Real-time stylization of image streams

OData support
Dr. Szécsi László
Department of Control Engineering and Information Technology

The main goal of computer graphics, since the very beginning of its history, was to create as realistic looking images as possible, that is, to simulate the workings of a traditional photographic camera. Artistic techniques however tend to abstract detail to communicate emotions, experiences and ideas to the observer. Non-photorealistic or stylized rendering is inspired by artistic styles found in drawings, paintings, cartoons and technical illustrations.

There is a growing interest in newer, stylized rendering styles, especially in games. However, modifying an application's graphics and engine to support new styles, even if it is open source, is a hard and time-consuming task. Therefore a general method to alter any application's rendering style non-invasively would be beneficial.

In this thesis I attempt to create a framework that intercepts the stream of OpenGL calls made by an application, reconstructs a meaningful representation of the scene geometry from it and also provides access to the frame buffer of the target. Different non-photorealistic renderers will then be explored that use this data to alter the application's graphics. The viability of the black-box approach, performance of the framework and test results will also be discussed.


Please sign in to download the files of this thesis.