Dual Optical Path Based Adaptive Compressive Sensing Imaging System.

Area of Science:

• Optics
• Image Processing
• Signal Processing

Background:

• Compressive Sensing (CS) is vital for image acquisition but requires high sampling accuracy.
• Noisy measurements from light paths and circuits degrade CS imaging quality.
• Existing CS systems face challenges in distinguishing measurement matrices and handling noise.

Purpose of the Study:

• To propose a novel dual-optical imaging system for simultaneous CS measurement and image acquisition.
• To mitigate noise interference in the CS sampling process.
• To enhance the quality of reconstructed CS images.

Main Methods:

• Utilized digital micromirror devices (DMD) for their bidirectional reflection characteristics.
• Developed a dual-optical system for same-angle CS measurement and image capture.
• Employed deep neural networks, training separate filter and reconstruction networks.
• The filter network addresses measurement noise; the reconstruction network handles image recovery.

Main Results:

• The proposed method effectively filters noise during the CS sampling process.
• Significant improvements in image reconstruction quality were observed across various algorithms.
• The dual-optical system demonstrated superior performance in noisy environments.

Conclusions:

• The dual-optical imaging system with DMD offers a robust solution for noise reduction in CS.
• Deep neural network integration enhances the accuracy and quality of CS image reconstruction.
• This approach advances the practical application of Compressive Sensing in imaging systems.