|Abstrakt:||Comparison of real lighting with the lighting computed by current global illumination methods brings up many questions concerning the correctness of these methods. While optimizing the methods and making the output images look nicer, physical and mathematical plausibility is often sacrificed.
Correctness of a method which uses several dozens of parameters is disputable, especially when the user has to set the parameters for every scene until the image looks real to him. As a result, there are more versions of "reality" and it is difficult to judge which one is the most real.
Images which cannot be distinguished from real images (photographs) are called photo-realistic. Global illumination methods are often compared by the quality of the output images and how much "photo-realistic" they are. The judge is a human operator. But can we really compare algorithms by such a subjective measure?
In this thesis we try to avoid these and other problems by taking a physically-based approach to global illumination. We give a mathematical definition and propose algorithmic solutions to the problem. Then we discuss the state-of-the art global illumination algorithms. Finally, we present design, implementation and improvements of an algorithm which solves the global illumination problem in newtonian sense.