Creating and sharing knowledge for telecommunications

Project: Deep learning-based Point Cloud Representation

Acronym: Deep-PCR
Main Objective:
The scope of this project is the design, development, implementation and assessment of a DL-based static PC compressed representation targeting both decoding and enhancement consumptions. This will pioneer DL-based solutions that simultaneously accommodate efficient fidelity-to-original PC coding, with compression efficiency improvements over available solutions, and effective enhanced-from-original PC processing using the same compressed domain stream. The expected outcomes are DL-based PC coding and enhancement solutions with the following features: i) efficient PC geometry and color codecs, with and without scalability capabilities, providing the highest fidelity to the original PC for the available rate; and ii) effective PC geometry and color denoising and super-resolution processors, generating enhanced PCs using the same compressed domain stream as the fidelity codec. These outputs should offer the best user experience for multiple application domains, considering different fidelity-to-original requirements. The project will advance the knowledge in the domain and disseminate its achievements with top international conference and journal publications.
Reference: PTDC/EEI-COM/1125/2021
Funding: FCT
Approval Date: 15-10-2021
Start Date: 01-01-2022
End Date: 31-12-2024
Team: Fernando Manuel Bernardo Pereira, Nuno Miguel Morais Rodrigues, André Filipe Rodrigues Guarda, Abdelrahman Seleem Mohamed Seleem, Mohammadreza Ghafari
Groups: Multimedia Signal Processing – Lx, Multimedia Signal Processing – Lr
Partners: IT
Local Coordinator: Fernando Manuel Bernardo Pereira

Associated Publications