Atmospheric turbulence is a key factor contributing to data distortion in mid-to-long-range target observation tasks. Neural networks have become a powerful tool for dealing with such problems due to their strong ability to fit nonlinearities in the spatial domain. However, the degradation in data is not confined solely to the spatial domain but is also present in the frequency domain. In recent years, the academic community has come to recognize the significance of frequency domain information within neural networks. There remains a gap in research on how to combine dual-domain information to reconstruct high-quality images in the field of blind turbulence image restoration. Drawing upon the close association between spatial and frequency domain degradation information, we introduce a novel neural network architecture, termed Dual-Domain Removal Turbulence Network (DDRTNet), designed to improve the quality of reconstructed images. DDRTNet incorporates multiscale spatial and frequency domain attention mechanisms, combined with a dual-domain collaborative learning strategy, effectively integrating global and local information to achieve efficient restoration of atmospheric turbulence-degraded images. Experimental findings demonstrate significant advantages in performance for DDRTNet compared to existing methods, validating its effectiveness in the task of blind turbulence image restoration.

Single-pixel imaging (SPI) is a new type of imaging technology that uses a non-scanning single-pixel detector to image objects and has important application prospects and value in many fields. Most of the modulators currently used in SPI systems are digital micromirror device (DMD) modulators, which use a higher frequency for binary modulation than other alternatives. When modulating grayscale information, the modulation frequency is significantly reduced. This paper conducts research on multiple discrete objects in a scene and proposes using binary patterns to locate and image these objects. Compared with the existing methods of using gray patterns to locate and image multiple objects, the method proposed in this paper is more suitable for DMD-type SPI systems and has wider applicability and greater prospects. The principle of the proposed method is introduced, and the effectiveness of the method is experimentally verified. The experimental results show that, compared to traditional SPI methods, the number of patterns required by the proposed method is reduced by more than 85%.

Geometric moment patterns are selected as non-orthogonal gray patterns used for single-pixel centroid detection. By using geometric moment patterns for structured illumination, the reflected light intensities collected by a single-pixel detector correspond to the zero-order and first-order geometric moments, which are leveraged to reconstruct and track the centroid of moving objects. However, the frame rate of the single-pixel tracking system is limited because geometric moment patterns are grayscale while the digital micromirror device (DMD) performs binary modulation much faster than grayscale modulation. Although spatial dithering addressed the problem by binarizing the patterns, the accuracy of tracking a small-sized moving object will decrease by the extra dithering error. Therefore, we propose a binary defocus method to improve the accuracy of tracking a small-sized moving object while ensuring a high frame rate. The key to the proposed method is to simulate grayscale geometric moment patterns by defocusing projection binary geometric moment patterns. The experimental results show that the proposed defocus method is still available for tracking small-sized moving objects even when the focused method fails, and ∼48% lower Centroid Estimation Error (CEE) than that of the focused approach when the frame rate is ∼7400 Hz. Meanwhile, our works also provide relationships between CEE of focused and defocus methods varies with the size of the moving object.