Construction of clustered HLOD with As‐Simplified‐As‐Possible boundaries

Abstract This paper presents a new method for constructing a clustered hierarchy of Level‐of‐Detail (LOD) with As‐Simplified‐As‐Possible cluster boundaries to represent large scale triangular meshes at varying resolution, making them suitable for interactive applications. Ensuring consistency between nodes in a clustered hierarchy remains a challenging problem, cracks may appear when dealing with nodes of different resolutions. Bypassing the issue by locking boundary vertices leads to under‐simplification causing quality loss or non‐volume‐preserving simplifications. To address these limitations, we propose a novel approach: a hierarchy construction that enables nodes boundaries simplification at each level, maintaining consistency between neighboring clusters. Building upon existing methods, we represent the model as a clustered hierarchy of Progressive Meshes. Nodes are formed by a partition of a progressive internal mesh and a set of progressive boundary portions in bijection with the node neighbors, allowing for continuous refinement of the boundary while ensuring the LOD mesh to be crack‐free (neighboring clusters in the view‐dependent LOD have consistent boundaries). This new degree of freedom results in an LOD of better quality and/or a more simplified model, and also reduces popping effects that may occur when abruptly simplifying locked boundaries. Our method focuses on the construction of this adaptive representation while ensuring that it is compatible with the standard rendering pipeline, and we provide an example of implementation.