We present a novel deep neural network (DNN) architecture for compressing an image when a correlated image is available as side information only at the decoder, a special case of the well-known distributed source coding (DSC) problem in information theory. In particular, we consider a pair of stereo images, which generally have high correlation with each other due to overlapping fields of view, and assume that one image of the pair is to be compressed and transmitted, while the other image is available only at the decoder. In the proposed architecture, the encoder maps the input image to a latent space, quantizes the latent representation, and compresses it using entropy coding. The decoder is trained to extract the common information between the input image and the correlated image, using only the latter. The received latent representation and the locally generated common information are passed through a decoder network to obtain an enhanced reconstruction of the input image. The common information provides a succinct representation of the relevant information at the receiver. We train and demonstrate the effectiveness of the proposed approach on the KITTI and Cityscape datasets of stereo image pairs. Our results show that the proposed architecture is capable of exploiting the decoder-only side information, and outperforms previous work on stereo image compression with decoder side information.

Neural Distributed Image Compression Using Common Information / Mital, N.; Ozyilkan, E.; Garjani, A.; Gunduz, D.. - 2022-:(2022), pp. 182-191. (Intervento presentato al convegno 2022 Data Compression Conference, DCC 2022 tenutosi a usa nel 2022) [10.1109/DCC52660.2022.00026].

Neural Distributed Image Compression Using Common Information

Gunduz D.
2022

Abstract

We present a novel deep neural network (DNN) architecture for compressing an image when a correlated image is available as side information only at the decoder, a special case of the well-known distributed source coding (DSC) problem in information theory. In particular, we consider a pair of stereo images, which generally have high correlation with each other due to overlapping fields of view, and assume that one image of the pair is to be compressed and transmitted, while the other image is available only at the decoder. In the proposed architecture, the encoder maps the input image to a latent space, quantizes the latent representation, and compresses it using entropy coding. The decoder is trained to extract the common information between the input image and the correlated image, using only the latter. The received latent representation and the locally generated common information are passed through a decoder network to obtain an enhanced reconstruction of the input image. The common information provides a succinct representation of the relevant information at the receiver. We train and demonstrate the effectiveness of the proposed approach on the KITTI and Cityscape datasets of stereo image pairs. Our results show that the proposed architecture is capable of exploiting the decoder-only side information, and outperforms previous work on stereo image compression with decoder side information.
2022
2022 Data Compression Conference, DCC 2022
usa
2022
2022-
182
191
Mital, N.; Ozyilkan, E.; Garjani, A.; Gunduz, D.
Neural Distributed Image Compression Using Common Information / Mital, N.; Ozyilkan, E.; Garjani, A.; Gunduz, D.. - 2022-:(2022), pp. 182-191. (Intervento presentato al convegno 2022 Data Compression Conference, DCC 2022 tenutosi a usa nel 2022) [10.1109/DCC52660.2022.00026].
File in questo prodotto:
File Dimensione Formato  
Neural_Distributed_Image_Compression_Using_Common_Information.pdf

Accesso riservato

Tipologia: Versione pubblicata dall'editore
Dimensione 909.41 kB
Formato Adobe PDF
909.41 kB Adobe PDF   Visualizza/Apri   Richiedi una copia
2106.11723.pdf

Open access

Tipologia: Versione originale dell'autore proposta per la pubblicazione
Dimensione 8.54 MB
Formato Adobe PDF
8.54 MB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

Licenza Creative Commons
I metadati presenti in IRIS UNIMORE sono rilasciati con licenza Creative Commons CC0 1.0 Universal, mentre i file delle pubblicazioni sono rilasciati con licenza Attribuzione 4.0 Internazionale (CC BY 4.0), salvo diversa indicazione.
In caso di violazione di copyright, contattare Supporto Iris

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11380/1286020
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 4
  • ???jsp.display-item.citation.isi??? 2
social impact