Abstract

Unstructured Lumigraph Rendering is a well-established and flexible image-based rendering technique to create novel views from a set of acquired images. However, the standard implementation may not provide a pleasing visualization when images are captured under suboptimal conditions. In our scenario, lightfield information is gained using images from endoscopes that observe the operating field during minimal-invasive surgery. Suboptimal lighting conditions and limited mobility of the endoscope lead to visual artefacts in the final visualization of the scene. To resolve these problems, we propose a set of new algorithms for computing the ranking criterions that are used by the Unstructured Lumigraph for visualization. Using these algorithms, we get satisfying results for visualization of endoscopic image data captured under the aforementioned conditions. To achieve real-time visualization, the vital parts of our algorithms have been implemented using OpenGL and GLSL, thus making use of the features and speed of modern graphics hardware.