Carleton University - School of Computer Science Honours Project
Winter 2020
Cut Detection Survey and Analysis for Long Form Video Game Streams
Jaime Herzog
SCS Honours Project Image
ABSTRACT
Shot Boundary Detection, or simply Shot Detection, is a fundamental part of research in the broader field of video analytics, used for essential video analysis tasks such as video indexing and content-based video retrieval. For professional video game live streamers who create hours of continuous content with significant downtime, identifying cuts in their streams is an important first step for automatically generating condensed stream highlights. In this project, I have summarized the nomenclature and methodologies established in the academic canon for Cut Detection research, implemented a sample of the most common approaches, Colour Histogram comparison and Edge Change Ratio using Canny Edge Detection, and compared their effectiveness when used on video game live stream content, as well as establishing methodological challenges to this new content form.