Visual quality assessment for motion vectors watermarking in the MPEG‑4 AVC domain
Résumé
Nowadays, to directly watermark a compressed video stream becomes a challenging research topic: low complexity, time saving and interporability are the three main advantages of such an approach. the present paper deals with the MPEG-4 AVC motion vector watermarking and presents the first reference assesment for the visibility of the related artefacts. The experiments follow the most popular types of quality metrics in video processing: pixel difference-based (signal to noise ratio, absolute average difference, peak mean square error, and image fidelity), correlation based (structural content, normalized cross coorlation, coorelation quality, structural similarity metric), and psycho-visual (digital video quality). The numerical results correspond to a corpus of 5 video sequences of about 30 minutes each.
Origine | Fichiers produits par l'(les) auteur(s) |
---|
Loading...