Model and Dictionary guided Face Inpainting in the Wild

Reuben Farrugia 1 Christine Guillemot 2
2 Sirocco - Analysis representation, compression and communication of visual data
Inria Rennes – Bretagne Atlantique , IRISA-D5 - SIGNAUX ET IMAGES NUMÉRIQUES, ROBOTIQUE
Abstract : This work presents a method that can be used to inpaint occluded facial regions with unconstrained pose and orientation. This approach rst warps the facial region onto a reference model to synthe- size a frontal view. A modi ed Robust Principal Component Analysis (RPCA) approach is then used to suppress warping errors. It then uses a novel local patch-based face inpainting algorithm which hallucinates missing pixels using a dictionary of face images which are pre-aligned to the same reference model. The hallucinated region is then warped back onto the original image to restore missing pixels. Experimental results on synthetic occlusions demonstrate that the pro- posed face inpainting method has the best performance achieving PSNR gains of up to 0.74dB over the second-best method. Moreover, experi- ments on the COFW dataset and a number of real-world images show that the proposed method successfully restores occluded facial regions in the wild even for CCTV quality images.
Type de document :
Communication dans un congrès
ACCV workshop on New Trends in Image Restoration and Enhancement, Nov 2016, Taipei, Taiwan. pp.17
Liste complète des métadonnées

Littérature citée [40 références]  Voir  Masquer  Télécharger

https://hal.inria.fr/hal-01388971
Contributeur : Christine Guillemot <>
Soumis le : mardi 15 novembre 2016 - 10:34:11
Dernière modification le : mardi 16 janvier 2018 - 15:54:20
Document(s) archivé(s) le : jeudi 16 mars 2017 - 11:57:10

Fichier

W10-01.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-01388971, version 1

Citation

Reuben Farrugia, Christine Guillemot. Model and Dictionary guided Face Inpainting in the Wild. ACCV workshop on New Trends in Image Restoration and Enhancement, Nov 2016, Taipei, Taiwan. pp.17. 〈hal-01388971〉

Partager

Métriques

Consultations de la notice

153

Téléchargements de fichiers

170