Phase retrieval is a critical technique for reconstructing optical field information. By reconstructing the phase of the light field from intensity measurements, this approach effectively address phase distortions induced by atmospheric turbulence and optical system imperfections, substantially enhancing the quality and multidimensional information content of remote sensing imagery, thereby providing critical support for disaster monitoring, resource exploration, and climate research. In recent years, the rapid development of deep learning has offered unprecedented support for computational optical imaging. However, traditional deep learning techniques in phase retrieval often rely on large sample datasets and deep network models, which pose numerous challenges in practical applications. To address this issue, this paper proposes a novel lightweight network, StarNet. Results of numerical simulations demonstrate that StarNet exhibits superior performance in handling higher-order modes and arbitrary intensity optical fields under random phase perturbations. It can effectively extract phase distortion information under few-shot conditions and accurately recover the intensity of the optical field. This few-shot phase-retrieval algorithm not only paves the way for practical advancements in computational imaging but also offers new approaches into data-scarce computational problems, such as remote sensing imaging and signal processing.

The scattering of light induced by hazy conditions severely degrades imaging quality, leading to reduced contrast, loss of detail, and blurring. These issues present severe challenges to critical fields such as autonomous driving, remote sensing, and object detection. Traditional Stokes vector-based dehazing algorithms are particularly susceptible to noise interference in polarized images, which considerably hampers their dehazing performance under dense hazy conditions. To address these limitations, this paper proposes a polarization-based dehazing approach that integrates the PBM3D denoising algorithm with the Stokes vector. This method effectively separates scattered light from target reflected light while suppressing noise and enhancing image details and clarity. Experimental results demonstrate that in real-world dense hazy conditions, the proposed method significantly outperforms traditional Stokes vector-based dehazing algorithms. This study not only enhances imaging quality in hazy conditions but also lays a foundation for future applications in extreme hazy conditions across domains such as autonomous driving and environmental monitoring.

Haze conditions degrade image quality via atmospheric scattering and absorption, posing challenges for optical imaging applications. In recent years, deep learning has emerged as an effective method for dehazing images. However, data-driven deep learning dehazing methods typically require large amounts of labeled training data and offer limited interpretability. In this paper, we propose a physics-guided dehazing method based on polarization imaging. By integrating polarization imaging with the physical model, the proposed method enables neural network training using only a set of hazy images captured at different polarization angles, thereby reducing the reliance on labeled training data. Experimental results show that the proposed method significantly outperforms commonly used deep learning dehazing methods in terms of contrast and mean gradient, while exhibiting strong generalization and physical interpretability. By integrating physical model and polarization imaging into deep learning, this method overcomes the limitations of traditional deep learning dehazing methods and paves the way for optical imaging in haze conditions.