Perceptually ModulatedLevel of Detail in Real Time Graphics
Patrik Johansson · KTH Publication Database DiVA (KTH Royal Institute of Technology) · 2013
As demands on real time 3D graphics increase we need new ways of improving rendering speeds. Significant speed gains can be achieved if we can remove polygons without affecting the visual fidelity of the image by removing the ones that will not be visible to the user anyhow. Level of detail (LOD) is one such method, where each 3D model is represented at several different detail levels, each containing a different number of polygons, so that the rendering engine can select the most appropriate one for the current view of the object. Traditionally the criterion used to decide what detail level should be used for the object has been the distance to the object from the viewpoint. The reasoning is that the further away the object is, the smaller it is likely to appear, and minor details are unlikely to be visible, and we can leave the details out without it being noticed. In this thesis we explore the use of other criteria for selecting detail level. We will discuss less common criteria that have been use in earlier works and how those criteria can be anchored in modern cognitive sciences. The result of the thesis is a LOD map, indicating what objects in a 3D scene should be represented at what detail level. The LOD map consists of two major parts; the acuity map, and the saliency map. The acuity map is calculated taking into considerations the shortcomings of the human visual system, while the saliency map is calculated using theories from the field of visual attention. When our demo application was tested, the results show that an average rendering speedup of 83% might be possible to gain with only moderate loss of visual fidelity.