Shadows are one of the most important elements to create realistic virtual environments. Shadows increase realism, play a large role in 3-dimensional visual perception and have an effect on expression of the scene.
But shadow generation process requires high complexity in time, especially for large-scale environments. Therefore it needs to be managed efficiently for real-time applications.
Shadows are natural phenomena caused by relations between objects and lights. The information of relationships can be saved as images, shadow maps. This method, shadow mapping, is very popular method because of many advantages.
Under static environments shadows are stationary because lights and all the objects in a scene are static. To generate static shadows, we make a shadow map once and calculate shadows with the shadow map. With moving lights or moving objects, under dynamic environments, we need to recalculate the shadow map for every movement because the relations are changed.
Like in a city environment, we usually meet semi-static environment that lights and most of objects are fixed and some objects are moving. But even with a few moving objects, we should recalculate shadow map for each movement in general shadow mapping process.
In this thesis, we propose a new real-time shadow generation method for large-scale and semi-static virtual environments. We first divide a scene into static and dynamic part, and generate shadow maps for each of them to calculate shadows. After that, we update shadow map for dynamic part only. And finally we verify the effectiveness of our method.