Application of deep-learning to polarization imaging technology for image restoration has led to many technological breakthroughs, especially in underwater image recovery and recognition. In this work, a four-input deep learning model with the Polarimetric Residual Dense Network is proposed for underwater image recovery. The diverse polarization component images are trained and tested in different processes in the network for the recognition and dehazing by considering the physical model of polarization dehazing imaging. Our study reveals that the proposed method can efficiently recover the hazed images, and provide good performance for improving the quality of image restoration even in a high-turbidity complex underwater environment.

Underwater ghost imaging based on deep learning can effectively reduce the influence of forward scattering and back scattering of water. With the help of data-driven methods, high-quality results can be reconstructed. However, the training of the underwater ghost imaging requires enormous paired underwater datasets, which are difficult to obtain directly. Although the Cycle-GAN method solves the problem to some extent, the blurring degree of the fuzzy class of the paired underwater datasets generated by Cycle-GAN is relatively unitary. To solve this problem, a few-shot underwater image generative network method is proposed. Utilizing the proposed few-shot learning image generative method, the generated paired underwater datasets are better than those obtained by the Cycle-GAN method, especially under the condition of few real underwater datasets. In addition, to reconstruct high-quality results, an underwater deblurring ghost imaging method is proposed. The reconstruction method consists of two parts: reconstruction and deblurring. The experimental and simulation results show that the proposed reconstruction method has better performance in deblurring at a low sampling rate, compared with existing underwater ghost imaging methods based on deep learning. The proposed reconstruction method can effectively increase the clarity degree of the underwater reconstruction target at a low sampling rate and promotes the further applications of underwater ghost imaging.

Image dehazing has always been one of the main areas of research in image processing. The traditional dark channel prior algorithm (DCP) has some shortcomings, such as incomplete fog removal and excessively dark images. In order to obtain haze-free images with high quality, a hybrid dark channel prior (HDCP) algorithm is proposed in this paper. HDCP first employs Retinex to remove the interference of the illumination component. The variant genetic algorithm (VGA) is then used to obtain the guidance image required by the guided filter to optimize the atmospheric transmittance. Finally, the modified dark channel prior algorithm is used to obtain the dehazed image. Compared with three other modified DCP algorithms, HDCP has the best subjective visual effects of haze removal and color fidelity. HDCP also shows superior objective indexes in the mean square error (MSE), peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and information entropy (E) for different haze degrees.

Visual navigation is an important guidance method for industrial automated guided vehicles (AGVs). In the actual guidance, the overexposure environment may be encountered by the AGV lane image, which seriously reduces the accuracy of lane detection. Although the image segmentation method based on deep learning is widely used in lane detection, it cannot solve the problem of overexposure of lane images. At the same time, the requirements of segmentation accuracy and inference speed cannot be met simultaneously by existing segmentation networks. Aiming at the problem of incomplete lane segmentation in an overexposure environment, a lane detection method combining image inpainting and image segmentation is proposed. In this method, the overexposed lane image is repaired and reconstructed by the MAE network, and then the image is input into the image segmentation network for lane segmentation. In addition, a convolutional skip triple attention (CSTA) image segmentation network is proposed. CSTA improves the inference speed of the model under the premise of ensuring high segmentation accuracy. Finally, the lane segmentation performance of the proposed method is evaluated in three image segmentation evaluation metrics (IoU, F₁-score, and PA) and inference time. Experimental results show that the proposed CSTA network has higher segmentation accuracy and faster inference speed.

The combination of a single-photon avalanche diode detector with a high-sensitivity and photon-efficient reconstruction algorithm can realize the reconstruction of target range image from weak light signal conditions. The limited spatial resolution of the detector and the substantial background noise remain significant challenges in the actual detection process, hindering the accuracy of 3D reconstruction techniques. To address this challenge, this paper proposes a denoising super-resolution reconstruction network based on generative adversarial network (GAN) design. Soft thresholding is incorporated into the deep architecture as a nonlinear transformation layer to effectively filter out noise. Moreover, the Unet-based discriminator is introduced to complete the high-precision detail reconstruction. The experimental results show that the proposed network can achieve high-quality super-resolution range imaging. This approach has the potential to enhance the accuracy and quality of long-range imaging in weak light signal conditions, with broad applications in fields such as robotics, autonomous vehicles, and biomedical imaging.

Down-sampling Fourier single-pixel imaging is typically achieved by truncating the Fourier spectrum, where exclusively the low-frequency Fourier coefficients are extracted while discarding the high-frequency components. However, the truncation of the Fourier spectrum can lead to an undesired ringing effect in the reconstructed result. Moreover, the original Fourier single-pixel imaging necessitated grayscale Fourier basis patterns for illumination. This requirement limits imaging speed because digital micromirror devices (DMDs) generate grayscale patterns at a lower refresh rate. In order to solve the above problem, a fast and high-quality Fourier single-pixel imaging reconstruction method is proposed in the paper. In the method, the threshold binarization of the Fourier base pattern is performed online to improve the DMD refresh rate, and the reconstruction quality of Fourier single-pixel imaging at a low-sampling rate is improved by generating an adversarial network. This method enables fast reconstruction of target images with higher quality despite low-sampling rates. Compared with conventional Fourier single-pixel imaging, numerical simulation and experimentation demonstrate the effectiveness of the proposed method. Notably, this method is particularly significant for fast Fourier single-pixel imaging applications.