UittrekselThis dissertation addresses the efficient compression of digital images from both rate- distortion and functionality perspectives. In particular, it focuses on compression of medical and holographic imaging modalities.
For the first image modaltity type, the main focus goes to the volumetric medical images, on which all modern hospitals heavily rely today – i.e. Computed Tomography (CT), Magnetic Resonance Imaging (MRI) and Ultra Sound (US) images. In fact, the completely digitized medical work flow, the improved imaging scanner technologies and the importance of volumetric image data sets have led to an exponentially increasing amount of data, raising the requirements for more efficient compression techniques with support for progressive quality and resolution scalability.
The second type of image modality represents the now upcoming holographic images, which are essentially interference pattern recordings. However, these record- ings require high-resolution acquisition – typically much higher than what is needed for regular photography – to accurately capture the interference patterns at an acceptable visual quality level. Moreover, such images exhibit rather atypical fre- quency distribution characteristics that often lead to inferior compression efficiency when resolving to classical image compression techniques. Thus, this too raises the requirement for an efficient representation technology that is able to cope with the increased resolutions and the specific data characteristics.
Finally, this dissertation links the proposed state-of-the-art technologies with actual standardization efforts in order to drive adoption and ensure interoperability. In this respect, the first contribution of this thesis is reflected by the creation of an International Standard, being ISO/IEC 15444-10 | Rec. ITU-T T.809, also referred to as JPEG 2000 Part 10 or JP3D. This specification extends JPEG 2000
Part 1 to add support for the isotropic compression of volumetric images.
As second contribution, this dissertation presents a first time exhaustive study on a variety of coding methodologies, beyond JP3D, that can be applied to efficiently compress medical volumetric images. In fact, it investigates the effectiveness of a number of existing state-of-the-art 2D image compression techniques by extending them to 3D and subsequently combining them in a single framework. More specifically, the framework combines JP3D with (1) two additional new wavelet filters, (2) directional wavelet support, and (3) a block-based intra-band prediction mode. Moreover, the conducted experiments also determine the optimal settings to be used with regular JP3D for the compression of the tested volumetric medical image modalities. These experiments are all presented in reference to JPEG-LS and H.265/HEVC, in order to provide a good point of comparison with current state-of- the-art.
The third contribution shows that the presented framework can also be used to efficiently compress holographic image data. By doing so, it demonstrates that JPEG 2000, combined with two potential extensions, can actually be used to efficiently compress holographic images. The first proposed extension enhances the existing Arbitrary Decomposition Styles (ADS) feature of JPEG 2000 Part 2 such that any decomposition style becomes possible, while at the same time lowering the implementation and signaling complexities. The second proposed extension introduces practical directional wavelet support to JPEG 2000.
Finally, the fourth contribution of this dissertation discusses (ongoing) work that was done in the context of standardization within JPEG of image coding technology. Namely, it discussed the contributions made to JPEG XR, JPEG Systems, AIC and JPEG XT, which were often logical precursors to or consequences of the other work that was performed for this thesis. Most notably, it discusses (1) a subjective quality assessment procedure that was originally developed for the evaluation of JPEG XR, (2) the upcoming publication of the International Standard, ISO/IEC 18477-9 – JPEG XT Part 9, Alpha channel coding, and (3) the JPEG Systems activity that focusses on the consolidation of common major technology that is used by many different JPEG standards.
|Datum Prijs||14 sep 2015|