MVMFCam: an all-in-focus optical synthetic aperture imaging system based on multi-view multi-focus computational imaging

Area of Science:

• Computational Imaging
• Optical Systems
• Machine Learning for Imaging

Background:

• Traditional imaging systems have a limited depth of field (DoF), resulting in blurred objects outside the focal plane.
• Existing methods for achieving an all-in-focus image require capturing multiple images with adjusted focus, which is unsuitable for dynamic scenes.
• Simultaneous capture of multi-focus images is crucial for dynamic scene all-in-focus imaging.

Purpose of the Study:

• To propose a novel all-in-focus optical synthetic aperture imaging system, MVMFCam, for dynamic scenes.
• To develop an end-to-end neural network (MVMF-Net) for fusing multi-view multi-focus (MVMF) images.
• To demonstrate the capability of MVMFCam for high-quality all-in-focus imaging in dynamic environments.

Main Methods:

• MVMFCam utilizes a synchronized array of nine sub-cameras, each focusing at different depths, to capture MVMF images in a single exposure.
• An end-to-end MVMF image fusion neural network (MVMF-Net) is proposed, comprising image alignment and adaptive fusion phases.
• Image alignment uses a feature transfer matching strategy, followed by fusion using a densely connected network with adaptive weights.

Main Results:

• MVMFCam successfully captures simultaneous multi-view multi-focus images, enabling all-in-focus imaging for dynamic scenes.
• The MVMF-Net effectively aligns and fuses the captured images to produce high-quality all-in-focus results.
• Experimental validation with 10 datasets confirms the system's capability for dynamic scene all-in-focus imaging.

Conclusions:

• MVMFCam overcomes the limitations of traditional imaging systems in capturing dynamic scenes with a limited DoF.
• The proposed MVMF-Net provides an effective solution for fusing MVMF images, achieving superior all-in-focus results.
• This technology holds significant potential for applications in microscopic imaging, close-range photography, and non-destructive testing.