Computational Abstraction of Films for Quantitave Analysis of Cinematography

Abstract

Currently, film viewers’ options for getting objective information about films before watching them, are limited. Comparisons are even harder to find and often require extensive film knowledge both by the author and the reader. Such comparisons are inherently subjective, therefore they limit the possibilities for scalable and effective statistical analyses. Apart from trailers, information about films cannot reach viewers audibly or visibly, which seems absurd considering the very nature of film. The thesis examines repeatable quantification methods for computationally abstracting films in order to extract informative data for visualizations and further statistical analy- ses. Theoretical background empowered by multidisciplinary approach and design processes are described. Visualizations of analyses are provided and evaluated for their accuracy and efficiency. Throughout the thesis foundations for the future automated quantification player/plugin, are described aiming to facilitate further developments. Theoretical structures of the website which may act as a gateway that collects and provides data for statistical cinematic research are also discussed.