Abstract: When pictorial information is presented, details of importance are typically emphasized. These include discontinuities in the geometry, highly curved regions, silhouettes, etc.

This work analyzes the probability that certain smooth surface regions or polygonal edges possess silhouettes. This probability analysis is then associated with the visual importance of the local neighborhood, which is capable of capturing discontinuities and highly curved regions.

A non-photorealistic rendering technique is subsequently proposed to take advantage of the silhouette-based importance. Based on this importance analysis, we present a completely automatic algorithm that creates line-art that captures visual features in the model in an appealing way.