Ghost imaging method based on blind noise estimation and iterative filtering denoising

This paper introduces an adaptive blind noise dynamic filtering for ghost imaging reconstruction (ABNDF-GIR), a novel method of optimizing ghost imaging data with a limited number of measurements, significantly improving image quality and peak signal-to-noise ratio (PSNR). To address the challenges of noise and undersampling, we first enhance the stability of the measurement matrix by using pseudoinversion and a unit matrix, and calculate correction terms for bucket detector observations to optimize the reconstruction process. A balanced all-one column vector is used as the initial value to accelerate convergence. For iterative computation, we propose a novel filtering and denoising technique, the adaptive denoising window-based guided filtering with BM3D (ADW-BG), which integrates blind noise estimation, block matching and 3D filtering, and guided filtering. This dynamic filtering method effectively preserves important details during each iteration, and can achieve high-quality target reconstruction even with fewer measurements. Extensive simulations and experimental results verify that our method is significantly superior to traditional filtering methods and various compressiv sensing algorithms, especially in edge preservation and texture detail enhancement. The proposed technique provides a key technical advancement for the application of ghost imaging in fields such as remote sensing and medical imaging, showing significant advantages in real-world imaging scenarios.