Silhouette detection in volumetric models based on a non-photorealistic illumination system

A basic step for many illustration techniques is the extraction of silhouettes and shape lines, in order to enhance the object shape visible to the observer. This paper presents a new and very fast method to obtain silhouettes and shape lines in non-photorealistic volume rendering processes, based on ray-tracing using an octree for optimization. This method consists of extracting contours based on image buffers acquired during the calculation of intersections in ray-tracing process with the model. These image buffers are then combined to produce silhouettes and shape lines from the model. A substantial improvement with regard to other silhouette extraction methods in volumes is that this method does not require the user to define thresholds; furthermore, it is not based on the z-buffer and thus does not depend on the precision of the latter. Neither the camera position nor the direction affect the performance of the algorithm. Finally, the algorithm detects silhouettes for objects which are inside other objects that are distinguished only by their hue.