Representation of Dense Volume Datasets Using Pointerless Sparse Voxel Octrees With Variable and Fixed-Length Encoding

Abstract

The paper deals with the problematics of the dense volume datasets representation, in which voxels are formed as multi-bit scalar values, and is oriented to the use of hierarchical data structures, not only as the medium for encoding and storing of volume datasets, but also for their lossless compression. The main contribution of the paper is in the design of the hierarchical data structure, that allows carving out of subtrees, that are homogenously filled not only by the symbol 0 or symbol 1, but any binary representation of voxel value from defined set of values. Designed data structure provides the possibility of fixed-length and also variable-length (Huffman) encoding of voxel values in leaf node level of the data structure and also in the non-leaf nodes. Results of tests, performed on medical volume datasets that were obtained by various non-invasive imaging techniques, including Computed Tomography and Magnetic Resonance Imaging, along with conclusions that were made, based on the test results, take place in the second part of the paper.