eprintid: 703
rev_number: 10
eprint_status: archive
userid: 17
dir: disk0/00/00/07/03
datestamp: 2017-01-15 22:04:19
lastmod: 2017-10-26 11:26:34
status_changed: 2017-01-15 22:04:19
type: report
metadata_visibility: show
creators_name: Badziahin, Dmitry
creators_name: Brosa Planella, Ferran
creators_name: Ferreira, Marina
creators_name: Gazzola, Silvia
creators_name: Herterich, James G.
creators_name: Krupp, Armin
creators_name: Lim, Sean W. X.
creators_name: Miklavcic, Stan
creators_name: Pronko, Rafal
creators_name: Skinner, Jim
creators_name: Bootland, Niall
creators_name: Coban, Sophia
creators_name: Riseth, Asbjørn Nilsen
title: Segmentation and Scene Content in Moving Images
ispublished: pub
subjects: data
studygroups: esgi107
companyname: Aralia Systems Ltd
full_text_status: public
abstract: The problem of scene content in moving images was brought by Aralia. The goal in this study group was to consider two problems. The first was image segmentation and the second is the context of the scene. These problems were explored in different areas, namely the Bayesian approach to image segmentation, shadow detection, shape recognition and background separation.
problem_statement: The goal of the study group is to consider some proposed techniques to improve the information extracted from a scene, as well as exploring new techniques for the problem.

There are two main areas of the problem that Aralia wish to focus on. The first is on image segmentation. Currently, it is performed by maintaining two moving averages of the scene. Aralia propose segmenting the moving images and then coalescing them using additional information such as a 3D depth map, a 2D or 2.5D texture of principal component analysis of illumination. An important part of image segmentation to be considered is shadow detection.

The second area is the context of the scene. Aralia apply scene content rules, typically to prevent impossible situations from arising. This includes people moving at too high a speed or objects changing their classification. Aralia wish to extend context analysis by identifying specific objects within the scene. Identification of the types of static objects in a scene may then be useful as a method to identify or narrow-down the full context of the scene.
date: 2015-03
citation:   Badziahin, Dmitry and Brosa Planella, Ferran and Ferreira, Marina and Gazzola, Silvia and Herterich, James G. and Krupp, Armin and Lim, Sean W. X. and Miklavcic, Stan and Pronko, Rafal and Skinner, Jim and Bootland, Niall and Coban, Sophia and Riseth, Asbjørn Nilsen  (2015) Segmentation and Scene Content in Moving Images.  [Study Group Report]     
document_url: http://miis.maths.ox.ac.uk/miis/703/1/Aralia.pdf