Silhouette-based Level of Detail: A comparison of real-time performance and image space metrics
Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesis
Context. The geometric complexity of objects in games and other real-time applications is a crucial aspect concerning the performance of the application. Such applications usually redraw the screen between 30-60 times per second, sometimes even more often, which can be a hard task in an environment with a high number of geometrically complex objects. The concept called Level of Detail, often abbreviated LoD, aims to alleviate the load on the hardware by introducing methods and techniques to minimize the amount of geometry while still maintaining the same, or very similar result.
Objectives. This study will compare four of the often used techniques, namely Static LoD, Unpopping LoD, Curved PN Triangles, and Phong Tessellation. Phong Tessellation is silhouette-based, and since the silhouette is considered one of the most important properties, the main aim is to determine how it performs compared to the other three techniques.
Methods. The four techniques are implemented in a real-time application using the modern rendering API Direct3D 11. Data will be gathered from this application to use in several experiments in the context of both performance and image space metrics.
Conclusions. This study has shown that all of the techniques used works in real-time, but with varying results. From the experiments it can be concluded that the best technique to use is Unpopping LoD. It has good performance and provides a good visual result with the least average amount of popping of the compared techniques. The dynamic techniques are not suitable as a substitute to Unpopping LoD, but further research could be conducted to examine how they can be used together, and how the objects themselves can be designed with the dynamic techniques in mind.
Place, publisher, year, edition, pages
2016. , 47 p.
real-time, level of detail, tessellation
IdentifiersURN: urn:nbn:se:bth-12755OAI: oai:DiVA.org:bth-12755DiVA: diva2:945408
Subject / course
DV1478 Bachelor Thesis in Computer Science
DVGSP Game Programming
Lopez Luro, Francisco, Universitetsadjunkt
Goswami, Prashant, Universitetslektor