High Quality Image Reconstruction of Point Models

The parallel and discrete nature of many image-based techniques, such as interpolation methods, are highly suitable for GPU implementations. An important advantage is their bounded complexity by the screen resolution as opposed to the data size. Consequently, the direct surface reconstruction of points using image-based methods is an attractive solution to interactively render large data-sets. In this paper, we propose a novel approach that raises the robustness and quality standards of previous related works. The algorithm achieves quality equivalent to surface splatting techniques, while maintaining high performance rates compatible with previous image-based methods. A hierarchical pixel structure is employed to render the projected samples efficiently with minimum artifacts. One of the main advantages of the method is the independency from the number of samples during the surface reconstruction. Furthermore, no extra object space data structure is needed, making the approach also memory efficient.