1. Introduction

The 2023 Optica Imaging Congress [1], held at the Park Plaza Hotel in Boston, Massachusetts, USA, brought together researchers with wide ranging expertise in optics and photonics. The event hosted multiple co-located topical meetings including the Computational Optical Sensing and Imaging (COSI) meeting [2], which traces its origins to a 1998 OSA topical meeting on Integrated Image Gathering and Processing [2]. Since its founding, the COSI topical meeting and members of the COSI community have sought to tackle longstanding imaging challenges through the concurrent design and joint optimization of the electro-optic chain and signal processing [3]. Through the years, COSI’s sphere of influence has expanded from wavefront coding and phase retrieval to accommodate a wide range of topics including but not limited to compressed sensing, super-resolution, lensless imaging, non-line-of-sight imaging, imaging through scattering media, neuromorphic sensing and quantum sensing.

The 2023 edition of the COSI topical meeting showcased the latest innovations in imaging science, emphasizing synergistic activities in optics, signal processing and machine learning. The technical program included a diverse collection of keynote, invited, and contributed talks from around the world, spanning topics that range from theoretical advances in photonics to imaging applications in medicine, defense and industry.

As in previous years [4,5], Optica solicited submissions for a feature issue, to follow the COSI topical meeting. The current feature issue comprises 30 articles in total, with 25 articles in Optics Express and 5 articles in Applied Optics. The following sections provide a summary of each contribution as it relates to recurring themes within the feature issue.

2. Non line-of-sight imaging

The nascent field of non line-of-sight (NLoS) imaging continues to be a hot topic within computational imaging. Sultan et al. [6] propose a variation of methods based on the spatial Wigner function to improve upon existing NLOS reconstruction techniques in the presence of an occluder, thus relying on the shadows, as well as the light fields, present in the scene. Hashemi et al. [7] look at the problem of identifying “clutter”, i.e. unwanted signals in the case of passive NLOS imaging, e.g., in the case in which the scene is illuminated by a simple LED screen. Liu et al. [8] also work with passive NLOS and specifically with polarised long-wavelength infrared light and show the ability to identify people at more than 6m distance from the last scattering surface, significantly further than in previous approaches. Han et al. [9] instead devise a method for improving NLOS image resolution that in turn depends on the temporal resolution in the case of active illumination. Their approach employs temporal ghost imaging techniques to improve the temporal resolution of their detection from ns to 10 ps. Milanese et al. [10] report progress on their LinoSPAD linear array for single photon counting with applications in NLoS imaging and LIDAR.

3. Imaging through scattering media

The design and optimization of modern imaging systems is predicated on the assumption that light is transported along uninterrupted ray paths originating at the illumination source, bouncing off the object and terminating at the image sensor. This assumption contrasts the myriad ways in which light is transported within scattering media such as fog, smoke, turbid waters, and biological media. The challenging problem of imaging objects embedded within scattering media is gaining increased interest within the COSI topical meeting, and the COSI community. The topical meeting featured two sessions on Imaging through scattering media. This feature issue includes an article by Lin et al. [11] who propose the use of a dynamic learning approach for polarization imaging through turbid media, by adaptively blending features from different polarization components to accommodate a wide variety of scattering scenarios.

4. Advances in lensless imaging and computational microscopy

The use of deep learning models to solve inverse problems in imaging science has attracted significant interest in recent times. The COSI feature issue includes four articles that reflect this development. Lee et al. [12] present a self-supervised learning for performing functional tasks with mask-based lensless cameras. They used contrastive representation learning has been used to extract semantic information from mask-modulated measurements with implicit priors. Rogalski et al. [13] introduce a physics-driven neural network for twin-image removal in digital in-line holographic microscopy, which does not require extensive experimental training data. Chen et al. [14] describe a deep-learning approach for structured illumination microscopy, where low-contrast simulation data was used in training to provide reliable super-resolution reconstruction under suboptimal illuminations. Hu et al. [15] put forth a deep-learning framework for super-resolution imaging of metallic nanostructures, expected to find applications in nano-scale fabrication and low-magnification scanning electron microscopy. Gong et al. [16] describe a data-driven approach to realizing full color wide-field imaging though a multimode fiber, with applications in clinical medicine and industrial monitoring.

The feature issue includes two articles that describe specific advances in microscopy. Hsu et al. [17] report on a volume holographic grating for Airy light sheet microscopes, demonstrating a low-cost alternative to spatial modulator-based beam shaping. Gupta et al. [18] report on an axicon probe-based common path optical coherence tomography system and the tissue characterization results using the proposed device.

5. Advances in compressed sensing

Compressed Sensing remains an active topic of research within COSI. The feature issue includes three articles that highlight the latest advances in compressed sensing. Zhang et al. [19] report the development of a single-pixel sensor operable to detect nanosecond transients with a slow detector. Guzman et al. [20] describe a deep learning framework for reconstituting high-speed motion/dynamics from compressive measurements acquired under pulsed infrared illumination. Zhang et al. [21] report differential compressed sensing ghost imaging, demonstrating improved experimental robustness against measurement errors in single pixel imaging systems. Pei et al. [22] describe a data driven approach to reconstituting video frames of ultra-fast dynamics using compressed ultrafast photography.

6. Advances in ptychographic imaging

Ptychography has remained a constant fixture in recent editions of the COSI topical meeting. The feature issue includes two articles that highlight the latest advances in conventional and Fourier ptychography. Eschen et al. [23] demonstrate the first structured illumination tabletop ptychography setup, enabling high resolution extreme ultraviolet microscopy. Divitt et al. [24] show a promising version of Fourier ptychography where patterned illumination is used to reconstruct distant and moving objects at high resolution.

7. Advances in metrology

The wide range of metrology applications examined in the feature issue demonstrate the increasing influence of COSI within optical science and engineering. Zhang et al. [25] describe a computational technique for ascertaining the size of an X-ray focal spot using an approach that is reminiscent of knife edge measurement of Gaussian beams. Tkachuk et al. [26] report a strategy for measuring the amplitude and phase of tightly focused free-space optical beams originating from photonic devices, with sub-wavelength accuracy and polarization sensitivity. Yue et al. [27] report the development of a machine-learned attention mechanism that can identify the topological charge of orbital angular momentum beams under the corrupting influence of turbulence. Kang et al. [28] describe a high-precision metrology technique for Integrated circuit and MEMS device inspection, which addresses shortcomings of state-of-the-art methods in composite fringe projection profilometry. Yang et al. [29] propose a data driven solution to the measurement of large-scale complex deformations using Digital Image Correlation. Staffas et al. [30] present a comparative assessment of the range accuracy and noise robustness of state-of-the-art LIDAR technologies (such as frequency modulated continuous wave and time of flight) in the light starved regime, using a superconducting nanowire single photon detector. Glandon et al. [31] report the development of a 3D LIDAR system operable to identify human subjects from their movement at increased standoff. Jeong et al. [32] describe a near-infrared eye tracker with a dedicated patterned mirror allowing for self-calibration and gaze extraction.

8. Spectroscopy and spectral imaging

The feature issue includes three articles that highlight advances in spectroscopy and spectral imaging. Sun et al. [33] describe the design of a sub-wavelength grating structure operable as an ultra-compact long wave infrared spectral filter with tunable characteristics. Singh et al. [34] report the characterisation of various explosive materials by means of terahertz time-domain spectroscopy. Gallastegi et al. [35] describe a hyperspectral imaging technique that exploits the wavelength dependence of atmospheric absorption for passive ranging using only thermal emissions.