Type-S Project - 3D Modeling Progress Report

9/12/2004 Mesh optimization for new texture procedure and smoother lighting calculations

The Old 3D model

The Problems

Sharp edges were produces by separating the mesh into objects. Although this allowed for fewer vertices and faster processing time, the model produced terrible aliasing effects which hindered the visual quality. Also each component of the car was separated for a particular part. The doors, headlights, windows, mirrors, etc. were modeled individually and textured individually. This was done with interactivity in mind in the final version of the game, however it increases the polygon count by a significant amount and causes inconsistencies when the model is textured with an environment map to produce reflections.

In the shaded version of the old model, lighting calculation abrasions can easily be seen. This is produced as a result of the placement of some vertices. Although structurally, the model is correct in the wire frame, there are not enough vertices to produce an effective lighting calculation. This makes the model appear "cheap" looking and ineffective when it comes to the final version. The separation of the individual parts is also problematic since it leaves holes in the model and in certain situations requires a 24 bit depth buffer instead of 16 bits to successfully sort the parts in depth.

The New Optimized Model

The Solutions

In the new model, most of the body work is kept in a single pieced object. This not only removes wholes and reduces aliasing effects, it also also allows for textures to create the complex parts such as the windows. The the new TGA file type allows for transparencies to be achieved in the texture, so a separate 3d model is no longer necessary. About 1000 vertices were also saved in creating the mesh for the front grill. In the original version, the grill was modeled structurally, however in the new version, the grill will be masked out by the texture, requiring only a single polygon for its map.

When shaded you can see that the edges appear creamier and smoother than the old version. This was achieved by increasing the resolution at the points of the edges to give the rendering code enough vertices to produce an accurate lighting calculation. As video cards progress, the model will never lose visual integrity, since this calculation is based on the structure and not on the texture of the model. This can best be seen at the front bumper modeling and headlights of the new version when compared to the old one.

Conclusion

The new 3d modeling tactic is to produce 3d model meshes that put emphasis on the "shape" of the model as opposed to the final appearance of the model when rendered. The texturing will allow for creating complex shapes through masking and defining alpha values, rather than creating them with polygons.

All content (c) Bedoworks Visual Studios 2004