Authors: Sarra Babahenini; Fella Charif; Foudil Cherif; Abdelmalik Taleb-Ahmed; Yassine Ruichek
Addresses: LESIA Laboratory, Department of Computer Science, University of Biskra, 07000 Biskra, Algeria ' LESIA Laboratory, Department of Electronic, University of Ouargla, 30000 Ouargla, Algeria ' LESIA Laboratory, Department of Computer Science, University of Biskra, 07000 Biskra, Algeria ' IEMN Laboratory, Université Polytechnique Hauts-de-France, 59300 Valenciennes, France ' CIAD (Connaissances et Intelligence Artificielle Distribuées) Laboratory, University of Technology Belfort-Montbéliard, 90010 Belfort, France
Abstract: In this paper, we introduce a novel and efficient algorithm based on saliency detection methods, our main contribution is a new manner to calculate the weight map by normalising the saliency values obtained from the input images, which makes it possible to differentiate the focused and defocused regions. We have experiment three techniques of computing the weight map using contourlet transform and low-rank and structured sparse matrix decomposition (LSMD) model. The performance of the proposed model is compared with that of the state-of-the-art multi-focus fusion methods by using fusion metrics. Our evaluation of a series of dataset image demonstrate that the proposed method provides an improvement both visual quality and objective assessment compared to existing methods.
Keywords: human vision system; visual saliency detection; saliency map; multi-focus; image fusion; weight map; contourlet transform; matrix decomposition.
International Journal of Signal and Imaging Systems Engineering, 2021 Vol.12 No.3, pp.81 - 92
Received: 23 May 2020
Accepted: 18 May 2021
Published online: 04 Oct 2021 *