.png){kind=icon}
.png){kind=icon}
Return
Scene-adaptive coded aperture imaging
.png){kind=icon}
PRE
AI
DOI：10.1007/s11042-017-5520-1.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Coded aperture imaging systems have recently shown great success in recovering scene depth and extending the depth-of-field. Tremendous efforts have been focused on designing good coded patterns. The ideal pattern, however, would have to serve two conflicting purposes: 1) it should be broadband to ensure robust deconvolution and 2) it should have sufficient zero-crossings for a high depth discrepancy. This paper presents a simple but effective scene-adaptive coded aperture solution to bridge this gap. We observe that natural scenes often exhibit a dominant edge direction that can be caused by structure, texture, shading, etc. We therefore apply depth-discriminative codes along the dominant direction and broadband codes along its orthogonal direction. To physically implement this scheme, we adopt a two-shot scheme. Firstly, we capture a scene image with a pinhole aperture and analyze image content to determine the dominant direction. Secondly, we apply the proposed coding scheme using the programmable Liquid Crystal on Silicon (LCoS). Experiments on a broad range of synthetic and real scenes show that our technique is able to combine advantages of the state-of-the-art patterns for recovering better quality depth map and all-focus images.
Keywords:
Computational imaging
Coded-aperture
Depth-from-defocus
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
3
Papers:
1.9W
Citations:
3.2W