In this paper, we present a simple and practical method for determining the front-facing polygons for static, dynamic, and deformable objects. The object space is considered a set of open-frustum regions. A bit string associated with each region records polygons for their extended positive half space intersecting the corresponding region. For a given viewpoint, the bit string of the set of the front facing polygons can be determined in a constant time. While objects continuously keep deforming and/or remain in motion in each frame, we update set of bit strings, determined as minimal as possible by the spatial coherence, making our method still perform efficiently.

Visibility culling techniques have been studied extensively in computer graphics for interactive walkthrough applications in recent years. In this paper, a visibility culling approach by exploiting hardware-accelerated occlusion query is proposed. Organizing the regular grid representation of input scene as an octree-like hierarchy, a 2-tier view frustum culling algorithm is to efficiently cull away nodes invisible from a given viewpoint. Employing the eye-siding number of nodes, we can quickly enumerate an occlusion front-to-back order and effectively maximize the number of parallelizable occlusion queries for nodes while traversing the hierarchy. As experimental results show, our approach improves the overall performance in the test walkthrough.