Postopek brezizgubnega stiskanja razčlenjenih vokselskih podatkov : doktorska disertacija

slovenian

doktorska dela

2011

Univerza v Mariboru, Fakulteta za elektrotehniko računalništvo in informatiko

The dissertation covers the area of voxel data compression. In the introduction we define the problem of compressing voxel data, describe the goals of the work and state the hypotheses we wish to confirm. The introduction is followed by the description and definition of voxel data, description of the devices for acquiring the data and description of Hounsfield's scale for analyzing voxel data. In continuation we cover the area of voxel graphics, methods for visualization of voxel data and an implementation of a ray-casting method for rendering voxel data. The implemented method utilizes the GPU, therefore we also illustrate the function of shaders. Next, an overview of methods for voxel data compression is given. We describe two methods that have been recently published and providea more detailed description of the QT-B method. In the core of the dissertation we describe the support techniques that we applied at the development of our method LoCoVox. We describe in details the JPEG-LS and JBIG standards together with LoCoVox and its web implementation VoxelServer, VoxelClient, and Voxel Decompressor. The next chapter presents experimental verification of LoCoVox. We applied the method to compress voxel datasets and compared the results to the general-purpose compression algorithms ZIP and RAR and the domain-specific method QT-B. The main advantage of our method is that it allows transferring designated areas of data. The method is significantly improved compared to QT-B and ZIP and is comparable to the RAR method in case we want to transfer the complete dataset. If it is sufficient to transfer only a part of data (for instance tissue, liquids, bones), the LoCoVox is considerably more efficient.

07.08.2023