Drawbacks

The main drawback with radiosity is the amount of time required for the calculation. This is due to the solving of the equation system, but also to the computation of the form factor.

For animated scenes where only the camera is moving, the time required is acceptable. But in the case of cinematographic scenes where objects move, the radiosity method is not a good solution.

Other drawbacks are linked to the hypothesis made to reduce the radiosity equation : the radiosity and form factor are constant on a patch. Here is a classical example :

Final Image	Solid View

A light source (a sphere) is placed in the bowl and the light 'slobbers' under the ball. If you look at the solid view, the problem can be easily explained. Some patche (uniform color squares) are both inside and outside the bowl. As the radiosity and form factor are supposed to be constant on a patch, the algorithms use a sample point (not necessarily the center of the patch). On the problematic patches, the sample point is taken inside the bowl and the color is applied outside too. Moreover the shading process applied for the final rendering amplify the defects.

As we have already seen, the diffuse light hypothesis is a limitation for the rendering methods. In order to get good results, rendering engines combine radiosity with other rendering method like raytracing which allow specular lights and other important effects. Here is an example provided by André Pascual, and rendered with Moonlight Atelier.

Here, differences between the images are less visible than those in the Comparison chapter, but the radiosity increase the physical correctness of the final image.