This thesis describes two new techniques for enhancing the rendering quality of cartoon characters in toon-shading applications. The proposed methods can be used to improve the output quality of current cel shaders. The first technique which uses 2D image-based algorithms, enhances the silhouettes of the input geometry and reduces the computer generated artefacts. The silhouettes are found by using the Sobel filter and reconstructed by Bezier curve fitting. The intensity of the reconstructed silhouettes is then modified to create a stylised appearance. In the second technique, a new hair model based on billboarded particles is introduced. This method is found to be particularly useful for generating toon-like specular highlights for hair, which are important in cartoon animations. The whole rendering framework is implemented in C++ using the OpenGL API. OpenGL extensions and GPU programming are used to take the advantage of the functionalities of currently available graphics hardware. The programming of graphics hardware is done using Cg, a high level shader language.

Background. In order to increase the visual fidelity of characters in video games, strand-based hair rendering within Unreal Engine 4 is explored in this thesis. Because of the complex nature of hair, explicit hair models can be very costly to render and simulate. It is theorized that utilizing customized hair and simulation settings, tailored to specific types of hairstyles could alleviate this issue while preserving visual fidelity. Objectives. The aim of this thesis is to provide an insight into what can be done to increase visual appeal and computer performance of physical simulation for different types of hairstyles and determine if customized settings may cause a significant impact on the perceived level of realism. The objectives of this thesis are to acquire a set of different strand-based hairstyles, determine a set of customized hair and simulation settings which can be applied to them, create a test scene inside of Unreal Engine 4 and render out a set of images and videos to be used in a user experiment, measure the performance of each customized setting and finally synthesize the acquired data. Methods. In order to achieve the aims and objectives of this thesis, a user experiment that utilizes the 2AFC method to let participants compare image- and video-pairs is performed as well as a performance experiment using the built-in profiling tools in Unreal Engine 4. In addition, a pilot experiment was performed in order to ascertain that the experiments would be feasible on the available hardware. Results. The results show that there was a significant difference in the perceived level of realism when different simulation settings was applied to the hairstyles, with customized settings being preferred to the default setting. The voting results on the image-pairs showed a preference for fine hair strands while the strand count did not have as much of an impact. It was shown that participants could easier distinguish between the different simulation- and hair-settings in long hair compared to short hair. The performance experiment showed that the amount of hair strands had the biggest impact on computer performance. Conclusions. Customizing hair and simulation settings to different types of hairstyles could provide a heightened perceived level of realism and a limited performance boost, from what could be derived from these experiments. Lowering the hair strand count was determined to be the most effective method of increasing performance. The performance of strand-based hair is currently not quite reaching a consistent 60 frames per second on the tested hardware, but with further optimizations it is believed that this could be acquired, especially on more powerful graphics cards. Future work should keep focus on increasing the stability of real-time strand-based hair simulation.