Video Rendering

Written by Paul Bourke
Example models by Thomas Gill
August 1993

Computer rendering usually consists of creating the geometry of the model with a 3D modelling package (eg: MicroStation, ArchiCAD, AutoCAD) then rendering this model with a package (eg: ModelView, StrataVision, Radiance, RenderMan). The rendering process requires extra information not normally provided by the modelling package such as the lighting environment as well as the surface colour, reflective properties, and texture. In many cases a physical scale model might be available, this model may accurately represent the geometry but it may be difficult to efficiently show different lighting and/or texture alternatives manually.

The following is an example of a computer based rendering technique which makes use of an existing physical model and powerful image manipulation software to illustrate lighting, colour, and texture. This technique avoids the necessity of building a computer based version of the geometry as well as the often high learning curve required to effectively create realistic computer renderings.

The outline of the process is as follows:

* Arrange the physical model and create the desired lighting effects. In the examples shown here it was only necessary to use a single external light source, this was simulated with a theatre light. The general layout of equipment captured using a video camera is shown in figure 5. The video screen shows the image from the camera, the user can capture any video frame.

* Capture a number of views of the model with a video capture system. Boards that create computer based animations or still images are readily available, they take a video source (video camera or video recorder) and form a colour bitmap representation that can be manipulated with a computer. The image quality is generally not as high as from scanning devices, this is entirely due to the limitation of video signals. The image in figure 2 for example is quite fuzzy partly due to the inability to focus across a large depth range. This fuzziness can be reduced using image processing techniques.

* Acquire any textures that will be applied to the surfaces of the model. This can be done in a number of ways, creating the textures with in a computer painting program, capture the textures using a video capture system as above, or more commonly, scan the textures using a colour scanner.

* Using an image editing package change the colour and texture of the objects in the captured images. For this process to be successful very powerful image manipulation software is necessary, many of the tools required are only found in the best packages. The most common operation is to select a region of the image and change the colour and/or texture of that region without changing the lighting characteristics (intensity). Of course, when changing the texture of a region the texture must also undergo perspective transformations corresponding to the depth changes of that region.

Figure 1

Figure 2

Figure 1 and 2 show the images from two different models captured with a video camera. The original images were colour but for this exercise the views of the model were converted into grey scale, the colour will be added afterward.

Figure 3

Figure 4

Figure 3 and 4 are two processed images from the same models shown in figure 1 and 2, although not the identical view or lighting configuration. In both examples textures have been laid on the surfaces, in figure 2 a human figure has also been added. This figure was scanned from a photograph and inserted into the scene with an perspective transformation. In both figures the lighting in preserved on the applied textures, this can be seen in the lower left corner of figure 2 and on the highlight of the curved bench in figure 1.

Figure 5