Mesh Simplification Using Hybrid Saliency

摘要

This research presents a view-independent saliency model, which takes the effect of both color information and geometric information into account. The proposed method can simplify a mesh by dropping low-attention places while protecting the high-attention places. Multi-platform game contents are becoming a trend in the gaming field of CG in recent years. One of the problems in developing such game contents is the difference of the processing performance between the platforms. A content that runs smoothly on high processing-performance platform may not run smoothly on the lower-performance platforms. This problem can be solved by mesh simplification, which reduces the number of polygons in the mesh while preserving the overall shape, volume and boundaries as much as possible. Fewer polygons require less processing performance, and the simplified mesh can be processed smoothly even in low processing-performance platform. The proposed method simplifies a mesh in three steps. First, it takes a triangular polygonal mesh for simplification as input. Then, it calculates a view-independent hybrid saliency for each vertex of the input mesh. The hybrid saliency is evaluated from the curvature, the mesh saliency, the maximum, minimum, and average values of multiview image saliencies. Last, it simplifies the input mesh by dropping low hybrid saliency vertices. The experiments indicated that the proposed method yields a better result than existing methods.