Underwater optical imaging is affected by light absorption and backscattering in water, thus yielding low signal-to-noise ratios and limited imaging ranges. This study proposes an image preprocessing method for underwater, time-gated, single-photon avalanche diode (TG-SPAD)-array-based images according to the Retinex theory, and a block-matching and 3D filtering (BM3D) algorithm to address uneven illumination and complex noise issues in small-diameter, light beam, underwater imaging. Specifically, images undergo rapid illumination correction in combination with time-domain transformation. Subsequently, the proposed method employs the BM3D algorithm in conjunction with an adaptive noise-fitted model and an improved Anscombe transform for denoising. The experimental results demonstrate that the proposed method outperforms various existing image preprocessing techniques in both subjective visual assessments and objective evaluation metrics. The proposed method considerably enhances the visual quality of TG-SPAD-array-based images and is well-suited for underwater, single-photon imaging applications and for the optimized processing of high-precision, 3D depth maps.
The division of focal plane polarization cameras is a widely used integrated polarization imaging system, where light intensity crosstalk between micro-polarizer pixels introduces measurement errors. Previous studies have only assessed qualitative crosstalk trends. Here, we model crosstalk as near-field, non-paraxial diffraction of an ideal metal grating. Using the angular spectrum method with spatial-domain discrete sampling, we establish how the F-number and incident angle influence crosstalk intensity distribution. We analyze coherence effects and propose a cosine-based intensity response. Macroscopic experiments validate our microscopic model, revealing polarization parameter deviations and system noise, confirming crosstalk calibration feasibility.
Monocular depth estimation technology has emerged as a critical component across a variety of outdoor applications like robotics, augmented reality, autonomous driving, and 3D reconstruction. Mainstream monocular depth estimation methods consistently face challenges in applications requiring real-time performances, as they exhibit considerable computational complexity, resulting in poor runtime performance. Here, we propose an innovative processing module named MDE-Lite. Based on that, we develop a lightweight yet effective depth estimation network named MBUDepthNet. Besides, we build a training scheme with multiple loss functions. Experimental validation on KITTI dataset demonstrates that our method not only rivals mainstream methods in terms of accuracy but also exhibits superior computational efficiency. Compared to the method using ResNet-18, our method achieves a 22% higher frame rate in terms of frames per second.
In response to the demand for small-size, high-precision, and real-time target distance measurement in platforms such as autonomous vehicles and drones, this paper investigates the multi-focal bionic compound eye (MFBCE) and its associated distance measurement algorithm. MFBCE was designed to integrate multiple lenses with different focal lengths and a CMOS array. Based on this system, a multi-eye distance measurement algorithm based on target detection was proposed. The algorithm derives the application of binocular distance measurement on cameras with different focal lengths, overcoming the limitation of traditional binocular algorithms that only work with identical cameras. By utilizing the multi-scale information obtained from multiple lenses with different focal lengths, the ranging accuracy of the MFBCE is improved. The telephoto lenses, with their narrow field of view, are beneficial for capturing detailed target information, while wide-angle lenses, with their larger field of view, are useful for acquiring information about the target’s environment. Experiments using the least squares method for ranging targets at 100 cm yielded a mean absolute error (MAE) of 1.05, approximately one-half of the binocular distance measurement algorithm. The proposed MFBCE demonstrates significant potential for applications in near-range obstacle avoidance, robotic grasping, and assisted driving.
To address the demand for convenient and high-accuracy calibration of modern polarization imaging systems, this study proposes a scene-based semi-supervised self-calibration algorithm. The algorithm fully exploits the polarization information perceived by the system, constructs three loss functions based on the light intensity, degree of linear polarization (DoLP), and angle of polarization (AoP), and employs a hybrid multi-objective optimization method that combines an evolutionary algorithm and gradient descent to achieve self-calibration. Compared to supervised algorithms, the proposed method uses scene information to match the image and obtain the rotation angle as the ground truth of the AoP variation. Combining the high accuracy of supervised optimization with the advantage of unsupervised optimization that requires no ground truth, the algorithm is convenient to use, achieves high accuracy, and is suitable for various four-channel polarization imaging systems.