A stereo camera comprises of a pair of cameras whose viewing directions converge, and it can be used to derive 3D structural information about the scene. If both cameras are fully-calibrated (intrinsic and extrinsic parameters), the 3D position of a point p visible in both cameras can be calculated via triangulation. The position p is estimated by computing the intersection point of two rays, rA and rB. The ray rA originates in the center of projection of camera A, cA, and passes the image plane in the position pA. The same construction is valid for ray rB from camera B. However, due to measurement noise, the rays will not intersect exactly at a single point. In this case, we can compute a pseudo-intersection point that minimizes the sum of squared distance to each pointing ray.
The epipolar geometry describes the image formation process in a stereo pair of cameras. It describes the fact that an image point pA in camera view A has a corresponding point pB in the camera view B, which lies on a line eB in image B, the so-called epipolar line. The epipolar geometry of a stereo pair is fully-specified by its fundamental matrix. The fundamental matrix can be inferred from eight point correspondences between two uncalibrated cameras, and it is directly available for fully-calibrated camera pairs. By using the fundamental matrix and the epipolar line, image correspondences can be computed using simple matrix multiplications, which reduces the problem to an one-dimensional search space along a line. The appearance, and the physical and kinematic properties of a real-world human body are the result of the interplay of many complex physiological components.
For example, the appearance of the skin is the result of structural pigmentation, light interaction on the body surface, and the deformation of muscles and connective tissues. The kinematic properties of the human body are mainly determined by its skeleton structure, i.e. bones and interconnecting joints. The kinematics also influences the physical shape of the person. Therefore, an authentic computational human model has to realistically represent the shape, kinematics and appearance of the real human. The surface geometry of the human body is typically modeled by means of a triangle mesh. A mesh is a collection of vertices, edges and faces that defines the shape of an object in computer graphics. The faces usually consist of triangles, which are connected by their common edges. In our projects, we acquired the geometric details of the human body by using a full body laser scanner. Our computational model of the shape is obtained by transforming the raw scans, i.e. triangulated depth maps, into a highquality surface mesh employing a Poisson reconstructionmethod. By using such scanning device, we are able to capture not only the coarse shape of the actor, but also fine details in the body shape and in the apparel. Moreover, such acquisition technology enables us to easily model different subjects.
Let's abandon the concept of a kinematic skeleton to represent the motion and deformations of the virtual actor. By doing this, the novel algorithms rely mostly on the high-quality model of the actor's shape to simultaneously capture rigid and non-rigidly deforming surfaces from multiple synchronized video streams.
Our website is not responsible for the information contained by this article. Articleinput.com is a free articles resource thus practically any visitor can submit an article. However if you notice any copyrighted material, please contact us and we will remove the article(s) in discussion right away.
Note: This article was sent to us by: Jamie W. at 12142009
1. Visual computing and its research areas
All articles are property of their respective authors. Please read our Privacy Policy!
© 2009 ArticleInput.com.