Realism of Procedural Lichens on Horizontal Stone Surfaces : Simulation and evaluation of lichens

Abstract: Humans use creativity and manual dexterity to create things, usually in a labor-intensive process. Procedural content generation (PCG) is an approach where computers do most of the manual work of creating virtual objects, while a human leads the creative process. One problem with PCG is that created objects tend to look too perfect and artificial, leading to research into simulating the aging process of created virtual objects. Simulating the growth of living organisms such as lichens on existing virtual objects has been a focus of research. Most previous techniques lack a clear evaluation of the realism of the results. In this thesis, a lichen growth simulation is implemented and evaluated in terms of perceived realism. Specifically, the target of the simulation is lichen growth on horizontal stone surfaces represented using a set of textures. The lichen growth simulation is slow but robust and flexible, capable of generating a wide range of output textures. 75 participants were recruited among university students and video game industry professionals, asked to compare the realism of lichens generated with different parameter settings. The investigated parameters are the scale of the lichens, the compactness of the lichens, and the number of lichen clusters (density). Participants perceived surfaces with a dense or medium-dense cover of small scale, semi-compact lichen clusters as the most realistic, while surfaces with a sparse cover of large scale, very compact or very lacy lichen clusters were perceived as the least realistic. The lichens perceived as the most realistic are also among the most computationally expensive to produce. This outcome tells us that there is a trade-off between realism and simulation time in the lichen growth simulation. The simulation is able to generate a wide range of lichen appearances, though the speed of the simulation can be greatly improved in multiple ways.