Abstract: This paper presents a novel splatting algorithm based on directional reconstruction filter kernels to enhance the 3-D object's edges in volume rendering. By utilizing the gradient information in volume data set, our algorithm can adaptively select directional ellipse Gaussian footprint functions. Compared with the traditional splatting, the new splatting algorithm can decrease the overlap of footprint functions, and render more dedicate profiles of the 3-D object at the same speed. In this paper, we present the new splatting algorithm in case parallel and perspective projections are given, and the error analysis in perspective projection. The experimental results prove our algorithm’s effectiveness.

Key words: volume rendering, splatting, reconstruction filter kernel, perspective projection