Thread: The lighting scene in 3d Animation

The lighting in a scene is one of the most difficult things to get right, and deserves close study for anybody serious about model¬ling and animation. Basic.lighting consists of ambient light - a general light present, such as the light from a cloudy sky - and then specific lights placed in a scene. Again, there is some considerable variety in the control you have over these lights, but this is not quite enough for realistic effects. A scene using standard lighting techniques simply looks rendered. Those images that are the most realistic use advanced techniques such as ray tracing and radiosity.

Both of these methods employ the physics of light. In ray tracing, each ray of light is followed back from the viewpoint to its origin, in order to determine its colour and intensity. For example, looking at a painting on a wall, following back to a point on that painting, there will be a particular colour in the paint¬ing itself; the light falling on it may be composed of ambient light streaming in through the windows covered with yellow curtains, and this light will be reflecting off the pale blue walls of the room. The point here is that the colour and intensity of any ray of light depends on many factors, simply due to the fact that light is reflected from many surfaces, passes through some transparent materials, each of these subtly affecting the final colour. Ray tracing takes as many of these factors into consideration as possible. This method is also particular good when reflective surfaces such as mirrors are involved.

Radiosity also uses basic physics to combine together the effects of multiple light sources and surfaces, but it does so in a general sense, and not from the point of view of a camera. The results are added to the scene itself, rather like a general bitmap that is applied all over. Effects such as shad¬ows, light coming through yellow curtains, and so forth, are handled well by this method, but it is only really suitable for diffuse surfaces, and not reflective surfaces - if the camera moves, the reflection will change, and so you will need to use a reflec¬tion map in such instances.

The big problem with radiosity is that if an object in the scene moves, then the results of the radiosity calculations are broken. These can take quite some time to compute - it is an iterative process, and the quality improves the longer the time you give the system. So, this method is suitable for applications such as an architectural fly¬through of a building or other structure. The radiosity calculations are done just once, and then the camera is animated and moved through the scene. The calculations needed for radiosity are generally more compute intensive than ray tracing, and so when objects in the scene are animated, ray trac¬ing is the preferred method.



The description that we have given here is from the point of view of software such as 3ds MAX. But there are other ways of repre¬senting objects in a scene. Maya is particu¬larly good with what is called NURBS content. NURBS stands for non-uniform rational B-splines, and this is a method whereby curves and surfaces are represented mathematically rather than by collections of triangles; you can think of the comparison here between image-based maps and proce¬dural maps. This is a procedural way ofrepre¬senting surfaces.

Usually, the NURBS method uses much less data than triangles. If you're building a 3D sphere with triangles you will need a high number of faces, otherwise it will look too jaggy. But if you define the sphere as a single surface based on a mathematical formula it will be much simpler in terms of raw data. This method is particularly suitable for organic surfaces - modelling a human figure from triangles is a difficult procedure, and NURBS gives you a different level of control that is much more suitable for such objects. It is also useful for smooth and perhaps complex surfaces such as the aeroplane on which the logo has to go. The specialities of these different programs make them more or less attractive in different fields.

Software like 3ds MAX is better for archi¬tectural or low-level polygon scenes, and it is this that is important for anyone developing applications that render in real-time, such as video games, virtual worlds, ete. Two points here: the most successful CAD program on the PC is AutoCAD, also by Autodesk, and as MAX and its predecessor 3D Studio have long been compatible with AutoCAD file formats, MAX has naturally been popular with AutoCAD users, such as architects and engineers. Also, with real-time rendering, such as in a game, to enhance performance it is important to have as small a number of triangles in the scene as possible. Mind you, the latest graphics cards for gamers, with several hundred processor cores, are capable of rendering triangles at rates that would probably not have been believed, even just ten years ago.

MAX is also well set up for character rigging, layered animation, etc. In short, if you're doing gaming, visualisation, virtual worlds - anything real-time - then you're probably better off using something like MAX. Also, pre-rendered animations like architectural fly-throughs where radiosity can be used.

NURBS-based software such as Maya is probably better suited to high-detail material that will be pre-rendered. For example, if you're creating a static scene that has lots of complex surfaces, liquids, mist, fog, funky lighting, and does not need to be rendered or manipulated in real-time. Similarly, if you are creating high detail lush scenes with crashing waves, complex curvy shapes, weird lighting, particle effects, dinosaurs, and so on, then software like Maya is probably more appropriate. CGI (computer-generated imagery) movies are, of course, pre-rendered so you would expect Bollywood to use Maya more than the games industry, which would favour MAX.