to difference angle. The cross-linked capillary filter fabricated in this way was applied to vertical fluid flow operation. This cross-linked capillary filter could hold and transmit the fluid by the switchover of in impressed pressure. Herewith the cross-linked micro capillary filter showed availability as high efficiency mixing device.     

Shape and motion from image streams under orthography: a factorization method

Inferring scene geometry and camera motion from a stream of images is possible in principle, but is an ill-conditioned problem when the objects are distant with respect to their size. We have developed a factorization method that can overcome this difficulty by recovering shape and motion under orthography without computing depth as an intermediate step.An image stream can be represented by the 2F×P measurement matrix of the image coordinates of P points tracked through F frames. We show that under orthographic projection this matrix is of rank 3.Based on this observation, the factorization method uses the singular-value decomposition technique to factor the measurement matrix into two matrices which represent object shape and camera rotation respectively. Two of the three translation components are computed in a preprocessing stage. The method can also handle and obtain a full solution from a partially filled-in measurement matrix that may result from occlusions or tracking failures.The method gives accurate results, and does not introduce smoothing in either shape or motion. We demonstrate this with a series of experiments on laboratory and outdoor image streams, with and without occlusions.