Acta Polytechnica Hungarica Vol. 18, No. 8, 2021 – 47 – Transforming Hierarchical Data Structures – a PSVDAG–SVDAG Conversion Algorithm Branislav Madoš, Norbert Ádám Department of Computers and Informatics, Faculty of Electrical Engineering and Informatics, Technical University of Košice, Letná 9/A, 042 00 Koši ce, Slovakia e-mail: branislav.mados@tuke.sk, norbert.adam@tuke.sk Abstract: This paper examines the issues of domain-specific hierarchical data structures, based on directed acyclic graphs, dedicated to the representation of the geometry of three- dimensional scenes. In this paper, the authors introduce two versions (out-of-core and semi-out-of-core) of an algorithm to transform hierarchical data structures – pointerless sparse voxel directed acyclic graphs, into sparse voxel directed acyclic graphs. Pointerless sparse voxel directed acyclic graphs are not suitable for immediate traversing, due to the absence of pointers to the child nodes; however, they are suitable for archiving and streaming, as they have a more compact binary-level representation. Sparse voxel directed acyclic graphs, on the other hand, allow quick traversing during visualization or other forms of processing, since their nodes include pointers to child nodes. The disadvantage of this, is that the binary-level representation, requires more operating memory or secondary storage space. Both hierarchical data structures – sparse voxel directed acyclic graphs and pointerless sparse voxel directed acyclic graphs – and both versions of the proposed conversion algorithm are described in the first part of the paper. Results of tests, performed on various models – previously surface polygonal models, stored in the Wavefront Technologies geometry definition file format (OBJ) – now voxelized to the respective resolutions, are summarized in the second part of the paper. The binary-level representation lengths of both data structures, along with the time consumption of both versions of the proposed conversion algorithms, are detailed in the last part of the paper. Keywords: pointerless sparse voxel octrees; PSVO; sparse voxel octrees; SVO; pointerless sparse voxel directed acyclic graphs; PSVDAG; sparse voxel directed acyclic graphs; SVDAG; hierarchical data structures; volume dataset; three-dimensional image; lossless data compression 1 Introduction Hierarchical data structures (HDS), based on the use of octant trees (octrees) and Directed Acyclic Graphs (DAGs), are popular solutions to represent the geometry of three-dimensional scenes, especially if the scenes are voxelized from three- dimensional polygonal surface models [1]. In voxelized scenes, the uncompressed