par Seznec, Nicolas;Losfeld, Armand 
;Van Bogaert, Laurie ;Teratani, Mehrdad
Référence Symposium on Information Theory and Signal Processing in the Benelux(44th: 27 May 2024: Delft, The Netherlands), WIC SITB 2024
Publication Publié, 2024-05-27
;Van Bogaert, Laurie ;Teratani, Mehrdad Référence Symposium on Information Theory and Signal Processing in the Benelux(44th: 27 May 2024: Delft, The Netherlands), WIC SITB 2024
Publication Publié, 2024-05-27
Publication dans des actes
| Résumé : | Plenoptic cameras are equipped with a micro-lens array that allows them to capture multiple viewpoints in the form of a plenoptic image, providing 3D information about the scene. This 3D information can be projected back with a glasses-free 3D display, such as the multi-layer display. However, the current method to generate the multi-layer images require first converting the plenoptic image into multi-view images, resulting in data and computational inefficiencies. Our proposed method directly converts plenoptic images to multi-layer images using a CNN, showing improvements in quality metrics, with an average gain of 1.25 dB for the PSNR (for plenoptic 1.0 camera model). |
