In this paper, a saliency estimation technique for omni-directional images is presented. Traditional approaches for estimating 360° image saliency rely on the exploitation of low and high-level image features, along with auxiliary data, such as head movement or eye-gazes. However, the image content plays an important role in saliency estimation. Based on this evidence, in the proposed method low-level features are combined with the detection of human faces. In this way it is possible to refine the saliency estimation based on the low-level features by assigning a larger weight to the regions containing faces. Experimental results on 360° image dataset show the effectiveness of the proposed approach.

Mazumdar, P., Arru, G., Carli, M., & Battisti, F. (2019). Face-aware saliency estimation model for 360° images. In European Signal Processing Conference (pp.1-5). European Signal Processing Conference, EUSIPCO [10.23919/EUSIPCO.2019.8902556].

Face-aware saliency estimation model for 360° images

Mazumdar P.;Arru G.;Carli M.;Battisti F.
2019

Abstract

In this paper, a saliency estimation technique for omni-directional images is presented. Traditional approaches for estimating 360° image saliency rely on the exploitation of low and high-level image features, along with auxiliary data, such as head movement or eye-gazes. However, the image content plays an important role in saliency estimation. Based on this evidence, in the proposed method low-level features are combined with the detection of human faces. In this way it is possible to refine the saliency estimation based on the low-level features by assigning a larger weight to the regions containing faces. Experimental results on 360° image dataset show the effectiveness of the proposed approach.
978-9-0827-9703-9
Mazumdar, P., Arru, G., Carli, M., & Battisti, F. (2019). Face-aware saliency estimation model for 360° images. In European Signal Processing Conference (pp.1-5). European Signal Processing Conference, EUSIPCO [10.23919/EUSIPCO.2019.8902556].
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11590/364036
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact