Feature extraction and statistical analysis of videos for cinemetric applications

Abstract

In this paper, we describe a framework for the extraction of low-level and high level information from movies in order to be used for cinemetric applications. The developed framework analyses the available video content and extracts characteristics related to color, motion, contrast, shot length, tempo, face to frame ratios etc. The extracted information is stored in MPEG 7 AVDP profile format, which is a standard description format that can be imported to related cinemetric applications. We applied the developed framework in a collection of downloaded videos, as well as 3 stereoscopic movies.