Efficient space-time sampling with pixel-wise coded exposure for high-speed imaging.

Area of Science:

• Computer Vision
• Image Processing
• Computational Imaging

Background:

• Cameras face a fundamental trade-off between spatial and temporal resolution.
• High-speed video cameras typically have lower spatial resolution than digital still cameras.
• Overcoming this trade-off often requires significant hardware cost increases.

Purpose of the Study:

• To propose techniques for sampling, representing, and reconstructing the space-time volume to overcome the spatial-temporal resolution trade-off.
• To develop a method that reconstructs high-resolution video from a single coded image.
• To adhere to practical hardware constraints for implementation on current image sensors.

Main Methods:

• Learning an overcomplete dictionary on video patches for sparse video representation.
• Developing a sampling function compatible with CMOS image sensor architectures.
• Implementing a prototype imaging system with pixel-wise coded exposure control using a Liquid Crystal on Silicon (LCoS) device.
• Evaluating system characteristics like field of view and modulation transfer function.

Main Results:

• The proposed method effectively reconstructs video from a single coded image while maintaining high spatial resolution.
• Simulations and experiments demonstrate the effectiveness of the sampling function and sparse representation.
• The prototype system shows promising performance across various scenes.

Conclusions:

• The developed techniques successfully overcome the spatial-temporal resolution trade-off in imaging systems.
• The approach offers a cost-effective solution for high-resolution video capture.
• Future implementation on CMOS image sensors with modified control units is feasible.