Shape-from-focus reconstruction using nonlocal matting Laplacian prior followed by MRF-based refinement. (July 2020)

• Record Type: Journal Article
• Authors: Ma, Zhiqiang; Kim, Dongjoon; Shin, Yeong-Gil
• Is Part Of: Pattern recognition. Volume 103(2020:Jul.)
• Publication Date: 2020-07
• Subjects: Shape from focus -- Depth reconstruction -- Matting Laplacian -- Image denoising -- Markov random field -- Edge-preserving; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2020.107302
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: Depth reconstruction using nonlocal matting Laplacian prior is proposed to obtain a reliable depth image with clear edges and fine details. We propose a MRF-based edge-preserving depth refinement algorithm to denoise the reconstructed depth image and suppress texture-copy artifacts, A closed-form solution is obtained in the depth refinement process. Our method outperforms the existing approaches in terms of robustness, and the ability in preserving edges and fine details while maintaining spatial consistency. Abstract: In this paper, we address the problem of depth recovery from a sequence of multi-focus images, known as shape-from-focus (SFF). The conventional SFF techniques typically exhibit poor performance over textureless regions, and it is difficult to preserve depth edges and fine details while maintaining spatial consistency. Therefore, we propose an SFF depth recovery framework composed of depth reconstruction and refinement processes. We first formulate the depth reconstruction as a maximum a posterior (MAP) estimation problem with the inclusion of matting Laplacian prior. The nonlocal principle is adopted in matting Laplacian matrix construction to preserve depth edges and fine details. As the nonlocal principle breaks the spatial consistency, the reconstructed depth image is spatially inconsistent and suffers from the texture-copy artifacts. To smooth the noise and suppress the texture-copy artifacts, a closed-form edge-preserving depth refinement is proposed, which is formulated as a MAP estimation problem using Markov random fields (MRFs). Experimental results over synthetic and real scene datasets demonstrate the superiority of our algorithm in terms of robustness, and the ability to preserve edges and fine details while maintaining spatial consistency compared to existing approaches.