Signalisierung für Merkmalbasierte Implizite Netzgrössencodierung Neuronaler Darstellung

Details

Veröffentlichungs-Nr.

EP4794332

Anmeldetag

12. Februar 2025

Veröffentlichung

19. August 2026

Rechtsraum

EP

Abstract

Method and apparatus are provided wherein one or more syntax elements are decoded providing for reconstructing at least one part of an image or a 3D object using at least one hybrid Implicit Neural Representation network, the at least one hybrid Implicit Neural Representation network comprising at least one mapping transformation that maps input coordinates of the image or 3D object to at least two successive grids of features, and an INR network that takes as input, features of the at least one grid of features and outputs values of the image or 3D object at the input coordinates, wherein dimensions of a current grid of the at least two successive grids of features are derived from a first grid of the at least two successive grids of features based on the one or more syntax elements. An extension of at least one dimension of a first grid of the at least two successive grids of features is determined providing an extended dimension of the first grid of the at least two successive grids of features such that dimensions of grids of the at least two successive grids of features are deteremined using the extended dimension of the first grid and input coordinates are mapped to grid positions of the grids of the at least two successive grids of features to reconstruct the at least one part of the image or 3D object.

Anmelder

Firma

InterDigital CE Patent Holdings, SAS

Land

🇫🇷 Frankreich