Theory and application of integrated, ubiquitous computer graphics.

点击翻译按钮获取中文摘要

The use of computer-assisted creation of visual digital media has seen huge growth. The improvement for example in the quality, speed and workflow of digital photography and video has allowed much of humanity to express itself visually. Digital photography and video content creation has become ubiquitous. On the other hand, virtually all other applications of digital visual communication that employ information visualisation, computer animation and graphics, and augmented and virtual reality still employ the traditional "broadcast" model: a few experts create content that is consumed by many. A "many creator/many consumer" model is still a far-off goal. The gulf between this goal and the science and engineering required to make computer graphics ubiquitous is also growing. Biomedicine, cultural and archaeological analysis and preservation, visual simulation, design, art, games, engineering, entertainment and manufacture are all in need of integrated, robust graphical solutions. Visual communication using realistic, interactive 3-D computer graphics remains the province of expert users. While the effectiveness of 3-D tools continue to improve, they address a growing yet relatively smaller set of users. The very success of realistic computer graphics both scientifically and industrially has been self-limiting: each successfully modelled visual phenomenon provides considerable generality, but is isolated in its conception from other models both in its semantics and its workflow. This allows a manner of scientific reductionism in furthering each field, and it facilitates a practical division of labour for experts, but it is wholly unsatisfactory for the purposes of creating integrated, multiply-coupled systems for autonomous environments as created by novice users. While there are some deep workflow and usability impediments, the fundamental problems in all areas of computer graphics, including complex geometric modelling, animation, deformation, viewing, lighting and rendering, is that complexity is extremely difficult to manage, and that they are separately considered in the creation of visual depictions. As graphics systems incorporate an increasing amount of physical simulation, it will become all the more difficult to predict outcomes, manage complexity, and insert user control. In the proposed research I plan to take my past experience in the mathematical and physical modelling of scale-sensitive visual phenomena to develop new methods to allow computational visual phenomena to interact with one another, and that would permit novices to create digital visual content so that they are better able to develop convincing visual narratives. New ways of looking at the problem are required, since the reduction of complexity of visual models cannot be accomplished just by improving interfaces and workflows. The areas we shall look at involve difficult problems in computer graphics: character and human body modelling, facial animation, visual modelling of natural phenomena, and control of light transport simulations. In each case, a reformulation of the problem will be required, because the usual "levels-of-detail" approach neglects the need for cross-scale and cross-model interactions. Since the work is intended to enable ubiquitous visual communication, I will make extensive use empirical techniques to guide modelling. I will employ emerging research in applied perception, as well as observing users using eye-tracking equipment to see, for example where artists look when they create content, and how audiences scan faces and motion. I will use the results of such observations to form a binding between the "seeing is believing" criterion employed by real people, and the object measures we require algorithmically.