Professional quality videos of live staged performances are created by recording them fromdifferent appropriate viewpoints. These are then edited together to portray an eloquent storyreplete with the ability to draw out the intended emotion from the viewers. Creating such competentvideos typically requires a team of skilled camera operators to capture the scene frommultiple viewpoints. In this thesis, we explore an alternative approach where we automaticallycompute camera movements in post-production using specially designed computer visionmethods.A high resolution static camera replaces the plural camera crew and their efficient cameramovements are then simulated by virtually panning - tilting - zooming within the originalrecordings. We show that multiple virtual cameras can be simulated by choosing different trajectoriesof cropping windows inside the original recording. One of the key novelties of thiswork is an optimization framework for computing the virtual camera trajectories using the informationextracted from the original video based on computer vision techniques.The actors present on stage are considered as the most important elements of the scene.For the task of localizing and naming actors, we introduce generative models for learning viewindependent person and costume specific detectors from a set of labeled examples. We explainhow to learn the models from a small number of labeled keyframes or video tracks, and how todetect novel appearances of the actors in a maximum likelihood framework. We demonstratethat such actor specific models can accurately localize actors despite changes in view point andocclusions, and significantly improve the detection recall rates over generic detectors.The thesis then proposes an offline algorithm for tracking objects and actors in long videosequences using these actor specific models. Detections are first performed to independentlyselect candidate locations of the actor/object in each frame of the video. The candidate detectionsare then combined into smooth trajectories by minimizing a cost function accounting forfalse detections and occlusions.Using the actor tracks, we then describe a method for automatically generating multipleclips suitable for video editing by simulating pan-tilt-zoom camera movements within theframe of a single static camera. Our method requires only minimal user input to define thesubject matter of each sub-clip. The composition of each sub-clip is automatically computedin a novel convex optimization framework. Our approach encodes several common cinematographicpractices into a single convex cost function minimization problem, resulting inaesthetically-pleasing sub-clips which can easily be edited together using off-the-shelf multiclipvideo editing software.The proposed methods have been tested and validated on a challenging corpus of theatrerecordings. They open the way to novel applications of computer vision methods for costeffectivevideo production of live performances including, but not restricted to, theatre, musicand opera.

EXPLOR_STEP=$WICRI_ROOT/Wicri/Musique/explor/OperaV1/Data/Main/Exploration

HfdSelect -h $EXPLOR_STEP/biblio.hfd -nk 000263 | SxmlIndent | more

HfdSelect -h $EXPLOR_AREA/Data/Main/Exploration/biblio.hfd -nk 000263 | SxmlIndent | more

{{Explor lien
   |wiki=    Wicri/Musique
   |area=    OperaV1
   |flux=    Main
   |étape=   Exploration
   |type=    RBID
   |clé=     Hal:tel-01119207
   |texte=   Automatic Rush Generation with Application to Theatre Performances
}}