5.4.2 Analysis of Shear-warp Factorization Algorithm

6 Summary The rendering process of discrete samples was described as a resampling process divided in two phases: reconstruction and traversal. Then, rendering acceleration techniques were classified into three different groups: using pre-computed information, controlling image quality, and simplifying viewing transformation. Finally, volume rendering algorithms were presented, divided into four different groups: polygon-based, texture-based, splatting, and shear-warp factorization. The discussion on reconstruction methods showed that interpolation-based methods are faster and more accurate than splatting-based methods. With respect to traversal, object-order methods are in general faster than image-order methods. Among the rendering acceleration techniques, it can be observed that all the algorithms presented use pre-computed information by exploiting coherence in object or image spaces, as shown in section 5.5. Most algorithms also use graphics hardware to accelerate rendering. The analysis of the algorithms showed that texture-based and shear-warp factorization algorithms are indicated to render seismic data sets. Texture-based algorithms have the additional advantage of being able to use the graphics hardware, which makes them more efficient. Both algorithms still have " weaknesses " , however. Texture-based algorithms require a great amount of memory, which is still expensive. The shear-warp factorization algorithm does not exploit the current graphics hardware at all. It was also mentioned that besides interactive frame rates, it is also important for interpreters to see seismic volumes and well data in the same visualization, which implies extending the rendering algorithm to handle geometry as well as volume data.The intermediate pixels between voxels are calculated using interpolation. After the composition step, the shared image is warped to generate the final image. 5.4.1 Algorithm steps 1. In a pre-procession step, create three RLE structures for the volumetric data, each of them aligned with one of the main axes. 2. Transform the volumetric data to sheared object space by translating and resampling each slice. The slicing direction is the most perpendicular to the direction of view. 3. Run through the volume and image simultaneously in scanline order. Render the voxels that are visible and non-transparent. 4. Composite. 5. Warp the intermediate image to image space. The shear-warp algorithm accelerates rendering by exploiting coherence in both object and image spaces. The synchronized access to the data structures that encode separately coherence in object and image spaces simplifies the rendering task and improves cache locality. One disadvantage is that the algorithm does not use the graphics hardware, because it does not use polygons as …