Zu dieser Karteikarte gibt es einen kompletten Satz an Karteikarten. Kostenlos!

Input: Set of images where on each image, there is an identifiable silhouette of the shown object. Each camera spans a generalized cone through the silhouette. The intersection of all these cones approximates the shape of the object.

Goal: For a new viewpoint, find out where the view ray intersects the approximated shape of the object.

In each source image, the projection of the view ray for one pixel in the target image is a line. We want to find the intersection points of this line with the silhouette. Thus, we discretize the silhouette into line segments and test for intersections: complexity per source image.

Note that if the camera position is fixed, the projected view rays in the source images are a pencil of lines through the camera epipolar point. Every time the camera position changes, perform a binning of the polygon edges into sectors. This reduces the intersection complexity from to .