Task-based automatic camera placement

Abstract

Placing cameras to view an animation that takes place in a virtual 3D environment is a di cult task. Correctly placing an object in space and orienting it, and furthermore, animating it to follow the action in the scene is an activity that requires considerable expertise. Approaches to automating this activity to various degrees have been proposed in the literature. Some of these approaches have constricted assumptions about the nature of the animation and the scene they visualize, therefore they can be used only under limited conditions. While some approaches require a lot of attention from the user, others fail to give the user su cient means to a ect the camera placement. We propose a novel abstraction called Task for implementing camera placement functionality. Tasks strike a balance between ease of use and ability to control the output by enabling users to easily guide camera placement without dealing with low-level geometric constructs. Users can utilize tasks to control camera placement in terms of high-level, understandable notions like objects, their relations, and impressions on viewers while designing video presentations of 3D animations. Our framework of camera placement automation reconciles the demands brought by di erent tasks, and provides tasks with common low-level geometric foundations. The exibility and extensibility of the framework facilitates its use with diverse 3D scenes and visual variety in its output.