The aperture-divided optical system is a significant imaging technique that enables real-time imaging with multiple channels. However, an increasing demand for multi-channel optics presents a substantial challenge for current refractive optical systems with complex structures and a narrow wavelength band. In this paper, we propose a modified design method that combines the strengths of an off-axis reflective system and an aperture-divided optical system to achieve high levels of integration and simplified structure. A design concept of integrated optical layout and local detail optimization is proposed here. We present an analysis showing how local sub-channels’ distribution affects the imaging characteristics. An integrated optical system, including a relay group constructed based on the Wassermann–Wolf differential equations and a telescope objective, is built first. The sub-aperture system utilizing distinct local surface regions is gradually established with a close connection. To demonstrate the feasibility and efficiency of the method, an integrated system with an F-number of 1.6 and an entrance pupil of 130 mm is presented with its design strategies. The aperture-divided system illustrates well imaging performance close to the diffraction limit in 3–5 µm at 33 lp/mm. The design strategy we have proposed not only has a broad application to multi-channel imaging but also provides valuable insight into to our knowledge, the new imaging technology.

Obtaining an unobscured starting point for off-axis four-mirror imaging systems is a non-trivial task that traditionally relies heavily on designer expertise and extensive trial-and-error. To address this challenge, we propose an automated framework capable of efficiently constructing initial off-axis four-mirror structures with diverse geometries. An evaluation function is formulated by combining first-order constraints, third-order aberration constraints using nodal aberration theory (NAT), and a "ray-quadrangle" obscuration error function. A hybrid global optimization strategy, named IACO _R -NM, is proposed by integrating the Improved ant colony optimization for continuous domains (IACO _R ) with the Nelder–Mead (NM) simplex method, enabling designer-independent global exploration and rapid obtain unobscured systems. The IACO _R quickly locates an unobscured, coarse global optimum, upon which NM performs refined local optimization. This framework allows designers to efficiently assess the optimization potential of different initial structures and rapidly select suitable geometries for further optimization. Using this method, a compact off-axis four-mirror freeform imaging system is constructed, achieving diffraction-limited performance. Moreover, comparative studies against other global optimizers and gradient-based local methods demonstrate the superior optimization capability of the proposed hybrid algorithm.

Interferometric measurement of large aperture freeform surfaces often requires sub-aperture stitching to achieve full aperture measurement. We propose layout preselection based on the Hessian response of the design surface to two planar translations and rotation about the optical axis. After removing the best-fit sphere, first-order Hessian responses form a scaled matrix that quantifies the weakest constraint on these variables. The resulting sensitivity field guides overlap placement toward regions with stronger constraints while weakly sensitive regions are covered within individual subapertures. Responses from adjacent overlaps are then assembled to rank feasible layouts by their weakest collective registration constraint. Global orientation selection and subaperture center adjustment refine the retained layouts. The method was evaluated on 180 design surfaces from the Zernike, XY polynomial, and Q-type families. After registration with three degrees of freedom, adjusted layouts yielded median translation and rotation errors of 5.42 $\mathsf{\mu }\mathrm{m}$ and 0.228 mrad, respectively, with a median stitching error root mean square (RMS) of 23.95 nm. All three interquartile ranges were narrower than for regular layouts. Using the recovered planar estimates to initialize registration with six degrees of freedom (DoF) reduced median subsequent iteration counts. These results support Hessian-based overlap planning for improved planar registration and initialization of full aperture reconstruction.

Stacked microlens array scanning imaging systems inevitably suffer from space-variant point spread functions caused by diffraction effects, resulting in severe image blur that is difficult for traditional deconvolution methods to effectively restore. To address this, we employ a sparse matrix to efficiently represent the image degradation process and embed it into an alternating direction method of multipliers optimization framework to establish a space-variant image deblurring model. Building upon this, we design a deep unfolding network, termed the multiple space-variant kernel network (MSVKNet), and compare its restoration performance against a method based on total variation (TV) priors and conjugate gradient (CG) iteration (TV-CG). Simulation results under various system parameters indicate that MSVKNet achieves performance comparable to or better than the TV-CG method in most cases, with an inference speedup of nearly three orders of magnitude. Furthermore, experiments using a dual microlens array imaging system operating in the visible spectrum validate the accuracy and practical utility of the proposed method.