The Enhanced Video Coding (EVC) workgroup of the Moving Picture, Audio and Data Coding by Artificial Intelligence (MPAI) organization aims at enhancing traditional video codecs by improving or replacing traditional encoding tools with AI-based counterparts. In this work, we explore enhancing MPEG Essential Video Coding (EVC) intra prediction with a learnable predictor: we recast the problem as a hole inpainting task that we tackle via masked convolutions. Our experiments in standard test conditions show BD-rate reductions in excess of 6% over the EVC baseline profile reference with some sequences in excess of 12%.
A Learnable EVC Intra Predictor Using Masked Convolutions
Gabriele SpadaroFirst
;Attilio Fiandrotti
Last
2023-01-01
Abstract
The Enhanced Video Coding (EVC) workgroup of the Moving Picture, Audio and Data Coding by Artificial Intelligence (MPAI) organization aims at enhancing traditional video codecs by improving or replacing traditional encoding tools with AI-based counterparts. In this work, we explore enhancing MPEG Essential Video Coding (EVC) intra prediction with a learnable predictor: we recast the problem as a hole inpainting task that we tackle via masked convolutions. Our experiments in standard test conditions show BD-rate reductions in excess of 6% over the EVC baseline profile reference with some sequences in excess of 12%.
File in questo prodotto:
| File | Dimensione | Formato | |
|---|---|---|---|
|
ICIAP_23_EVC_Intra.pdf
Accesso aperto
Descrizione: Versione finale accettata per pubblicazione
Tipo di file:
PREPRINT (PRIMA BOZZA)
Dimensione
410.91 kB
Formato
Adobe PDF
|
410.91 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
