Real-Time Non-photorealistic Color Quantization
Author(s)
Mandel, Maximillian
Date Issued
May 3, 2024
Abstract
A non-photorealistic rendering effect changes an image render through the use of various graphics post-processing algorithms. This is done with the goal of stylizing an image in order to make it appear more visually interesting, diverse, or to highlight specific elements within an image. This thesis describes in detail a series of graphics processing algorithms intended to be ran sequentially in a post-processing pipeline. The proposed stylization of images in real time uses techniques such as edge detection, ordered dithering, and color quantization. This effect, titled real-time color quantization, allows existing renders to take on a new and unique look with the intent of replacing or modifying the core art-style of a real-time application. This thesis focuses on analyzing and developing algorithms that can generate this stylistic effect while being suitable for various real-time applications, leveraging existing knowledge in the field and existing graphics processing techniques. Various individual algorithms and their logics and implementations are proposed. In addition to this, a Unity project has been developed that demonstrates each of the proposed algorithms effects both individually and with each other. By implementing these algorithms in a Unity environment, this thesis provides both a proof of concept and a framework example for how these algorithms might be implemented into an existing project. This Unity demo is used to run various experiments on the algorithms, testing the impact they have on performance in general and determine whether they are suitable for real time. Generally, the proposed algorithms have a noticeable performance impact on the application, the significance of which varies widely depending on the specific device being used to run the application. In general, more powerful devices that are not limited by the CPU will experience a more significant reduction in performance, although the experiments indicated that the performance will never drop below what is considered to be real time.
Major
Computer Science
Honors
Computer Science, 2024
First Reader(s)
Kapfhammer, Gregory
Other Reader(s)
Luman, Douglas J.
Department
Computer and Information Science
Type of Publication
Senior Project Paper
File(s)![Thumbnail Image]()
Name
SeniorThesis.pdf
Size
8.93 MB
Format
Adobe PDF
Checksum (MD5)
41ee7d8e19218a57e31f740e24953866