ABSTRACT
Throughout VBR literature, different technical terms are used to refer to the same things; see Figure A.1. Multiple video streams represent the input data to all video-based rendering algorithms. These video streams are recorded simultaneously by several acquisition/recording cameras placed around the dynamic scene. A single image recorded by one camera is referred to as a light-field image, source image, input image, or reference image. Typically, position, orientation, and internal imaging characteristics of the acquisition cameras have been determined beforehand, so the recorded footage consists of calibrated image data. The acquisition cameras are synchronized, so a time frame corresponds to all images recorded at the same time instant. The rendered new image is the output image, desired view, or target view. A viewing ray is the ray from the viewpoint’s center of projection through one pixel of the output image into the 3D scene. Finally, object space or scene space refers to the 3D space of the recorded scene, as opposed to the 2D image space.
