< Terug naar vorige pagina

Publicatie

L-infinite-driven coding of semi-regular meshes

Boek - Dissertatie

Providing users with an immersive viewing experience can prove useful in entertainment, medical care, military simulations, or education. This has been a focus of hardware manufacturers in the past decade, which resulted in a number of novel visualization technologies. Among them, we remind the growing number of commercially-available Head-Mounted Devices, 3D screens and, more recently, curved screen TVs. Such devices provide immersion by creating a 3D effect of the visualized scene. On the software side, achieving an immersive viewing experience has prompted extensive research in the creation of realistic 3D environments, and accurate 3D representations. One of the most useful tools in this sense are 3D polygonal meshes. A wide range of important domains rely on polygonal meshes, the most popular being the gaming industry. E-learning and tele-presence, or advanced 3D conferencing systems demand highly accurate and realistic 3D representations of humans. These applications have the power to drastically change the way in which people communicate by providing a genuine immersive 3D experience. In 3D movie productions the subjects are offline-scanned with very high accuracy. Their 3D representations need to be stored for further use or transmitted at a remote site, where they undergo subsequent processing steps (such as animation, immersion in a virtual 3D scene, scene composition and rendering using multiple subjects, etc.). 3D CAD is widely employed in the automobile and aircraft industry. 3D representations are also employed for the virtual reconstruction of urban or topographic landscapes. In 3D watermarking one can provide efficient protection against illegal use, distribution or tampering of 3D representations. In the particular case of mesh geometry watermarking, imperceptible watermarks can be embedded in the mesh geometry to protect the 3D models against copyright violation. 3D medical imaging can compensate for various inadequacies of classical 2D imaging. Finally, 3D printed prosthetics are gaining popularity given their ability to be accurately fitted to each patient. In all these application scenarios, 3D polygonal meshes are a common 3D representation model. Through them, the accuracy and the level of detail in the virtual representation become functions of the mesh resolution. Thus, increasing the resolution allows for complying with the available processing power or for adding as much detail as the user requires. The processing power of graphics hardware has witnessed a tremendous increase in the past decades, which in turn triggered a boost in the employed mesh resolutions, to the point where virtual objects consisting of millions of vertices are commonplace. For this reason, efficient compression algorithms are mandatory. In addition, all these applications have another common point: the quantization errors need to be very efficiently controlled at the level of each vertex. Many 3D mesh coding technologies have been proposed in the past, however very few offer local control of the quantization error. This dissertation proposes novel approaches for L-infinite-driven semi-regular mesh coding, in which the individual quantization errors are accurately upper bounded. In this context, novel techniques for estimating the L-infinite distortion in 3D semi-regular mesh coding are proposed. Moreover, most existing L-infinite-driven compression systems make high-rate assumptions with respect to quantization and, as a consequence, employ uniform partitions. This dissertation investigates optimal L-infinite-constrained scalar quantization without making classical high-rate assumptions. As a result, a new class of
Jaar van publicatie:2016
Trefwoorden:3D