Today many systems exist to generate geometric models of existing scenes and objects. However, no accurate data about surface appearance such as colors and textures is stored in this process. Such data can be captured as a series of images that, collectively, capture all surfaces of the object.
This work introduces a method to compute a minimal set of camera positions for this purpose. Taking images from the computed positions can then be used to derive a complete set of surface appearance data. A slightly different application of the presented method is the computation of a minimal set of viewpoints for reference images to be used in image-based rendering methods.
First a method to determine an optimal set of viewpoint regions for a given scene is introduced. It uses a hierarchical visibility method to preprocess the scene. Then a technique to find an optimal set of viewpoint regions is presented and the solution is used to derive an optimal set of viewpoints. Results and visualizations of the computed solutions are presented.