Huajun Feng

Automatic lens design remains challenging under multiple competing constraints. For complex aspheric systems, the increasing number of optimization variables and strong parameter coupling often make the optimization process susceptible to local minima. To address this problem, we propose an automatic aspheric design framework centered on feasibility-aware exploration. The core component is an SVD-guided perturbation strategy that explores low-sensitivity aspheric directions while bounding the induced sag variation, thereby improving local-minimum escape and reducing the risk of ray-tracing failures. To further enhance global exploration, a structured sparse differential evolution (SSDE) strategy is introduced to search within structural and aspheric parameter subspaces. The framework is implemented with approximately orthogonal Qcon surface parameterization and continuous material optimization in a balanced latent space. Experiments on a Cooke triplet spherical system, a four-element aspheric system, and two challenging six-element aspheric systems demonstrate the effectiveness of the proposed method. Compared with patent designs and existing automatic design methods, our framework achieves superior RMS spot-size performance under stricter manufacturability constraints. These results suggest that the proposed framework offers a robust and scalable solution for automatic optical system design, with potential for extension to more challenging and diverse design scenarios.

Compared with traditional periscope telephoto systems, micro catadioptric systems feature a compact structure, long focal length, and large aperture capability, showing potential in portable photography systems. However, they suffer from scarce initial structures, as well as image quality degradation caused by tolerances and central obscuration. To address these challenges, this study proposes a method for generating initial structures of three-element catadioptric systems with Q-bfs aspheres. Given design indicators, the optical power distribution is first determined via Delano diagrams. Subsequently, gradient descent and differential evolution algorithms are applied alternately to optimize the sampled system population. Finally, the optimal individual in the population is selected, and a digital network is adopted to compensate for image quality degradation caused by assembly tolerances and central obscuration. Two design examples are presented to verify the feasibility and effectiveness of the proposed framework.

Coded ptychographic microscopy (CPM) is a lensless imaging technique that captures low-resolution diffraction images at various lateral positions and reconstructs them into high-throughput micrographs. The forward imaging model typically assumes the light source to be a plane wave. While this assumption simplifies calculations, it hinders the miniaturization necessary for maintaining imaging quality. A reduced axial distance from the light source to the sample can cause the plane wave approximation to fail, introducing ambiguities. In this Letter, a compact prototype has been developed, employing an accurate computational method to estimate the spatial position of the light source for the spherical wave model. A translation calibration method was implemented to precisely calculate the object’s displacement. During reconstruction, iterative decoupling of the light source and object mitigates image degradation caused by a reduced source-to-sample distance. Simulations and experimental comparisons demonstrate a 25.5% resolution improvement over typical CPM based on plane waves under the same close-range conditions. This approach has the potential to significantly advance the miniaturization of CPM and enhance its applications in digital pathology and cytometric analysis.

Previous deblurring methods mostly tackle global motion blur due to camera shake but struggle with local motion blur from object movement, facing challenges like the random motion blur locations, data imbalance, directional ambiguity, and positional uncertainty. To fill the vacancy of real-world local motion deblurring, we establish ReLoBlur, the first real-world local motion deblurring dataset. ReLoBlur is captured by a synchronized beam-splitting photographing system and annotated via our developed Local Blur Foreground Mask Generator (LBFMG). To bridge the gap between local and global motion deblurring, we propose a Local Blur-Aware Gated network (LBAG) with gate blocks to focus deblurring on blurred regions, and a Blur-Aware Patch Cropping Strategy (BAPC) to address the data imbalance problem. Acknowledging directional ambiguity and positional uncertainty from shooting errors and non-uniform object motion, we enhance LBAG with LBAGp, guided by center-related distance, and optimized by a symmetric minimization loss. Extensive experiments prove the reliability of the ReLoBlur dataset, and demonstrate that LBAG and LBAGp achieve better local motion deblurring performance compared to state-of-the-art (SOTA) CNN-based deblurring methods.

Against the backdrop of the rapid development of the miniature camera lens manufacturing industry, with an annual output exceeding 4 billion units (including various lenses for mobile phones, automobiles, security, etc.), the design of highly aspherical miniature camera lenses has become increasingly crucial. Traditional optical design methods rely heavily on designer intervention and are optimized step by step through a trial-and-error approach, which consumes a great deal of time and effort. To address this, this paper innovatively proposes a design method based on the modified basin-hopping (MBH) algorithm, enabling the rapid and automatic design of highly aspherical miniature camera lenses without the need for designer intervention throughout the process. Users only need to input design requirements, and the program can automatically generate multiple solutions. Designers simply need to select from the output solutions. In this paper, this method is successfully applied to design two different types of miniature aspheric camera lenses: a 4P (4 Plastic lenses) mobile phone lens with highly aspheric surfaces and an ultra-wide-angle ADAS (advanced driver assistance systems) lens. The designed lenses exhibit excellent imaging quality. This method effectively shortens the development-to-market cycle and reduces the engineering design threshold, fully demonstrating its innovation, effectiveness, and practicality.

