In the real world, light particles called photons are emitted from light sources, travel through space, and interact with the objects in the environment until absorption. Some of them eventually reach the camera sensors (or our eyes), which in turn record the light energy and convert it to an image. This process can be simulated on a computer to produce a realistic image of a virtual (i.e. non-existing) environment with physically correct lighting.

The physical laws that govern light transport and the mathematical models that formalize the image rendering problem are generally well understood. However, carrying out an accurate light transport simulation on a computer can be a challenging task in practice – billions of light particles need to be tracked, each particle can interact with arbitrarily many objects, and typically only a small fraction of all emitted particles eventually make their way to the camera.

The main focus of my work can be very obscurely shortly summarized as investigating methods for the efficient discovery of light transport paths that contribute energy to the camera. Such methods can tremendously increase rendering efficiency, as they aim to concentrate the computational effort where it pays off most. Higher efficiency allows for handling larger environments, simulating more complex lighting, and producing more visually stunning images.