ISBN 978-3-8439-5150-0

978-3-8439-5150-0, Reihe Ingenieurwissenschaften

Andreas Weinlich
Compression of Medical Computed Tomography Images Using Optimization Methods

372 Seiten, Dissertation Universität Erlangen-Nürnberg (2022), Softcover, A5

Zusammenfassung / Abstract

As for natural images, compression of medical computed tomography images to reduce necessary storage space or transmission data rate can save both time and money in clinics. For lossless and high-quality compression, pixel-wise prediction techniques like backward-adaptive autoregression are improved and extended in various ways, e. g., by optimizing context regions, estimating probability distributions for entropy coding, or exploiting the third dimension of volume images. The developed Open Source framework Vanilc proves to outperform most established codecs like HM or VTM at high image qualities. For medium-quality, deformation compensation, constraining motion to heart muscle movement, is proposed to replace block-based video compression for dynamic cardiac data. Using further novel techniques, the Nvidia CUDA implementation surpasses rate-distortion performance of modern codecs like HM.