Compressed Sensing-Based Correlation Imaging Program

Compressed sensing technology has revolutionized computational ghost imaging applications by enabling high-quality image reconstruction at extremely low sampling rates, overcoming limitations of traditional imaging approaches.

The core principle involves performing compressed sampling of target scenes using random measurement matrices, followed by reconstruction of original images from limited measurements through optimization algorithms. This approach significantly reduces data acquisition requirements while maintaining image quality. Key implementation includes generating random measurement matrices using functions like randn() in MATLAB and applying optimization solvers such as l1-minimization algorithms.

In correlation imaging systems, computational ghost imaging relies on structured light modulation and signal correlation operations. The system first projects a series of random speckle patterns generated by spatial light modulators onto targets, then collects reflected or transmitted intensity signals using single-pixel detectors. Code implementation typically involves controlling DMD (Digital Micromirror Device) patterns and synchronizing detector readings through hardware interfaces.

Reconstruction algorithms commonly employ optimization methods like L1-norm minimization combined with sparsity priors. These algorithms effectively overcome information loss from under-sampling by solving convex optimization problems using packages like CVX or SPGL1, enabling accurate target reconstruction from minimal measurement data.

This technology demonstrates unique advantages in low-light imaging and scattering media penetration applications, providing novel solutions for scenarios where traditional optical imaging approaches face limitations. Practical implementations often include noise reduction techniques and adaptive sampling strategies to enhance reconstruction quality.