利用最小周長函數之三維網格骨架擷取

Abstract

我們利用最小周長函數(Minimum Slice Perimeter function)，提出了一個三維網格骨架擷取演算法。最小周長函數是一個用來量測物體局部體積資訊的純量函數，藉由最小周長函數，我們可以找到每一個頂點的所相對的潛在骨架節點的所在位置，我們利用這些骨架節點的位置，將網格中的每一個點移動到它自己相對應的骨架節點的位置，可以將輸入的三維網格轉換至一個外觀上接近於骨架的網格，也就是所謂的骨架網格(skeleton mesh)，接著我們利用多層次精細度(LOD)簡化架構，來化簡這個骨架網格。我們利用多層次精細度簡化的目標是為了要保留骨架網格的整體外觀，並不同於傳統的多層次精細度化簡，它是專注於物體表面細節的維持。我們提出的演算法不僅僅簡單，並能夠有效的產生合理且不具多餘分支的骨架。

We propose a novel algorithm to extract curve-skeletons from 3D meshes using the minimum slice perimeter (MSP) function. The MSP function is a scalar surface function to measure the local volume information. Through the MSP function, we can find a potential skeleton position to each vertex. We can transform the input mesh into a so called skeleton mesh whose shape is close to a skeleton by moving each vertex to its corresponding potential skeleton positions. After we obtain the skeleton mesh, we apply the LOD simplification to reduce the skeleton mesh. Our LOD simplification aims to preserve the global shape of the skeleton mesh, rather than local shape and features in traditional LOD simplification. The proposed algorithm is simple yet effective in generating reasonably good curve skeletons that have no branches due to surface noise.