Compression vidéo fondée sur l'apparence

Karl Schwerdt 1
1 PRIMA - Perception, recognition and integration for observation of activity
GRAVIR - IMAG - Graphisme, Vision et Robotique, Inria Grenoble - Rhône-Alpes
Abstract : This thesis describes a new technique for the compression of digital video data, called Orthonormal Basis Coding (OBC). Algorithms from computer vision, data compression, and pattern recognition are combined to form its three-step encoding scheme. OBC gathers information about image content without using models. Instead, it is based on appearance. Appearance-based techniques use orthonormal basis space representations of objects, usually in eigenspace, and exploit geometrical properties of these object representations. From a sequence of images of an object, each image contributes a point in the orthonormal basis space used. Focusing on one object from a sequence means that we need to normalize our video input data to one principal object. Given the application areas for video compression, a talking head as principal object is a natural choice. We show that OBC is a valid alternative to conventional video compression techniques. If the input video stream is well normalized to the principal object, OBC outperforms conventional compression techniques.
Document type :
Theses
Complete list of metadatas

Cited literature [79 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00004704
Contributor : Thèses Imag <>
Submitted on : Tuesday, February 17, 2004 - 10:03:34 AM
Last modification on : Wednesday, April 11, 2018 - 1:52:11 AM
Long-term archiving on : Friday, April 2, 2010 - 8:21:23 PM

Identifiers

  • HAL Id : tel-00004704, version 1

Collections

INRIA | IMAG | UGA

Citation

Karl Schwerdt. Compression vidéo fondée sur l'apparence. Interface homme-machine [cs.HC]. Institut National Polytechnique de Grenoble - INPG, 2001. Français. ⟨tel-00004704⟩

Share

Metrics

Record views

524

Files downloads

681