As optical systems become increasingly complex, accurate and fast alignment is becoming more critical. Active alignment (AA) techniques dynamically optimize optical positioning using real-time image quality feedback, ensuring better performance and higher production yields. However, in large-scale manufacturing, such as smartphone lens production, existing AA techniques struggle to balance speed and accuracy. In this paper, we propose a pipeline for AA using physics-informed deep learning. Our pipeline consists of two main components: a physics-informed tolerance estimation neural network (TolNet) that estimates tolerances from point spread functions (PSFs) and an optical optimization module that determines the adjustment parameters for AA. TolNet is trained with a strategy that combines data-driven and physics-driven losses, enabling it to produce physically plausible results. TolNet performs tolerance estimation with exceptional speed, completing the process in less than 0.01 seconds, while the optical optimization module is also highly efficient, requiring less than 3 seconds. We conducted extensive experiments to validate the performance of our proposed method, which offers a promising solution to improve the efficiency and accuracy of AA in large-scale manufacturing.

Recently, the joint design of optical systems and downstream algorithms is showing significant potential. However, existing rays-described methods are limited to optimizing geometric degradation, making it difficult to fully represent the optical characteristics of complex, miniaturized lenses constrained by wavefront aberration or diffraction effects. In this work, we introduce a precise optical simulation model, and every operation in pipeline is differentiable. This model employs a novel initial value strategy to enhance the reliability of intersection calculation on high aspherics. Moreover, it utilizes a differential operator to reduce memory consumption during coherent point spread function calculations. To efficiently address various degradation, we design a joint optimization procedure that leverages field information. Guided by a general restoration network, the proposed method not only enhances the image quality, but also successively improves the optical performance across multiple lenses that are already in professional level. This joint optimization pipeline offers innovative insights into the practical design of sophisticated optical systems and post-processing algorithms. The source code will be made publicly available at https://github.com/Zrr-ZJU/Successive-optimization

Differentiable optics integrates optical design with image processing, enabling joint optimization of optical systems and neural networks through task-specific evaluation metrics. However, controlling lens tolerance sensitivity during optimization remains an unresolved challenge, often resulting in significant performance degradation due to manufacturing and assembly tolerances. In this work, we propose a fast sensitivity control method tailored for different optical models. Our approach addresses sensitivity from three key aspects: tolerance sensitivity, uniformity of tolerance degradation, and symmetry of tolerance boundaries. Using the wavefront-focusing method, we enhance optimization stability by compensating for defocus and focusing the evaluation on the optimal image plane. Additionally, the optical path differential method eliminates the need for repeated ray tracing of perturbed systems, significantly reducing computational costs. Experimental results with two advanced smartphone lenses, a wide-angle and a telephoto lens, validate the effectiveness of our method. This work provides a practical solution for desensitization optimization in differentiable optics, offering new possibilities for optical desensitization and joint design in computational imaging tasks.

Image matching is widely used in many practical applications, such as high-precision map reconstruction, UAV localization, and autonomous driving. Since the optical lens directly influences the captured image, which in turn affects image-matching performance, designing optical lenses that are better suited for image-matching tasks under similar budget or size constraints is a highly valuable and practical problem. In this paper, based on the differentiable optics framework, we propose modifying the point spread function (PSF) of the optical lens to be isotropic across different fields of view (FoV) by introducing intuitive PSF-based loss functions. This modification benefits feature detection and description performance. Experimental results show that our method significantly improves the feature-matching performance of the given optical design, even though it does not achieve the highest image quality.

Push-broom imaging systems, including Linear Array (LA) and Time Delay Integration (TDI) sensors, are extensively used in remote sensing for high-resolution image acquisition with continuous spatial coverage. However, platforminduced jitter introduces significant distortions and blurring, especially in TDI systems with multiple integration stages, where the cumulative effects of jitter are more pronounced. Traditional jitter models often struggle to accurately simulate these effects in the presence of measurement noise, hindering effective image restoration. In this paper, we propose a novel jitter-aware restoration framework that addresses these challenges in both LA and TDI push-broom imaging systems. Central to our approach is the introduction of an Equivalent Jitter Model (EJM) that is robust to measurement noise. By averaging time-shifted jitter curves across multiple integration stages, the EJM effectively smooths out noise-induced fluctuations, providing a reliable characterization of jitter effects. Leveraging this model, we develop JARNet, a two-stage restoration network that combines optical flow correction with spatial-frequency residual learning to mitigate geometric distortions and motion blur. We also design a custom data synthesis pipeline to generate realistic jitter-degraded datasets, facilitating effective training of the network. Experimental results on both synthetic LA and TDI datasets demonstrate that JARNet outperforms state-of-the-art methods in terms of PSNR, SSIM, and GMSD metrics. Our framework offers a robust solution for restoring high-quality remote sensing images degraded by jitter, significantly advancing the state-of-the-art in this domain. The source code will be made publicly available upon publication at https://github.com/naturezhanghn/EJM.