Toward More Effective Viewpoint Computation Tools

Abstract

The proper setting of cameras is an essential component in many 3D computer graphics applications. Commonly, viewpoint computation tools rely on the specification of visual criteria on a number of targets, each expressed as a constraint; then on the use of an optimization-based technique to compute a 7-degrees of freedom camera setting that best satisfy this set of constraints. Proposed methods can be evaluated in terms of their efficiency (required computation time), but there is a clear lack of a proper evaluation of their effectiveness (how aesthetically satisfactory the generated viewpoints are). In parallel, current methods rely on the maximization of a single fitness function built as a weighted sum (i.e. a pure tradeoff) over the satisfaction of each single criterion considered independently from all others. In contrast, cinematographers' sense of the effective satisfaction of a viewpoint is far from a tradeoff between visual criteria. These issues call for the provision of means to better evaluate the overall satisfaction of a composition problem, and methods to improve the search of a satisfactory viewpoint. In this paper, we present a work in progress which targets to steer computation tools in this direction. We first propose a range of aggregation functions which supplement the classical tradeoff function, and enable to express evolved relationships between criteria. We then propose to aggregate the individual satisfactions of criteria in hierarchical way instead of simply summing them. We finally propose to reduce the search to camera positions (i.e. from 7D to 3D), while constraining the framing more strongly by separately optimizing its orientation and focal length.