Fusion of computed point clouds and integral-imaging concepts for full-parallax 3d display

Resumo

During the last century, various technologies of 3D image capturing and visualization have spotlighted, due to both their pioneering nature and the aspiration to extend the applications of conventional 2D imaging technology to 3D scenes. Besides, thanks to advances in opto-electronic imaging technologies, the possibilities of capturing and transmitting 2D images in real-time have progressed significantly, and boosted the growth of 3D image capturing, processing, transmission and as well as display techniques. Among the latter, integral-imaging technology has been considered as one of the promising ones to restore real 3D scenes through the use of a multi-view visualization system that provides to observers with a sense of immersive depth. Many research groups and companies have researched this novel technique with different approaches, and occasions for various complements. In this work, we followed this trend, but processed through our novel strategies and algorithms. Thus, we may say that our approach is innovative, when compared to conventional proposals. The main objective of our research is to develop techniques that allow recording and simulating the natural scene in 3D by using several cameras which have different types and characteristics. Then, we compose a dense 3D scene from the computed 3D data by using various methods and techniques. Finally, we provide a volumetric scene which is restored with great similarity to the original shape, through a comprehensive 3D monitor and/or display system. Our Proposed integral-imaging monitor shows an immersive experience to multiple observers. In this thesis we address the challenges of integral image production techniques based on the computerized 3D information, and we focus in particular on the implementation of full-parallax 3D display system. We have also made progress in overcoming the limitations of the conventional integral-imaging technique. In addition, we have developed different refinement methodologies and restoration strategies for the composed depth information. Finally, we have applied an adequate solution that reduces the computation times significantly, associated with the repetitive calculation phase in the generation of an integral image. All these results are presented by the corresponding images and proposed display experiments.