van der Linden, Jarno2009-04-082009-04-081999-08Computer Science Technical Reports 168 (1999)1173-3500http://hdl.handle.net/2292/3502Image-based modeling can simplify the scene modeling process in computer graphics by assuming that it is not the geometry of a scene that is of interest, but rather the look of a scene. Instead of obtaining detailed surface geometry and colour data, visual data alone is acquired. This can be done with relative ease through photography. Light Fields are a recent development in how such information is stored. Images of a scene are used to create a four dimensional radiance function. A light field determines what a scene looks like when viewed from anywhere outside of the space enclosing the scene. With light fields, as in many other image-based modeling and rendering systems, there is a trade-off between quality and memory. A high quality result requires storage of a large number of samples obtained from high-resolution input images. The number of images required can be reduced by using image warping techniques based on image flow. Image warping computes intermediate images from the existing images in the system, thereby giving the illusion of a higher sampling density than that actually used during acquisition. In this paper, we show how to use image flow in light fields to produce high quality output from only a small set of input images. Unlike previous work, we do not generate intermediate images, but rather use image flow to render a new view directly. High quality new views can be generated from as few as 8x8 images. We demonstrate several rendering methods for image-flow augmented light fieldshttps://researchspace.auckland.ac.nz/docs/uoa-docs/rights.htmTechnical ReportFields of Research::280000 Information, Computing and Communication SciencesThe author(s)