Shuffled rolling shutter for snapshot temporal imaging.

Area of Science:

• Computer Vision
• Image Sensors
• Signal Processing

Background:

• Rolling shutter mechanisms in CMOS detectors capture images sequentially, leading to space-time distortions.
• Traditional rolling shutters limit the ability to reconstruct high-speed dynamic scenes from single captures.

Purpose of the Study:

• To introduce a novel shuffled rolling shutter mechanism for CMOS detectors.
• To enhance the sampling of the space-time datacube for improved video reconstruction.
• To enable the recovery of high-speed videos from single images.

Main Methods:

• Proposing a hardware modification involving pixel shuffling within each scanline of the rolling shutter.
• Utilizing tensor completion and compressive sensing algorithms for video recovery.
• Developing a design methodology for optimal sampling schemes and comparing with random shuffling.
• Conducting simulations and optical emulation experiments.

Main Results:

• The shuffled rolling shutter significantly improves space-time sampling compared to traditional methods.
• Demonstrated successful reconstruction of high-speed videos from single images.
• Validated the effectiveness of optimal sampling schemes over random shuffling.

Conclusions:

• The proposed shuffled rolling shutter is a viable hardware modification for CMOS detectors.
• This technique overcomes the limitations of traditional rolling shutters in capturing dynamic scenes.
• Enables high-speed video reconstruction previously unachievable with standard rolling shutter systems.