A new approach of point-based rendering

A new approach of point-based rendering is presented for objects with complex shape. By this approach, an object is represented by a set of discrete sample points, each point stands for a sampled surface area whose boundary is assumed to be a circle when viewed along its normal. An algorithm is proposed to build a point-based model with level of details. Efficiency of the rendering process is ensured by performing adaptive sampling of the model without calculating the sample points on the fly. To get rid of the aliasing effects due to point sampling, we have developed a delta-z-buffer and a flag buffer to smooth the boundary of two adjacent sampled areas and to fill the holes that may appear inside the image of an object. Experimental results show that our new approach is capable of rendering complex objects consisting of only 50,000 points with acceptable image quality, and high rendering efficiency