A partitioning strategy that decomposes a given 2D or 3D model, in to 'approximately convex' pieces, is discussed. The strategy is based on the premise that for some models and applications, some of the non-convex features can be considered less significant, and allowed to remain in the final decomposition while removing others. The model is decomposed if its concavity exceeds the threshold . The result shows that if an application can sacrifice a little convexity, then algorithm can produce fewer components then the exact convex decomposition, in significantly less time.
